building a new pc and i need the gods to ok my specs.

[paulselhi]

I would go with a minimun 1TB hard drive but 2 TB HDD arre fairly inexpensive

Also check your cooling the 920 will suffer at 4 Ghz on stock cooing, if it even makes it that far

If you go for 12 or even 24 gb ram and use a third party cooler such as Noctua check the clerance spec for the cooler as some ram may not fit if it has large heat spreaders ( the cooler may block some ram slots for large ram chips), this may not be the case if you use the low priced water coolers such as thae one from corsair however though it will cool a 4 ghz well it does not cool any better than a top end air cooler

[imashination]

He wont be able to overclock it to 4ghz if he has 12/24 gigs of ram, so that kinda becomes moot doesnt it?

[paulselhi]

Who says so ? It may be harder to get that o/c ( 4ghz is a breeze with 6 GB RAM) but i don't think it is impossible, just more work getting there adjusting the voltages

This a quote from corsair:

Our 2000 kits run at 2000 just fine in the ASUS RE II with 3 modules. 2000 with 6 modules is pushing it and that is why we don't sell and guarantee 2000 with 6 up.

If you are wanting 12GB guaranteed tested and stable, you need to purchase a 12GB kit. Currently our fastest 12GB kit is the 1600C9. And, 1600 will easily allow you to OC to 4ghz.

at this posting

http://vip.asus.com/forum/view.aspx?board_id=1&model=Rampage+II+Extreme&id=20090609201148174&page=1&SLanguage=en-us



Here is an example of some settings thta may or may not work for the OP

The Rig:
Asus Rampage II Extreme
i7 920
Scythe Mugen 2 Heatsink
12GB Corsair Dominator DDR3 PC3
X-Fi Titanium Fatal1ty Pro
BFG GeForce GTX 295 1792MB
Seasonic M12 700W

The Target:
CPU: 4GHz overclock
RAM: 1700MHz 8-8-8-24 1.6v

http://www.computerlounge.co.nz/forum/Default.aspx?g=posts&m=7671

But i agree it would be difficult, probably a 3.6 - 3.8 ghz O/C would be a good target with 12 GB RAM

[PhoenixCG]

how much ram am i shooting for, my old comp had 3gb which wasn't enough. i've never OCed a processor before so i have no clue about the ram problem, then again i've never built a pc before either, unforteunately though if i want something soon i'll have to build it myself.

the 4ghz. is something i read on an OC forum i found. appariently the 920 can get to 4ghz. with fans, or 4.5ghz on water. no idea yet on how all that works, but i'm sure i can find a tut somewhere.

if you guys have hardware suggestions please list them, the hardware i've listed seem to be a big bang for little bucks. thanks.

[paulselhi]

Also why are you thinking about the rampage MB when the gigaybte x58a -ud3r is half the price

I am running a gigabyte i7 930 with 6 gb ram and run 4.0 Ghz as my stable day to day over clock, i have had 4.2 ghz stable and probably could go higher but i found i needed load line calibration on which may cause unhealthy cpu voltage spikes. Since the MB manual warns against having it on and from what i have gleaned from searching the net i have decided to keep LLC off which means the cpu votage drops as per intel specs

I may well up to 12 Gb and if needed take an O/C hit as many of the sims scemes i am playing with need high amounts of memory and i think a 200-400 Mhz drop in cpu (3.6 -3.8 O/C) is not that drastic

[PhoenixCG]

the rampage is supposed to be an OCers dream, and it supports 24gb of ram. is it not worth the $350?

[paulselhi]

how much ram am i shooting for, my old comp had 3gb which wasn't enough. i've never OCed a processor before so i have no clue about the ram problem, then again i've never built a pc before either, unforteunately though if i want something soon i'll have to build it myself.

the 4ghz. is something i read on an OC forum i found. appariently the 920 can get to 4ghz. with fans, or 4.5ghz on water. no idea yet on how all that works, but i'm sure i can find a tut somewhere.

if you guys have hardware suggestions please list them, the hardware i've listed seem to be a big bang for little bucks. thanks.

I think 4.5 ghz is asking a bit much and would require a high end water cooler ( very expensive)

4- 4.2 ghz is more realistic on a good air cooler (Uk £50+), these air coolers can be very large and with fans can block some of the ram slots that is why the large ram kits with large heat spreaders may limit the slots you can use, water coollers solve this issue by being compact blocks which just cover the cpu and pipe water to a radiator on the case side, corsair do a cheap watercooler prebuilt kit which is good enough for a 4 Ghz overclock

I built mine on a budget and basically followed this spec but added a noctua nh-d14 cooler, a corsair 650 W PSU and a xclio windtunnel case ( the Noctua cooler is one of the best and one of the biggest !! this you really need a large full tower case,, the xclio is a cheapie but has excellent air flow with 2 x240mm side case fans and plenty of venting the noctua comes with 2 fans and i added another cheapie exhaust fan)

http://www.overclockers.co.uk/showproduct.php?prodid=BU-044-OE

O/C to 4.0 Ghz is really very simple and there are loads of sites that will give you starting specs

[paulselhi]

I think the rampage is more geared for hard core enthusiast gammers, the cheaper gigabyte board will do all you need, it also has the 6GB/s SATA if you need it in the future as well as 6 RAM slots

I think 6 GB is good enough for most uses but as i said heavy simulation work may benefit from larger ram

[PhoenixCG]

just noticed the rig you listed, 12gb sounds ok, the spiffy sound card is a no go though (i'm partially deaf in one ear so i've never seen it as a good investment), is the gtx 295 a big deal, i thought c4d didn't really use gpus. good stuff, you've given me some stuff to think about. thanks.

[paulselhi]

Bye the way have you looked at the price of 24 GB RAM ? You would need 4GB Sticks but i cannot find them !!

this kit is 3 x 8 GB ( does this mean the board cpuld take 48 Gb ?)

http://www.crucial.com/uk/store/partspecs.aspx?imodule=CT3KIT102472BB1067%3fcpe%3dpd_google_uk

[imashination]

hmm, half tempted to make myself a new machine now ;-) I was generally holding back because 6 gigs was a bit measly for a new machine and hadnt heard of anyone getting a nice speed from a 12 gig machine.

[paulselhi]

Mash, dare i say it ??, i think you were probably right about the 4.0 Ghz with 12 GB Ram.

But it may well be possible, thing is as you use more ram slots the o/c ability comes down so using all slots with 24 GB would more than likeley really limit the overclock. Howver if you had the dosh for 24 GB DDR3 Ram kits then i suppose you would also have the dosh for quad xeon rigs

[paulselhi]

Here are 6x 4gb stick from corsair for the I7

http://www.corsair.com/store/item_view.aspx?id=1242072

note the large heatspraders/ heatsinks, i think these can be removed

Note with any memery i woud be advisable to check with your MB manufacturer to confirm that the memory is compatable for that motherboard

[PhoenixCG]

if ocing at 24gb is impossible it's not really a problem because if i could afford the ram i could probably afford a i7 985.

i looked up the gtx 295, $550 that must be one great card. is it crucial?

i found an alienware with:

i7 920
12gb ram
gtx 260
500 gb hdd
dvd-rw
window 7 home

$1,650

i don't even think i could build a comp like that for that price. do you think i could still oc it easily?

[paulselhi]

If you look around you will find the parts cheaper my build was

Corsair 650 W PSU
HITACHI 1TB Hard Drive
Intel Core i7 930 2.8G s1366
6 Gb Corsair 1600Mhz RAM
120mm Case Fan
Gigabyte Motherboard GAX58AUD3R
CPU Cooler NOCTUA/NH-D14
DIGIMATE 23 inch monitor
XCLIO Full Tower Case with 2x 240 mm Fans

£862 Inc tax and delivery = $1325

The Noctua cooler is a bit pricey at £66 = $101

to me the card was not important so i reused an old nvidia 8800 GT

[meleseDESIGN]

You realized that you need more RAM.
How does it come?
How much polys does your models/scenes count?
I guess you wont go higher than 3.000.000 with zBrush and C4D, but I'm just guessing, am I right!?

If so, you're well suited with 6GB.

;)


i realize that the ram is probably a little low and i plan to add more asap (the mobo supports up to 24gb).

[imashination]

Mash, dare i say it ??, i think you were probably right about the 4.0 Ghz with 12 GB Ram.

Stranger things have happened ;-)

i looked up the gtx 295, $550 that must be one great card. is it crucial?

Dont buy the 295, its an SLI card, Get the 285 if you want a high end card or a 260 if you want to save some money, you wont see much difference in 3d apps.

I guess you wont go higher than 3.000.000 with zBrush and C4D, but I'm just guessing, am I right!? If so, you're well suited with 6GB.

You seem to be very confused about how much ram is needed. 6 gigs will let him work in scenes generally up to 20 million polys which is far above a typical 3d job. My last big project was 200 square miles of central london as a fly-over, that clocked in at 16 million and still happily rendered on 32bit machines, so cant have clocked more than 2.7 gigs ram usage.

Anyway, just ordered a bunch of new parts myself :)

i7 930 - slightly more expensive than the 920, but will make overclocking easier with 12 gigs.
gigabyte ud7r mobo - as paul said, theres no point spending silly amounts on a motherboard, just get what does the job
12 gigs corsair memory 1600
Crucial c300 128gig ssd
titan fenrir cooler, should be a lot easier to fit than the hulking noctua one.
Radeon 5870 - I dont use maya or max so I dont care ;-)

Ill be slapping this all in with some bits i had laying around gathering dust
seasonic 600watt psu
antec 100 case
2 1tb drives

anyone wanna buy an 8 core 3ghz mac pro? ;-)

[paulselhi]

The Noctua was fairly easy to fit but it does take up a lot of room hence the full tower case

I wouild suggest a better cooler than the fenir since the 930 tends to run hotter than the 920 and continual rendering may get the better of the wolf, Prolimatech Megahalems is rated quite highly

[meleseDESIGN]

I'll bit 1000€ but only if it comes with a warranty!

:deal:

anyone wanna buy an 8 core 3ghz mac pro? ;-)

[meleseDESIGN]

As for the GTX 295.
There is no benefits over a single GPU card for most 3D packages.
I bought my Asus Mars only because at the same time I bought a Octane license.
Octane is able to benefit from more GPU's and it will render up to 10x faster per GPU (means 20x with my Asus Mars) as any CPU based render engine and with the same quallity as well.

;)

[sentry66]

the rampage is supposed to be an OCers dream, and it supports 24gb of ram. is it not worth the $350?


The rampage is nuts. Don't get it unless you want to go down the water cooling path or some crazy extreme stuff. Stepping down to the Asus P6X58D or any of the cheaper ones would more than suite your needs. I'd pay special attention to the heatsinks on the chipsets and how many power capacitors it has. There's a lot of great boards out there, but with bad heatsinks that'll run really hot possibly screwing you up

12 gigs or ram or more will work fine. Just don't push things as hard and keep the memory settings conservative. Also be aware if you get memory that has tall heatsinks attached to them such as the corsair dominator memory make sure they'll work with the CPU cooler you're planning on using. Often times the closest memory slot to the cpu gets blocked by the CPU heatsink or fan so you can only use regular sized memory in that slot. Actually the dominator memory conflicts with almost every CPU heatsink out there with the closest memory slot

I hit 4.2 ghz so easy with my 930. Just make sure you don't use vcore voltages over 1.4 and that'll ensure you won't fry your chip over time. Also beware that there's a lot of bad advice out there saying to use vcore, pll, ICH, and IOH values that might be way too high for most people since they're gamers running 2x or 3x SLI video cards.

Going over 4.2 ghz starts getting tough.

Make sure you disable C-state or it'll reboot your machine the instant it sees you're drawing too much power beyond spec. However I found leaving C1E and speedstep enabled and using the offset instead of fixed v-core to be great at lowering temps back down when I wasn't running the machine full tilt.

[PhoenixCG]

why would they make mobos that make you choose between running cool, or using your ram slots? maybe atx has gotten to small.

not interested in pc games except for starcraft 2, so i'll stick to the evga gtx 260.

930 huh, i thought intel blocked everything but the 920 from ocing. i've got a lot to learn.

is the antec 1200 a good case, i hate cases with lights and windows, but i don't seem to have a lot of other good options.

thanks.

[sentry66]

all of today's cpu heatsinks will let you use all your ram slots with normal sized memory. It isn't until you get into the more exotic memory with the large heatfins that they won't physically fit in the first ram slot with today's large aftermarket cpu heatsinks. It's not really the motherboard's fault at all

all i7's have their multiplier locked except the extreme editions, but you can increase the BLCK to overclock any of them anyway. You can always bump the multiplier up by 1 point to permanently use the multiplier that turbo mode uses. Most overclockers disable turbo mode.

stock BLCK is always 133 and you can push it up to 200. Most boards and hardware have problems going above 211 BLCK and I believe 221 is the wall no one has gotten past.

920's are good and can be pushed far, but it's a little easier with the new 930, intel engineers have even hinted at that when they submitted the 930 to review sites.


I run my 4.2 ghz 930 at the stock 21 multiplier and 200 BLCK. I can make it clock faster but it really takes a lot more fine tuning and voltage which creates more heat and IMO it's just not worth it for pro 3D work - though maybe for gaming or poorly threaded software. I can run 4.4 ghz as long as all 8 threads are not being maxed out. Everyone's mileage will vary though. I lapped my processor and heatsink which most people aren't willing to do and my PC case is focused on cooling so I leave all my large storage drives in another computer.


case-wise, the antec 1200 IMO is a good case, but IMO bit expensive. It's going to depend on what you need out of a case with drive bays and cooling. From looking at it, I wouldn't say it does anything particularly innovative but it does have a lot of storage space. IMO smaller cases are better for cooling and need less fans to create good airflow.

Some of the CoolerMaster and NZXT cases are IMO a bit more modern with their features and cooling considerations, but you'll trade off some drive bays compared to the antec 1200. The other thing to keep an eye on is not all fans these cases come with are equal. Some push a lot more air than others. Antec generally always makes solid cases though.

[PhoenixCG]

i looked up the x58a-ud3r mobo and a reviewer said it supports the 920, 950, 970, 985, and the i9. so if i get a 930 i'd have to find a new mobo which would be a shame since he gave that one a 10/10. is the 930 worth it?

i've been looking at corsair dominator ram which has the big cooling fin on top because i thought that would be necessary to control the heat. can i get by with a regular ram stick without melting them?

the antec 1200 may not be necessary anymore, maybe a 300 could work. the 1200 was really only necessary if i was getting the rampage 2 extreme.

[paulselhi]

It supports the 930, i have both working fine, if you only plan to use there slots and a big cooler than the large heat spreads are not a problem, if you plan to use all the slots go with corsair standard sticks

this is what i use

http://www.overclockers.co.uk/showproduct.php?prodid=MY-169-CS&groupid=701&catid=8&subcat=1389

You may be able to use all ram slots with a large air cooler you need to check the coolers specs and most makers show clearance diagrams. I think i can use all 6 slots but would have to install ram before the cooler which means removal of cooler ( in my case this is also MB removal) to change ram

Of course a water cooler will not or should not block any ram slots, the corsair H2O has been reviewed O/C to at least 4.0 Ghz

Have a look at the range of cases used here for inspiration

http://www.overclockers.co.uk/productlist.php?groupid=43&catid=1270

There are 3 reasons i would go with a full tower case

1. I can easliy fit a large cooler in
2. Ease of access
3. Air flow is improved

I too am not fond of overlit cases but for a budget case the xclio windtunnel ticks all the right boxes and though the fans are lit it is not over the top

http://www.xclio.com/products/case-windtunnel.htm#details

[PhoenixCG]

right now i'm looking at the megahalem to cool the 930, and a cooler master haf 932 case. i haven't found the xclio in a US shop yet but i'm still looking

[sentry66]

the haf 932 I think is a pretty nice case or something like the stormsniper, scout, NZXT hades, thermaltake element G, antec p193, and lian-li PXx1000 are pretty good too, but it's a personal taste thing. I absolutely love the NZXT Panzerbox, especially if you set it up right, but it might not have enough storage for a lot of people and they're tough to find right now anyway. It has some of the lowest temperatures ever seen in a case, but only you set it up well, which a lot of reviewers didn't.


I'm not sold on the xclio windtunnel because I'm not a fan of the primary intake fans being from the side without a direct path for the air to exit the case. Instead it creates turbulence and eventually fumbles it's way out. Those designs pretty much always run the motherboard chipsets the hottest when used for long sessions.

When reading reviews pay special attention to what cooler is used and how it's oriented when these sites take temp readings. Some sites always use the stock intel cpu cooler for the sake of consistency, but it's design is extremely different from most of the aftermarket coolers. Also don't take any temp readings seriously if their install is a total mess with cables all over the place and not tied down out of the way or if they decided to install their hard drives right in the way of the fans etc. No one in their right mind who overclocks high would do those things.

[paulselhi]

The xclio has vents on both sides, top and rear. i have added a 120mm exhaust to the rear which augemts to noctuas second fan blowing out

Air flow in the xclio windtunnel is excellent

[meleseDESIGN]

I was just referring to zBrush, because most of the time you wont export model to C4D, Maya, Max with higher poly count above 3 millon polys.
When I export 3,5 million poly object to Maya or C4D both apps using around 1,5GB of my RAM (when I look in the taskmanager), a 14,5 mio. poly model needs around 4,5GB RAM all in smooth shade view here.

So I'm wondering how you managed a 20 mio. poly scene on a 32-bit machine?
I guess you work only in wireframe then, right?

:surprised



You seem to be very confused about how much ram is needed. 6 gigs will let him work in scenes generally up to 20 million polys which is far above a typical 3d job. My last big project was 200 square miles of central london as a fly-over, that clocked in at 16 million and still happily rendered on 32bit machines, so cant have clocked more than 2.7 gigs ram usage.

[imashination]

So I'm wondering how you managed a 20 mio. poly scene on a 32-bit machine?
I guess you work only in wireframe then, right?

Wireframe or shaded really doesnt have anything to do with ram usage, its purely how much space the structure takes up. C4D takes 100 megs to launch fresh with no scene loaded. 1 million polys take almost exactly 100 megs, both as a saved file and in memory. 20 million polys will require almost exactly 2 gigs of ram.

Given that a 32 bit OS will allow you to use 2.7-4 gigs of ram depending if its mac or windows, this is enough space for the rest of the app, undo buffers and textures. If you start copying and pasting or deleting the entire mesh then your undo buffer will rapidly start filling up as it has to make a copy of the entire mesh if you copy or delete the entire thing.

Sidetracking back to the machine. Im happy to report that Ive managed to get the i7 to 4ghz on a fully populated 12 gig machine. For anyone else wanting to built something similar:

gigabyte x58a UD3R motherboard
i7 930 cpu
12 gigs (6x2) Corsair XMS3 PC3-12800C9
Titan Fenrir V2 cooler
Seasonic 600watt

To overclock it, just crank the bclk to 182MHz and the multiplier to 22. Bingo 4.010 GHz machine. Bingo, faster than most 16-thread xeon machines and costs less than a grand :) No need to piss around with voltages or anything else, its been rendering and prime95'ing for 2 hours now, skipping around 75c which is fine, but I havent put the case together yet and all the fans are set to low.

With any luck Ill get the same for the mac on ebay and my total cost might even net me a profit ;) God bless mac re-sale values.

[meleseDESIGN]

Strange things, both mode show different results in the taskmanager here - have to figure out why.

Anyway, you have a nice machine there. It's a good example for the OP.
For the Mac you will get probably twice as much as you have spend for your new machine, I'm hoping it for you!

;)

[paulselhi]

To overclock it, just crank the bclk to 182MHz and the multiplier to 22. Bingo 4.010 GHz machine. Bingo, faster than most 16-thread xeon machines and costs less than a grand :) No need to piss around with voltages or anything else, its been rendering and prime95'ing for 2 hours now, skipping around 75c which is fine, but I havent put the case together yet and all the fans are set to low.



Are you saying you are getting 4.0 Ghz on stock voltage? what does cpuz show as the voltage whilst doing a cinebench ? Is HT on ? Is LLC on ? what is the ram voltage ?

You would be on to a winner if you just used default bios settings and raised the bclk !!

[imashination]

100% stock settings. All Ive done is set the bclk to 182MHz and multiplier to 22. It does still clock down to 2ghz when the cpu isnt under load so its generally cool and quiet, idling at 45c. Honestly I spent longer setting my harddrive to be the first boot drive than I did on the overclocking ;-)

Volts are default/auto across the board, 1.328 for the cpu and 1.632 for the ram whilst under load

HT is on

Whats an LLC?

btw I bought 2x sets of 3 sticks for the memory rather than buying the official 6-stick 12 gig set which had a £100 premium. CB render score is 6.90 which puts is bang on with what the other overclocked 920s are getting. Radeon 5870 arrives tomorrow, lets see if I can set the opengl score record :) http://www.cbscores.com/

[paulselhi]

I have just tested your settings and they seem to work using optimised defaults as stating bios and then just adjusting the bclk and multiplier as you said

LLC is load line calibration Intel implements voltage reduction on load so you should see the CPU voltage reduce when rendering (vdroop and vdrop) but with some overclocks the voltage drop is too much for the higher CPU o/c and so higher starting voltages are needed to allow for the drop

LLC effectively tries to even out the voltage thus requiring a lower voltages in the first place as it will not drop so much. However this is not the Intel spec and it has been argued by some that to keep the voltages stable can lead to LLC crating voltages spikes which at high starting voltages may cause damage to a CPU. Many people however happily use LLC

On the ud3r LLC standard is basically OFF as per Intel specs allowing for vdroop ( stop sniggering at the back !!), the level 1 and level 2 = LLC ON, non Intel spec, vdroop off

So LLC can help to keep temps down as you don't need such a high CPU voltage, it also makes overclocking easier and more stable, some higher overlocks are difficult to achieve without LLC and very high voltages

[PhoenixCG]

why did you use the fenrir over a megahalem? why did you choose that brand of psu? what are you thinking case wise, i'm thinking about the same parts as you with the exception of megahalem, gtx 260, and a corsair psu. i'm thinking haf-932, or the scout if i can fit everything in there without melting everything.

[imashination]

why did you use the fenrir over a megahalem? why did you choose that brand of psu? what are you thinking case wise, i'm thinking about the same parts as you with the exception of megahalem, gtx 260, and a corsair psu. i'm thinking haf-932, or the scout if i can fit everything in there without melting everything.

The fenrir and megahalem are virtually identical in size, noise and cooling ability, but the fenrir is half the price.

The psu is one I already had laying around the house, but originally picked it for reasonably quiet operation and because seasonic is a decent brand. I like to pick parts which are simple and work, theres no point paying for braided hoses, detachable modular cables and blue leds.

I use an antec 900, again its what I had here, but I wouldnt hesitate to buy it again. The entire front surface is a mesh grill like the mac pro. 2 large intake fans and a giant 10 inch exhaust fan. Plus, the psu is mounted at the bottom, it stays nicely out of the way as far as cooling goes, the top of the case its a straight line exit route for the heat

[paulselhi]

Cooling and overclocking go hand in hand but it is really worthwhile pushing things to extremes?

My o/c at 4.0 Ghz is considered a standard for the I7 920/930 but it is still an impressive % raise from the default 2.8 Ghz, i can get to 4.2ghz stable and have booted to windows at 4.4 Ghz but failed stress test and did not pursue stabilising that

At 4.0 Ghz a scene renders in 2m 47 secs and at 4.2 Ghz the same scene renders in 2m 36 secs, now for a 1000 frame animation that would be a saving of about 3 hours, of course i would generally expect frames to be at say 5mins a frame +, so the savings would be obviously greater. If i was doing this for a living i would weigh up the possible decrease in life expectancy of my £230 CPU against the increased income of a faster end product turn around. I am sure £230 is peanuts to a professional render job

But 4.0 Ghz is so easy to reach and have stable, it is also "cool" maxing at 66 C on my Noctua under an unrealistic stress test Prime95, i would never expect this level of stress on real world rendering or even simulations, now Mash is getting around 75 C ( OK not optimised) on his fenir which is acceptable for the CPU but if he were to o/c to say 4.2 Ghz ( without further optimisation) i think temps would be unacceptable

I Have 4.2 Ghz stable but I am not happy with the settings i have used as temps are way to high for my liking at 79 C max. I would imagine the fenir would have these even higher. If i cannot get a stable 4.2 Ghz at good voltages and temps i will stick with 4.0 Ghz, i would hope for 75 C max under heavy torture

So if 4.0 Ghz is a good overclock for you then go with a cheaper fenir, if you want to really push the CPU go with a better cooler or even a high end water setup but don't expect miracles as I think 4.2 - 4.3 is realistic tops on air and 4.4-4.5 tops on high end water

[PhoenixCG]

how often am i gonna be replacing cpus at 4ghz. i thought that stable temps mean normal life for the cpu. i once tweaked my quadro to get a boost out of it, but after about a year it was fried and i spent $700 to replace it.

[sentry66]

I think a lot of people are really timid with the whole overclocking thing worrying way too much about cpu life.

The cpus are designed to last like 10 years at full tilt at stock clock 24/7

if you overclock 50% higher, you'll shave life down to maybe 6 years running full tilt 24/7

I don't know about you, but who here is still using 6 year old computers and find them useful for rendering modern scene files that need 8 gigs of ram to render?

My machines aren't always running hard 24/7, and only run hard when I have a project that's ready to render.

The i7's are designed to overclock via turbo mode. They're also all the same base chip which is why overclockers buy the lowest end chip in that processor family. The i7 920 is the exact same base chip as the i7 975 except the 975 has a higher default multiplier. The 975 turbos up 3.6ghz stock. It also won't last as long as the 920 but do people honestly worry about that when they pay over 3x the price for the 975? This is a case where at the end of the day, more expensive does not equal faster.

They do it to cut costs. I once had an intel engineer tell me that back when they had pentiums and pentium pros, that the chips were all pentium pros to start out with, but then they just disabled the L2 cache on the regular pentiums because that was cheaper than making a separate fabrication for the pentiums. If they put that chip out today, a lot of motherboards would give you the option to re-enable that L2 cache, yet I bet there'd still probably be people saying you should pay the higher prices for the "real" pentium pro.

Temperature-wise, Intel is extremely conservative because they can't control everyone's pc case, fan setup, motherboard, and where they set their A/C at. All the processors can take a beating temperature-wise. Cooling solutions have become very advanced compared to 10-15 years ago and there's a lot more products available.


What burns up cpus and causes permanent damage is too much voltage for prolonged periods of time. There's pictures on the net of people who've done that - physically burned the motherboard socket and cpu, but those people were all nuts anyway trying to hit insane overclocks without the cooling needed to handle it which is why people bring out the dry ice and liquid nitrogen when they use voltages that are that high...and they can only run it like that for a few hours anyway before their cooling evaporates. Keep the v-core as low as possible and just high enough to power the cpu at full blast and you'll be more than fine.

for most mortals, speedstep will kick in and throttle the i7's back. And if that doesn't work, chances are you'll crash anyway before real damage even thought of happening

I personally don't sweat a $300 cpu. My overclocked systems have never died on me and were tossed long before they stopped functioning. If my i7 930 did stop working due to overclocking, chances are it'd be at least 2-4 years from now at which point it's time to upgrade or move on anyway. To date, people have been overclocking the i7 for almost a year and a half now and still running strong.

Computers all have an expiration date. You can coddle them, but eventually even the most exotic expensive machine will be outperformed by a $500 best buy special 3-4 years later. IMO it's more of a waste to not get the full power out of a computer during the useful lifespan it has. Throwing away a processor that ran stock its whole life seems like throwing away a shirt that you wore just a few times then tossed because it was no longer in fashion.

[imashination]

What he said ^^. Frankly, just treat it like crap, when it dies in a few years, who cares, youd be able to replace it with the same chip at that point in time for £50 or more likely upgrade to a newer 6/8 core chip.

[paulselhi]

What he said ^^. Frankly, just treat it like crap, when it dies in a few years, who cares, youd be able to replace it with the same chip at that point in time for £50 or more likely upgrade to a newer 6/8 core chip.

ah so overclocking an I7 is like making love to a beautiful woman.....

for those not in the UK...

http://www.youtube.com/watch?v=1BXIbl2F6gs

[imashination]

You know I very almost wrote that, but the lady was lingering over my shoulder ;-)

[PhoenixCG]

A) tell your mom to go back to her room.
B) are they ever going to make a dual i7 mobo.

[sentry66]

no, it's really doubtful they'll ever do a dual i7 board because it'll cannibalize the xeons.

Back in the day when Abit came out with the BP6 dual celeron board, that really pissed Intel off. They've since gone to special lengths to ensure a motherboard company won't ever be able to make a dual board for a processor they didn't intend to run as a dual setup.

[imashination]

B) are they ever going to make a dual i7 mobo

There is, they renamed it to the xeon and you will pay accordingly

[PhoenixCG]

xeons are great, i love how mine makes fan noises, but nothing connected to it will boot up (mouse, tablet, monitor). xeons like to tease you.

[paulselhi]

xeons are great, i love how mine makes fan noises, but nothing connected to it will boot up (mouse, tablet, monitor). xeons like to tease you.

Installing a xeon is like making.....

http://www.youtube.com/watch?v=Hr3hP0CeiQo&NR=1

[meleseDESIGN]

Not untill yet.
But maybe there is an Skulltrail coming in the near future with that you can do overclocking, like with Socket 771 CPUs?

;)



B) are they ever going to make a dual i7 mobo.

[paulselhi]

Just an update i am now running the 930 at 4.2 Ghz stable 24x7 on a x58a-ud3r

I mentioned in my earlier posts that i was not happy using LLC to even out the intel driven vdrop and vdroop. Just to recap if say you set your cpu vcore to 1.4v in bios when windows starts ( slight load on cpu) intel implements vdrop to lower the cpu voltage, when you start rendering ( high load on cpu) intel implements vdroop to lower the cpu voltage even further

This voltage drop means that you have to start out with far higher voltages ( perhps 1.4 - 1.5 v) for the cpu as vdroop will drastically lower the cpu voltage and may lower it so much that the overclock becomes unstable

LLC basically reduces or may even turn vdroop off so you can set the bios cpu far lower, there are conflicting reports on this but with LLC on ( vdropp off) i can run 4.2 Ghz at load with a cpu-z indicated voltage of 1.36 v which is below intel maximum. So i am happy with this.

Mash has stated he overclocked using auto voltage setings, many people on the net warn against this as auto may boost some voltages beyond recommended limts, not sure, perhaps those who say against auto just want to justify all their tweaking, but i have manually set important volatges cpu, qpi, vtt dram just to be safe

My temps are max 71C when rendering and simulating and 75-76 C full load stress testing

[meleseDESIGN]

Interesting thread.

One question to the overclockers gods here.

How's actually the performance boost durring the workflow with an overclocked processor running at 4GHz compared to 3GHz, is it noticeable better?

;)

[imashination]

Its 33.3333333% better ;-)

[sentry66]

yeah, the i7 scales very linearly with clockspeed increases.

[meleseDESIGN]

That's interesting!

Personaly I can't imagine sculpting in zBrush or modeling in C4D/Maya/Max/XSI could get any faster done as with an 3GHz CPU allready. Applying color adjustments in PhotoShop aso. feels like doing it in realtime with only 3GHz as well.

All I can notice is when it comes to rendering, there I feel more GHz has an noticeable impact.
Where is the biggest performance boost for more GHz as 3?
While working with PhotoShop? Maya? C4D? Modeling? Texturing? Painting? Compositing? Lighting? ect.? I'm just wondering because most of what I'm doing here occurs pretty much in realtime with a 3GHz CPU, my graphics card is the only negative aspect I noticed durring my workflow with bigger scenes.

Any idea?

;)

[sentry66]

Photoshop is super quick these days, especially in 64 bit with large files. I can't even say I really care so much about photoshop speed anymore, but perhaps with more 16 and 32 bit stuff I'd probably care more.

I mainly notice the big gap in clock speed with maya and After Effects. I don't know how many times in maya I execute a command - say a polysmooth, or waiting for a swatch to update in the hypershade, or trying to move lattice points around for a dense object, selecting 5 million faces on a 10 million polygon object, or doing a boolean, or doing several iterations of expanding/contracting component selections on dense objects - hell just switching to component mode on a dense object spikes up one lone processor, or in general sometimes I get too far ahead of the computer to keep up in maya where I've clicked a bunch of commands or hit a bunch of hotkeys and the computer just can't keep up. I find myself waiting on that single processor task to finish so I can move on. The other cores just sit there while one is floored. I always have my task manager open all the time on my 2nd monitor so I can keep an eye on things and that's what I've noticed in my years using maya.

When I say one processor is being used, I mean the percentage of one core out of all your cores since the windows kernal usually tries to distribute single-threaded tasks across multiple cores even though there's no real benefit to the end user since it's still just a single thread worth of computing. If you have 8 total threads and 12.5 % total power is being used, that's just one thread worth of work even though you might see multiple threads spike up in the graphs

With After Effects, most effects filters also only use one processor core. Some might use two. A few might use 4, some even might use all of them, but usually they just use one core.

Rendering in AE or maya is one thing, but working in them on your file is heavily bottlenecked by being limited to just using one or two processors. Any functions within the programs that use all processors are generally extremely quick. For my work, I probably wouldn't care if an already fast function is 2x as fast, but I will care if several common slow functions are 2x as fast.

I wish maya was as quick as mudbox or zbrush and I wish AE was as multithreaded as Nuke. But they're the apps I use. Until maya and AE are more multithreading, for my job I feel my time is more valuable than the computer's time. I'm willing to sacrifice some rendering speed to have snappier apps. I figure if it allows me to get my file done in 1 day so it can start rendering it that first night, that's better than having to spend 2 days on it and losing the first night that it could been rendering.

[imashination]

I'm just wondering because most of what I'm doing here occurs pretty much in realtime with a 3GHz CPU, my graphics card is the only negative aspect I noticed durring my workflow with bigger scenes.

Your gfx card isnt the limitation, your cpu is. Its extremely rare that any scene you have is limited by your gfx card, its limited by how fast the cpu can feed the gfx card. Hence the higher speed cpu will make the biggest difference, not the gfx card.

[meleseDESIGN]

Well, how much work time could you save durring an 10 hour work day with 1GHz more? 10 minutes, 1 hour or ???
I mean, compared to the lower renderspeed with an single CPU don't you even lose time in the end?

;)

[paulselhi]

As i stated earlier

At 4.0 Ghz a scene renders in 2m 47 secs and at 4.2 Ghz the same scene renders in 2m 36 secs, now for a 1000 frame animation that would be a saving of about 3 hours

and that is just going from 4.0 to 4.2 Ghz so saving 3 hours on a 46 hour render. The savings in render time from 3.0 ghz to 4.0 ghz would be far more impressive of course

[sentry66]

depends on what you're doing that day I suppose. If you're doing tons of render tests, then a dual xeon will be better, but upping the clock speed easily speeds rendering up as well. I have a maya test render file that I've used for my own personal benchmarking.

i7 930 at 2.8 ghz, turbo enabled, it renders in 291 seconds
i7 930 at 2.8 ghz, with turbo disabled, it renders in 301 seconds
i7 930 at 4.2 ghz, with turbo disabled, it renders in 201 seconds

That's a 44.7 speed increase I got in rendering from overclocking from 2.8ghz to 4.2ghz and it cost less than $100 for a nice heatsink, a few extra fans, nice thermal paste, and a little bit of trial and error time to make sure the system was stable and used the least amount of vcore voltage possible.

Your monthly power bill will go up a little, but your work will get done quicker so the machine might sit at idle more often. Overall 4.2ghz will cost just under a 50% increase in power consumption over stock. Here's my rough approximation of power usage over a 5 hour period:

stock 2.8 ghz i7
5 hours at full load of 216 watts = 1080 watts when file finishes rendering
0 hours at idle
==1080 watts total

4.2ghz i7
3.5 hours at full load around 380 watts = 1330 when file finishes rendering
1.5 hours at idle around 165 watts = 248
==1586 watts total

dual xeon 2.8 ghz (95 watt version)
2.5 hours at full load around 350 watts = 875 watts when file finishes rendering
2.5 hours at idle around 95 watts = 238 watts
==1113 watts total


A bit of a side note, turbo boost mode will run more often with a good heatsink and will draw more wattage while in turbo mode. The stock heatsink will run out of cooling ability quick and drop the clockspeed back down to the normal rated speed when one of the cores hits a certain temp. So that 291 seconds time I got with turbo enabled would probably be more like 295-296 seconds with the stock intel heatsink

I think it's fair to assume a dual 2.8ghz xeon quad-core would render my test scene somewhere around 150 seconds

If you're doing a lot of what I mentioned earlier with smooths, component selections, etc, then the faster ghz will be faster every time until those functions get multithreaded, so it's going to vary for everyone.

I use really heavy models, so modifying them usually involves doing lots of single threaded actions. I use lots of procedural textures and calculating them at render time is mostly single threading before all the processors kick in to start drawing the frame.


So overall, when compared directly to each other:
4.2ghz quad core = 50% faster at most tasks
dual 2.8ghz xeon quad core = 33% faster at rendering and multithreaded tasks

Maya 2011 is great because it finally multithreaded artisan sculpting and the non-linear deformers both of which I use all the time, but I'm still using 2009 because of all the bugs. As time goes on, clockspeed is going to have less impact on my daily work when things become more and more multithreaded. I really do look forward to that day.

[imashination]

Well, how much work time could you save durring an 10 hour work day with 1GHz more? 10 minutes, 1 hour or ???
I mean, compared to the lower renderspeed with an single CPU don't you even lose time in the end?

So by that logic why dont you use a 2ghz cpu and save some money. or a 1ghz cpu and save even more money. Saying that a faster cpu has no use because some things you do only take a few seconds anyway is just ludicrous.

No, it wont make opening your email faster, or applying a blur in ps more than a second or two faster, but if i tried to swap your 3ghz cpu for a 2ghz one, you wouldnt be very happy would you?

[meleseDESIGN]

I'm absolutelly with you guys. Higher clock rate as 3GHz might has some impact to a few tasks.
But I'm still wondering how much time you can save durring a work day if you have a 4GHz instaed of an 3GHz CPU.

What if you would have an 10GHz CPU installed?
Would you even notice the higher clock rate?
I mean, where is the limit? What clock rate do we really need to work as fast as possible?
There must be a limit, I'm quite sure about it.

Or is it just the fact that we can't have enough clock rate, even if we don't need more?

We can't have enough for rendering, but how is it for daily tasks?

;)

[paulselhi]

I wonder why you ask as you have a dual xeon X5570 with 48 GB Ram !! We are concerned with 3D issues here, rendering, simulating etc these are the daily tasks

We have given you render times for 3.0 4.0 4.2 ghz we have discussed single threaded operations and multi threaded operations. An overclock from 3.0 to 4.2 ghz will not slow down your system !! It will improve single thread operations and greatly improve multi thread operations

[meleseDESIGN]

I'm just confused, because I was told a 4GHz single CPU system would perform better as an 3GHz Dual CPU system. It gave me a reason to think about switching the system or get an new 4GHz system. But only if it is really worth. That's why I was asking if someone could mention how much work time I will save with an 4GHz system, compared to my current Dual 3GHz system.

If it's only 10 minutes or an hour I wouldn't mind to sit a bit longer in the chair to finish my work.

;)

[sentry66]

meleseDESIGN, your system right now is top notch and IMO comparable to a 4+ghz system (in terms of my needs anyway). How comparable is going to depend on your work. I'd absolutely keep using it because it is a very nice workstation and not at all worth spending more money on a brand new system.

Your system will be faster at some things while the 4+ghz i7 systems will run faster at other things. It's just a trade off, and people should choose which setup to get based on the software they use and what specialization they mainly do in their daily work. For me, it works out better to have a faster clockspeed system because of what I do and the software I use (Maya, AE, and PS) and being easy on the budget never hurts

I'd love to have 48 gigs of ram, but I have to admit other than After Effects multiprocessor rendering, I might not know what to do with that much right now with the work I'm currently doing. Maybe AE CS5 might change my tune some since it could just load tons of 1080p footage and almost never have to reload frames. Right now I've used 11-14 gigs of memory at most when multiple apps were open with large files loaded. If I had 48 gigs of ram, I'd be curious to try turning virtual memory off so the hard drive is never touched for paging, but I don't think that'd offer much speed benefit anyway. $200 for a 8gig stick is a crazy smoking good deal though.

[paulselhi]

There are no 4ghz I7's they are 3.0ghz I7's overclocked. So overclock your xeon..simples :)

[earwax69]

Yep, more Ghz help when you work in your interface, more cores help while rendering. For exemple, for After Effects, 2 years ago, I chose a 3ghz duo2core over the 2.4ghz quad2core. Its overall faster in after effects. The quad would have render faster in Lightwave though.

I think now in 2010 Intel turboboost combine the best of both world.

[tuna]

It is very task specific. As far as animation goes, with more Ghz I can playback more characters and higher resolution in real time than lower clock speeds. More cores helps me very little in comparison.

[sentry66]

the other thing is that as more and more processor cores are added, the more efficiency goes down. This has just always been the case since the cores all have to share resources

Here's some maya render benchmarks I did with my same test file, with hyperthreading turned off to ensure virtual threads wouldn't interfere, running at 4.2 ghz:

1 core 895 sec
2 cores 444 sec = 0.78% faster than perfect 200% efficiency (447.5 sec)
3 cores 304 sec = 1.9% slower than perfect 200% efficiency (298.3 sec)
4 cores 237 sec = 5.9% slower than perfect 200% efficiency (223.75 sec)

I'd probably guess by the time you got to using 8 cores, you've probably lost 9-12% efficiency.

hyperthreading seems to make mental ray renders about 18% faster since that file renders in 201 seconds when turned on

This a really interesting cinebench site that compares systems.
http://www.cbscores.com/

The overclocked systems obviously aren't on top for rendering, and no one has an overclocked 980x on there yet (which scores an 11 at 4.3ghz http://www.overclockersclub.com/reviews/intel__core_i7_980x/7.htm), but they do fairly well considering they probably cost considerably less. They do have the highest openGL scores though, even outdoing systems with faster graphics cards

[AdamT]

Wireframe or shaded really doesnt have anything to do with ram usage, its purely how much space the structure takes up. C4D takes 100 megs to launch fresh with no scene loaded. 1 million polys take almost exactly 100 megs, both as a saved file and in memory. 20 million polys will require almost exactly 2 gigs of ram.

Given that a 32 bit OS will allow you to use 2.7-4 gigs of ram depending if its mac or windows, this is enough space for the rest of the app, undo buffers and textures. If you start copying and pasting or deleting the entire mesh then your undo buffer will rapidly start filling up as it has to make a copy of the entire mesh if you copy or delete the entire thing.

Sidetracking back to the machine. Im happy to report that Ive managed to get the i7 to 4ghz on a fully populated 12 gig machine. For anyone else wanting to built something similar:

gigabyte x58a UD3R motherboard
i7 930 cpu
12 gigs (6x2) Corsair XMS3 PC3-12800C9
Titan Fenrir V2 cooler
Seasonic 600watt

To overclock it, just crank the bclk to 182MHz and the multiplier to 22. Bingo 4.010 GHz machine. Bingo, faster than most 16-thread xeon machines and costs less than a grand :) No need to piss around with voltages or anything else, its been rendering and prime95'ing for 2 hours now, skipping around 75c which is fine, but I havent put the case together yet and all the fans are set to low.

With any luck Ill get the same for the mac on ebay and my total cost might even net me a profit ;) God bless mac re-sale values.


Cool, thanks for providing the specs. I just priced it out and it would come to $1,950 with all-new parts, OS, and blu-ray burner, all prices from Newegg except the cpu heatsink.

[AdamT]

Mash,


I was thinking about maybe going with a GTX 260 instead of the Radeon you chose, since it seems to perform about the same in Cinema at half the cost. Did you choose the Radeon for Cinema purposes or was it something more sensible ... like gaming? :)

[imashination]

What, blowing terrorist's heads off at a high frame rate is a crime? ;-)

[AdamT]

All's I can say is ... YOU ROCK, MASH!! :bowdown: :bowdown: :bowdown:

Heh heh.

I just got done assembling my "Mash Box" ©, which I heavily plagiarized from the specs you listed for your system. I ended up with:

Same:
cpu
heatsink
mobo
ram

Similar:
Lian Li PC-P50 case
Corsair 650W modular PS
Hitachi 2 TB HD
Geforce GTX 260

Other:
LG Blu-Ray writer

Total cost including Win7 HP and shipping was just under $2,000.

I also used the OC settings you suggested and it's working like a charm! I've been running Prime95 for over an hour with the case closed, and cpu temps have stayed between 61 - 66C ... most of the time in the 61 - 63 range. Incredible, and it's actually pretty quiet, too.

Can't remember the CB11.5 scores right now but they were excellent. I'll post them in a little while.

Thanks a million for posting the specs.

FYI on the Lian Li case, it's very nice but if I had it to do over I would go for one that is NOT tool-less. Most of the really clever retention mechanisms turn out to be a big PITA in practice.

Re: the cpu cooler: it came a day after the rest of the components, so I assembled the system first with the stock cooler, then replaced it with the motherboard already mounted in the case. I missed my calling; I shoulda been a neurosurgeon. :surprised But damn, that thing is head and shoulders above the stock cooler ... literally and figuratively.

[AdamT]

OK, CB11.5 results:


CPU: 6.91
OGL: 48.69

Hmm, maybe I shoulda sprung for that Radeon card. :)

[imashination]

If thats clocked to 4ghz then its spot on. agreed on the tool-less cases, every one ive come across has been cheap plastic that doesnt hold anything in when the machines moved. I received a toolless machine for review once, 3 harddrives broke loose during transit and wrecked the inside of the machine

[paulselhi]

Adam what are you using to measure your temps, if they are correct then your fenir is out cooling some the top end water coolers. Somehow i doubt if the temp readingas are accurate

[tuna]

Adam what are you using to measure your temps, if they are correct then your fenir is out cooling some the top end water coolers. Somehow i doubt if the temp readingas are accurate

I think it has something to do with them raising the multiplier to 22x so they don't have to raise the base clock speed (what is it called any more? not FSB?) very high and therefore running on stock voltage. Also if he turned off hyperthreading it will reduce temps by 10c (although I don't think he did).

I found Real Temp gave me accurate cpu temps while speedfan was a lot under. So give them both a try.

Which is weird that it's outperforming (very slightly, by 0.01 :P ) my 21x multi 191mhz @4ghz OC on the same hardware (except msi x58 mobo) which should be making the RAM run faster. I also have a gtx470 and the OGL score is slightly less too. So Cinebench might just be a bad benchmark for these things.

[edit] Okay I had my RAM ratio wrong and was suffering a bit of performance, its all fixed now.

[paulselhi]

Realtemp is used by most overclockers as an accurate I7 temp monitor. speedfan is considered..well..unreliable

[earwax69]

Fenrir + 4ghz, 60C... cant believe it.

I got 70 at 3.8ghz, 60 at 3.6ghz under full load and I think my mugen 2 is well installed.

If it's the case, how do you push the multiplier to 22? Mine cant go more than 21, which is default for i7 cpu. Only Turbo-boost will push it to 22 while using a single core.

[paulselhi]

The I7 930 has a higher mutliplier

[tuna]

If it's the case, how do you push the multiplier to 22? Mine cant go more than 21, which is default for i7 cpu. Only Turbo-boost will push it to 22 while using a single core.

I think if you leave turbo boost on you can force the multi to be 22 in your OC settings.

[earwax69]

The I7 930 has a higher mutliplier

NooooooooooooOOOOOOOOOOOO!!!



I think if you leave turbo boost on you can force the multi to be 22 in your OC settings.


Turbo-boost is on and I cant force the multiplier to 22. Do you mean turbo-boost off?
You know what, im going to try right now.

EDIT: no, just tried to enbale and disable tons of stuff and nothing do it, with +/-, I cant go over 21.
Do you think the motherboard is somewhat limited?

[tuna]

Hmm, maybe it doesn't need to be on. It apparently should be unlocked for 21x or 22x.


NooooooooooooOOOOOOOOOOOO!!!
lol

[imashination]

Realtemp shows vastly different temps compared to every other app I have, and the bios. All my other apps and the bios show 40c idle, 75 under full prime load (realistically 70c under load, nothing actually makes as much heat as prime95)

But realtemp shows 54 idle and 95c under load. Considering my bios is set to auto shut down at 90...

[paulselhi]

What version of RT are you using ? I am using 3.40 and it works fine for me giving comparable readings to other apps, You may need to adjust the RT confiuration, try the sensor test

[imashination]

3.40 And whats a TJ Max? sounds like a chav clothing store.

[paulselhi]

It is the max temp specified for the chip should be 100 C for the i7 930

[tuna]

EDIT: no, just tried to enbale and disable tons of stuff and nothing do it, with +/-, I cant go over 21.
Do you think the motherboard is somewhat limited?


Okay I tried the 4ghz@ 22x multi, and my cinebench score (cpu) is 6.51 pts. compared to 6.90 with 4ghz@ 21x multi (because of 182mhz vs 191mhz base clock speed & resulting memory speedup). So I doubt AdamT's score if he followed Mash's setup exactly.

You have to set "intel EIST" to "disabled" on my motherboard in order to change the multi to 22x

[earwax69]

Intel EIST disabled and it's not working for me. Cannot pass 21. not big deal really, I just wonder why.

Would be nice to get higher clock with less voltage.

My realtemp TJ max is 99. perfect for the i7.

[paulselhi]

What Mobo do you have? and when you say it is not working at 22 do you mean you cannot select 22 or your system will not boot ?

My guess is the 870 only uses 22 when in turbo mode, the 930 on a gigabyte MoBo allows you to set 22 but i think the 920 could only be set to 21 with 22 being used for turbo

[earwax69]

MSI P55M-GD45

paid 85$ for it. All my voltage are at AUTO... maybe its that.

the multiplier stop at 21 when using the +/- in the bios.

Bios look like this; http://www.legitreviews.com/images/reviews/1102/MSI_P55_GD65_b_cell1.jpg

[paulselhi]

If the multi stops at 21 then 22 only kicks in if you have turbo enabled in the bios and will only boost one core if the others are idle

http://www.intel.com/support/processors/sb/cs-029908.htm

So basically for rendering when all cores are at max 100 % i doubt if turbo boost will be of any use, Better to disable it and overclock, even then you don;t have to use the highest multiplier, 920's and i think 930's are happier at odd numbered mutipliers

[tuna]

Would be nice to get higher clock with less voltage.


But that higher clock speed will not equal more performance, because the base clock speed is still going to be lower, which will slow down the memory, etc. If you want cooler, then reduce voltage instead of using auto, or reduce the clock speed.

[paulselhi]

All that matters for performance if the cpu frequency, how you get there is a matter for stability. 20 x 200 or 21 x 190.5 will both give an overclock of 4Ghz. RAM has a very negligable effect when rendering in terms of speed

[tuna]

All that matters for performance if the cpu frequency, how you get there is a matter for stability. 20 x 200 or 21 x 190.5 will both give an overclock of 4Ghz. RAM has a very negligable effect when rendering in terms of speed

Incorrect.

191x21 = cinebench score of 6.91

182x22 = cinebench score of 6.51

I just did this benchmark and of course it makes a difference.

It's not just the speed at which your RAM is, but the speed at which your cpu communicates to the RAM and the PCI, etc. It speeds your entire system up with a faster base clock.

[paulselhi]

0.4 !! Wow !!, however you need to run each cinebench several times to avearge out the score for each configuration

also you settings for overclock are not the same

191x21 = 4.011 Ghz cinebench score of 6.91

182x22 =4.004 Ghz cinebench score of 6.51

[tuna]

0.4 !! Wow !!, however you need to run each cinebench several times to avearge out the score for each configuration

also you settings for overclock are not the same

191x21 = 4.011 Ghz cinebench score of 6.91

182x22 =4.004 Ghz cinebench score of 6.51


My settings for OC aren't the same because you CANT have them the same, they're always going to be slightly different. But since you're being pedantic, so shall I.

I tweaked RAM timings a bit and here are the 3 latest benchmarks:

191x21 @4011mhz = 6.93pts
183x22 @4026mhz = 6.63pts
182x22 @4004mhz = 6.59pts

I ran them all multiple times but cinebench results aren't to enough decimal points to show any fluctuation.


And that small increase might be important depending on what you're doing. For me I think it matters because it directly translates into increased playback FPS, where an extra frame or two is a noticeable and nice thing.

What I haven't tried yet is getting more mhz out of the 22x multi at the same voltage I have now on 21x. If it's much higher than the current mhz, then I could see a speed increase, but I'd have to re-do prime95 testing for hours and hours.

[paulselhi]

Speedfan is definately giving false temp readings using it's default configuration

[earwax69]

Killing turboboost doesnt make a difference. Im stuck at 21 anyway. Not big deal. Im sure it would be different on a Gigabyte motherboard. Im still very happy with my i7. It's the best thing I bought since my 30inch Dell monitor.

[imashination]

191x21 @4011mhz = 6.93pts
183x22 @4026mhz = 6.63pts
182x22 @4004mhz = 6.59pts

Actually theres a bit of a big mistake there and the others are correct. Ram speed has virtually no impact on render speed, between 183MHz and 191MHz the difference is a fraction of a percent at most.

You have a bios setting somewhere to step up the multiplier as high as it goes (22) when under heavy load. This doesnt just kick in when a single core is in use, this is something completely different.

When your system is set to 183x22, this is its highest speed, it wont increase the ram speed and the multiplier is already maxed out. When your system is set to 191x21 as a base, under load the motherboard is still ramping the multiplier up to 22, so in fact under load you have 191x22 which gives 4.2GHz.

The difference between 183x22 and 191x22 is 4.4%
The difference between 191x21 and 191x22 is 4.4%
The difference between CB 6.93 and 6.63 is 4.4%

Your machine is just running faster than you thought it was ;-)

[tuna]

Actually theres a bit of a big mistake there and the others are correct. Ram speed has virtually no impact on render speed, between 183MHz and 191MHz the difference is a fraction of a percent at most.

You have a bios setting somewhere to step up the multiplier as high as it goes (22) when under heavy load. This doesnt just kick in when a single core is in use, this is something completely different.

When your system is set to 183x22, this is its highest speed, it wont increase the ram speed and the multiplier is already maxed out. When your system is set to 191x21 as a base, under load the motherboard is still ramping the multiplier up to 22, so in fact under load you have 191x22 which gives 4.2GHz.

The difference between 183x22 and 191x22 is 4.4%
The difference between 191x21 and 191x22 is 4.4%
The difference between CB 6.93 and 6.63 is 4.4%

Your machine is just running faster than you thought it was ;-)

Nope. I'm looking at cpu-z right now as I'm rendering 100% all 8 hyperthreads, its at 21x.

If I set the Multiplier to 22x at 191, it crashes real quick with prime95.

Turbo boost (if it was enabled, which it isn't), would only kick the multi to 22x if less threads were being used as well. It doesn't kick it to 22x when all 4 cores are at 100%.

Go look at some other people's OC benchmarks and stability results (at hardforums or extreme systems). Almost everyone uses 21x. I'm pretty sure the entire overclocking community hasn't been duped into wrong multiplier results.

RAM speeds are not the only things affected by a faster Bus speed either; The QPI Link is running at 3447Mhz The entire system's cpu/chipset IO has a higher bandwidth as a result.

[paulselhi]

With 21 set as mutiplier on my 930 it DOES kick to 22 when all processors are 100 % and turbo boost is on giving a boost from 4Gz to 4.2Ghz when rendering

However volatges and other O/C settings have to be in place to accomodate this 4.2 Ghz from the start else the system will crash when it ups the cpu frequency

Personally if i have need of the 4.2 i use a seperate profile with a O/C to 4.2 at 21 multi with turbo bosst of and voltages et to handle this

To be honest 4Ghz is rock stable and runs at around 65 -68 C on heavy renders sometimes hitting low 70's when doing heavy simulations, load voltege is 1.28 V so I am more than content to live with 4Ghz

[tuna]

With 21 set as mutiplier on my 930 it DOES kick to 22 when all processors are 100 % and turbo boost is on giving a boost from 4Gz to 4.2Ghz when rendering


Oh, thats cool then.

But it's definitely off on my system, cpu-z constantly shows it at 21x, and I can't get this chip stable @4.2 at these voltages so it's not gonna happen. I'm also really happy with its performance @4ghz. :)

[imashination]

But it's definitely off on my system, cpu-z constantly shows it at 21x, and I can't get this chip stable @4.2 at these voltages so it's not gonna happen. I'm also really happy with its performance @4ghz. :)

Yes, but when you set it to 4.2, its overclocking to 4.4 as the multiplier kicks up to 22 ;-)

[tuna]

Yes, but when you set it to 4.2, its overclocking to 4.4 as the multiplier kicks up to 22 ;-)

Shush, you!

[CGTalk Moderation]

This thread has been automatically closed as it remained inactive for 12 months. If you wish to continue the discussion, please create a new thread in the appropriate forum.