PDA

View Full Version : Any reason to buy AMD PC?


HONG KONGer
01-12-2003, 11:58 AM
Hyper-Threading+Dual channel DDR=unbeatable
Do U think so?

LeeTN
01-12-2003, 05:06 PM
Not necessarily, no. MHz and GHz speeds are not the end all of processing, nor is HT. That said, AMD is planning to include HT in the new Hammer chips later this year.

One troubling fact about AMD is that they have all but said that the current Athlons, Athlon MPs and the MPX chipsets are dead. With no future upgrade path, I would wait, if possible, until the new Hammers become available before investing in a new AMD system.

Xeons and Pentiums are great, but they are more expensive in certain set-ups. If you have to have a new system right now, and can afford it, I would go with the Intel solution simply because there is at least a pathway to 3.2 GHz (announced) and probably a little more before the current slots are played out.

HONG KONGer
01-13-2003, 02:10 AM
> AMD is planning to include HT in the new Hammer chips later this year.

If it is true it would be amazing. But when will Hammer release?

> 3.2 GHz (announced)

When will be out and what's the difference with the 3.06GHz?

elvis
01-13-2003, 03:59 AM
Originally posted by HONG KONGer
Hyper-Threading+Dual channel DDR=unbeatable
Do U think so?
no. i think it's all marketing crap. intel need to stop relying on buzzwords and SSE2 and actually make a chip with decent FPU performance rather than stupidly high clock speeds that are outdone by other chips 500MHz slower.

that's what i think. :)

GregHess
01-13-2003, 04:59 AM
When will be out and what's the difference with the 3.06GHz?

It will be out around the 3000+ XP (Barton) release. Though this is slated for Jan-Feb, the actual date will vary depending on who releases what first. (As always the two companies try to take away each others steam)

The 3.06 is a different chip in a variety of regards.

a) It is currently one of the hottest chips on the market, producing an equivilant amount of heat per surface area as any AMD chip on the market.

b) It requires more power to do this, and thus requires that a board support HT, or at least support the additional power requirement for this chip to function. Many boards which support HT, may or may not support the 3.06 HT because of this.

c) It has the lowest throttling temp of any P4 (69C). This of course means, that at 69C the processor begins reducing its megahertz to reduce its heat production.

AMD still has the best price/performance ratio on the market, and still exceeds Intel's in performance...in certain areas. (Viewport OGL for example, is currently the 2800+ XP on an nforce2 in first place).

Remember that unlike the Intel chips, AMD chips perform extremely well in any unoptimized software. Programs which do not have any advanced instruction sets (like SSE2) will benefit much more from an AMD processor (due to the superior architecture) then they will from the bloated Intel pipeline.

Sieb
01-13-2003, 05:56 AM
I agree.. It seems Intel these days relies alot on its speed and its name. But they still have yet to solve the FP situation which is all that matters to us. Intel has the OEM market to itself basicly and still refuses to listen real users. AMD still listens to customers and has stayed isolated enough that they can concentrate on their work. Unlike intel trying to get their name in on everything. A 2800 still beats the pants off other systems I have tested. It will be even nicer when the Bartons show up. And once hammers are here, I believe they will be well worth their cost.

Speed hardly matters much these days when most games and programs still only require a 1Ghz system. I have yet to see anything need faster. I still see high end servers for sale that are slower and I still run a 1.4Tbird. But if you want to pay the premium for a fully rigged HT system, go for it. But you can get a lot more for your money if you go AMD. And you won't have to keep buying new mobos with each new processor revision.

A KT4 mobo with a 2800+ and some 3200XMS memory runs for roughly 600+- bucks.. The P4 3.06 HT processor alone is over 600 bucks.. :) And MPs will always be cheaper than Xeons and still perform better.

My only issue with AMD is that I wish they would release processors rated at better speeds. Instead they release a 2Ghz, then a 2.08, then a 2.17....

HONG KONGer
01-13-2003, 12:23 PM
Benchmarks on the web seems to show completely different things. Please look here:

www.tomshardware.com/cpu/20021216/images/image021.gif
www.tomshardware.com/cpu/20021216/images/image023.gif
www.tomshardware.com/cpu/20021216/images/image033.gif
www.tomshardware.com/cpu/20021216/images/image034.gif

www.anandtech.com/cpu/showdoc.html?i=1746&p=16
www.anandtech.com/cpu/showdoc.html?i=1746&p=17
www.anandtech.com/cpu/showdoc.html?i=1746&p=18

elvis
01-13-2003, 12:34 PM
read these forums (or just about anywhere else) and you'll come across the fact that tom's hardware isn't worth the paper it's written on (so to speak).

i would disregard anything printed on that site. so would many others.

remember also that benchmarks are indications, not proof. we all model/render different things for different reasons. due to this everyone's results are going to be different.

GregHess
01-13-2003, 12:48 PM
Tom's hardware is like the local whore. Everyone's had a go, and you can smell her coming for miles.

Here's some links to real sites.

www.2cpu.com
www.3dluvr.com (cough self plug)
www.aceshardware.com
www.tech-report.com
www.anandtech.com
www.overclockers.com
www.ars-technica.com

Slightly biased...

www.hardocp.com
www.maximum3d.com
www.hexus.net
www.digi-life.com
www.xbitlabs.com

Incompetent/Liars

www.tomshardware.com

Thalaxis
01-13-2003, 04:00 PM
Originally posted by LeeTN
Not necessarily, no. MHz and GHz speeds are not the end all of processing, nor is HT. That said, AMD is planning to include HT in the new Hammer chips later this year.
[b]

No, they are not. They are not planning on introducing HT for quite a while... in fact they've clearly stated that they will implement dual cores first, which won't even be feasible before they move to the .09 technology node, which won't be for a minimum of another year, and that will probably just for samples by then.

[b]
One troubling fact about AMD is that they have all but said that the current Athlons, Athlon MPs and the MPX chipsets are dead. With no future upgrade path, I would wait, if possible, until the new Hammers become available before investing in a new AMD system.

Xeons and Pentiums are great, but they are more expensive in certain set-ups. If you have to have a new system right now, and can afford it, I would go with the Intel solution simply because there is at least a pathway to 3.2 GHz (announced) and probably a little more before the current slots are played out.

They haven't said anything of the kind -- yet. In fact, they've said quite the opposite; the Athlon XP will continue to be their bread and butter through 2003. The Hammer won't replace their volume chip until next year; this year it's going to be largely their high-end, low volume cachet part (Athlon64) and their low-end server part (Opteron), the latter of which they are hoping to push upwards into the mid-range and eventually one would assume into the high-end markets.

Now that Granite Bay motherboards are becoming available, you can get a 3.06 P4 machine with 4 GB of inexpensive memory that is still able to outpace the best that AMD is currently shipping.

Thalaxis
01-13-2003, 04:04 PM
Originally posted by elvis
no. i think it's all marketing crap. intel need to stop relying on buzzwords and SSE2 and actually make a chip with decent FPU performance rather than stupidly high clock speeds that are outdone by other chips 500MHz slower.


Hm... in light of the fact that it's competing head-to-head with the
Power4 and Itanium2 in most respects, I think it's safe to say that Intel is doing something right.

elvis
01-14-2003, 12:30 AM
Originally posted by Thalaxis
Hm... in light of the fact that it's competing head-to-head with the
Power4 and Itanium2 in most respects, I think it's safe to say that Intel is doing something right.

no. intel did something right a long time ago and are currently getting by on the quality that was associated to their name 5 years ago.

the fact that you have to pay double the price for a, intel system that outperforms an AMD box by maybe 5% is not what i'd call a sensible price to performance ratio, nor a good ROI.

AMD was the best thing that happened to the CPU market. they single handedly forced intel to drop their prices and increase the rate at which they produced hardware. with no AMD we'd all be sitting on less than 2GHz machines with intel telling us to wait another year. it's a well known fact for years they were holding back on chips to increase their profits. in the early days of the PII-233 there were plenty of inside voices about intel's PII-400MHz chips they had in their labs, at working volume. yet we as the public didn't see them for nigh on 12-18 months.

reguardless of who has the performance crown, AMD are a sensible alternative for those of us who don't have gold-lined pockets, as well as being a pain in the arse for intel, which suites me just fine. :)

here in australia i'm looking at a price list with an intel celeron 2.0GHz at AU$179, and an athlonXP 2000+ at AU$175. which would you buy?

Thalaxis
01-14-2003, 03:38 PM
The fact that the Athlon is a great processor doesn't change the fact that so is the P4. The reason it's priced the way it is is simply a reflection of Intel's marketing muscle.

You can deride Intel's marketing all you want, and I would just sit back and laugh in agreement, but their engineering teams have done an excellent job with the P4.

You could argue (and I would agree) that the P4 has reached such stellar performance levels only because of the fact that the Athlon turned out to be such a beast.

The upshot is that we now have 2 good alternatives to choose from in a desktop processor, and if AMD survives the next year and launches the Hammer successfully, then this will continue for some time.

So far, it seems like there is good reason to be optimistic about the Hammer... but it also seems like we won't have much longer to wait to find out, since they plan to launch the server versions in just another couple of months.

GregHess
01-14-2003, 04:39 PM
You can deride Intel's marketing all you want, and I would just sit back and laugh in agreement, but their engineering teams have done an excellent job with the P4.

I completely disagree with this statement.

The Pentium IV is an excellent example of how engineering can go wrong, and what lengths a company will take to convince everyone that its not actually their fault.

The Pentium IV which is being compared here (mainly the 3.06 HT) is a 4th generation P4...unless you count all the subtle iterations, then its more like the 12th generation.

So since were talking about Intel's excellent engineering here...lets start at the begining and look at the changes to the architecture....

The original Pentium IV's released to the market were Socket 423 Williamante's. These chips had 256k l2 cache, and lost a FPU unit to double the lenght of their pipeline.

Whats doubling the length of the pipeline do? It gives alot more elbow room for raising the core frequency.

But what is the pipeline? Why's it important? Here's a quick/poor analogy.

Take an assembly line. Normally it takes 10 steps for a product to go from one end to the other. Suddenly they double the size of the assembly line, now it takes twice as long for the same product to go from step 1 to step 20, as it did to go from step 1 to 10. But you can now fit more people at any given step. (Cause now there are 20 steps instead of 10).

If you look at this at a PR perspective, AMD calls it IPC, or instructions per cycle...or the amount of work a chip does at a given megahertz.

Jumping back to the Pentium IV's....

The inital releases came out at 1.5, 1.4, and 1.3. The first and inital response to the Pentium IV was unexpected. Returns. Tons and tons of returns. Why were people returning these chips? 1.5 gigahertz is a lot right? No its not....not when it takes twice as long to execute an operation. The high % of inital returns was due to the fact people's older computers were faster then their replacements.

Imagine owning a ghz pentium 3, buying a 1.5 gigahertz P4 for 3k, and having it be SLOWER then the chip you just replaced. In most instances with the initial P4 arrival, if you owned a 800-1000 p3, you'd be extremely disappointed in the new machine, and in some instances extremely angry. (My boss got one, it was 25% slower then his ghz machine). Add into the fact that these machines were using Rambus, and the normal Intel markup, you were seeing 4,000 USD machine's beaten by year old machines. Thats poor engineering. Did I mention the original chips required specific motherboards, cases, cooling, and power supplies....just to work?

But Intel knew this. They knew the P4's handicap was its pipeline. But the pipeline was extended for a reason...so they could raise the core speed up. They covered up this fact through PR, and lots of talk about SSE2...which wouldn't see actual real use for another year or two.

Within a few months the Pentium IV Williamante would reach 2.0 Gigahertz, and end its life as Intel abandoned the current socket, and moved to Northwood, Socket 478.

(Notice I haven't mentioned AMD yet. At this time the 1.0-1.333 Tbird's were shipping, and on average outperformed everything Intel threw at them by a MASSIVE amount).

Its good to note here that Intel was developing the P3 line as well here...turning it into Tualatin. Tualatin scaled up past 1 gigahertz, all the way to 1.4 gigahertz...until Intel realized it was slaughtering their xeon line. A Dual 1.4 Tualatin-M can and will outperform a Dual 2.0 Xeon in many server tasks...and most definitely in price. So Intel cut the Tualatin off at 1.4 gigahertz (It could be at 2.0 gigahertz easily now) to prevent it from outcompeting the P4/Xeon line both in performance and cost. Excellent example of good engineering vs bad engineering within the same company. PR saved the day though :).

(This is a really long response, so I'm editing it now and then so I don't lose something, so it'll just grow with time).

So here comes northwood. Time to buy a new motherboard. (As with most intel stepping upgrades). All the previous Williamante chips were stuck at 2.0 ghz, with no possible upgrade routes. Northwood was a new socket, with a smaller die, and 512k l2 cache. Basically not a P4, this was almost an entirely different chip. Starting at 2.0 Gigahertz, these chips would scale up to around 2.6 before intel would cut them down.

At this time engineering problems in bandwidth and latency became apparent, as the Northwoods were bottlenecking due to lack of bandwidth. Both memory and front side bus. Intel began pushing SSE2 on everyone they could fine to try and reduce the crippling effects of the P4 architecture.

Why's SSE2 so important? Remember that 20 stage pipeline? Well imagine that if something was using SSE2, it only had to go through 14 stages of that pipeline, and could skip 6. Thats the equivilant effect of SSE2 if properly integrated.

So back to the northwood. All of intel's engineer's are freaking out cause Intel also wants a "budget" pentium IV solution. So they quickly throw together plans for a P4 Celeron, as well as an SDRAM solution for the Pentium IV. (Worst Idea ever). On the mainstream they figure out away to get the QDR FSB to 133 megahertz, bringing the FSB to 533 and helping reduce some of the bottlenecks in the system.

Note on SDRAM P4 Platforms: On most tests, a SDRAM equipped P4, will run 500 megahertz slower then a DDR or Rambus equipped P4. Ex. a 2.0 P4 running SDRAM, is equivlant to 1.5 P4 running Rambus.

So now its time for another motherboard. The old ones didn't work with the 533 FSB. Yay. So far we've had a 500 megahertz upgrade route, and then a 600 megahertz upgrade route, and now...wow, we can upgrade a whopping 200 megahertz...before we need another new motherboard.

The final revision of the P4 is Hyperthreading. Blah blah blah. PR. Blah blah blah. PR. Hyperthreading is basically a logical processor which allows the primary cpu to better deal with idle cycles and secondary threads. On average it can improve performance 0-15%. (There are cases where its shown more then 40% improvement, but this is only theoretical benchmarks). On average its around a 5-7%. Whats that 5-7% cost everyone? Another motherboard, a new psu, and a 700 USD processor. [Reminds me of 1996 all over again].

So down to the point.

The Intel architecture is crap. It only performs best at extremely high clock frequencies, and only then if the application, or program is threaded with SSE and SSE2 instructions. Its saving grace is that Intel was able to convince a majority of oem's to go with their cpu over AMD's. (Gateway, Dell, Etc). This puts Intel cpu's all over the market place in many consumer boxes. Does this make it a superior architecture? No. It just means a large % of the consumer population was unaware of an alternative solution. Remember that universal agreement on anything is usually a sign that something is wrong :).

The best example of anti-consumerism is to walk into a local computer shop, or even best buy. Ask them what % of systems they build are AMD...you'll be surprised. Bestbuy sells a greater % of AMD systems then Intel, even when the AMD system's aren't available off the shelf. (They build custom boxes)

Intel's hold on the oem's is about to change very soon as AMD as already grabbed many an OEM for the Clawhammer and Hammer releases. Then we'll see what happens when superior architectures are pittied against cheap imitations.

So why would you want a Intel cpu?

1. If your application is SSE2 threaded (Most Cg app's rendering engines), it'll give some nice performance.

2. Intel cpu's tend to run cooler then their AMD counterparts (Except the 3.06 HT).

3. Intel cpu's tend to be more heat tolerant.

4. This means on average intel systems can be quieter (if designed correctly).

5. There are some incompatibilities with certain AMD chipsets with a small mount of hardware. If your hardware falls in this catagory, you'd best go Intel.

6. Intel systems are not more or less stable then any AMD solution.

Now over to AMD...

Why get an AMD system over an Intel one?

1. AMD processors execute more instructions per cycle, doing more work in less megahertz. This means if your running any application which isn't SSE2 threaded, and/or the majority of your applications do not feature intel optimized code, that the AMD processor will be far superior in terms of performance then any Intel solution. (Example...scientific tests, dna analysis, phelogency tests, older cg apps...like max3, max4)

2. AMD chips tend to be more inexpensive then their Intel counterparts.

3. AMD is currently holds the OGL performance crown. (2800+ XP on an nforce2 board). And will accelerate the viewport faster then any other solution.

4. Newer revisions of AMD motherboards feature additional overheating features such as active fan control, throttling, and shutdown features...matching them up against Intel's.

5. The KT133A chipset has support from a 600 Duron, all the way to a 2200+ XP. AMD chipsets tend to have much greater longevity then Intel solutions, meaning you can save money by not having to upgrade over a longer period of time. (The nforce2 boards have support for the upcoming bartons). This is also because socket A has remained the same for a very long time.

6. AMD systems are not more or less stable then Intel ones.

What gives me the right to type up this type of post? I've been working with machines since 1997. I've been writing reviews and been an active hardware consultant since 1998. I've been working as an admin for a 125+ computer department for the last 2-3 years. I've also built and worked with every generation of P4 and Athlon. I've tested them in an untold # of apps, and am still testing them (Currently working with boxxtech on Dual 2.8B Xeon's and 2400+ MP's).

Snorks rule.

Thalaxis
01-14-2003, 05:06 PM
Originally posted by GregHess

I completely disagree with this statement.


While I can dispute your points, first I just want to say thank you for being civil and rational in your disagreement.

(I've seen too many flamewars over this sort of thing...)


The original Pentium IV's released to the market were Socket 423 Williamante's. These chips had 256k l2 cache, and lost a FPU unit to double the lenght of their pipeline.


The 2nd FP unit wasn't cut for the pipeline, it was cut so that they could produce the processor on a technology node that it wasn't intended for. They built it on a .18 micron process because they had no choice; the PIII had hit its limit (7 stages will only get you so far), and the 1.1 GHz PIII ended up being recalled because of that.

What you're looking at is a rush job, not marketing issue.


Whats doubling the length of the pipeline do? It gives alot more elbow room for raising the core frequency.


Your analogy is flawed. The point behind having a longer pipeline is, indeed, to enable higher clock speeds... because the more you break up each task, the simpler it is. The slowest task is that sets the limit for the entire processor's clock speed, which is why Intel pulled the scheduler out of the critical path and went with a trace cache.

Even comparing to the Athlon, their decision bears some fruit (I'll get to IPC in a moment); even though it takes more clock ticks to get an instruction from one end of the pipeline to the other, the P4 has enough of a clock speed differential to retire it sooner anyway.


If you look at this at a PR perspective, AMD calls it IPC, or instructions per cycle...or the amount of work a chip does at a given megahertz.


AMD's design isn't that far different from the P4's in some ways; they do, after all, have a 15-stage pipeline, and the hammer has a 17-stage pipeline.

Even though the Athlon has more parallel execution resources (9-issue superscalar vs 6, for example), it doesn't in practice show this... partly because, unlike the P4, it can't feed its own core well enough (memory), and because its scheduler can't find enough parallelism to pull this off anyway.

Intel's approach was to design for multithreading. The validation process for that was VERY extensive, and several iterations of core revisions were specficially intended to cure some problems that they found with their implementation of that particular technology. But they went with that design because it makes the scheduler's job considerably easier, which leads to better utilization of available resources.


But Intel knew this. They knew the P4's handicap was its pipeline. But the pipeline was extended for a reason...so they could raise the core speed up. They covered up this fact through PR, and lots of talk about SSE2...which wouldn't see actual real use for another year or two.


That sort of problem has been the case with almost every new core they've shipped. But in actuality, the P4's problem wasn't the pipeline. That huge pipeline is one of the P4's strengths. The P4's biggest problem wasn't the lack of utilization of SSE2, either.

The real problem is that it didn't get along well with poorly optimized code. The PIII and Athlon are designed to perform well with crappy code. The P4 was not. That was a large part of the reason that the P4 didn't perform well out of the box, Intel were counting on their ability to convince developers to recompile their software, while the marketing drones conveniently ignored that little snafu.

In any case, the proof's in the pudding: if you look at the actual performance data for the PIII and the P4, it turns out that the P4 even with the Willamette core was actually delivering close to 50% more floating point performance per clock cycle than the PIII.

It suffered more in the integer side of things than in the floating piont side of things initially, though; this is because FP performance tends to be bandwidth-limited, while integer performance tends to be branch-limited.


Its good to note here that Intel was developing the P3 line as well here...turning it into Tualatin. Tualatin scaled up past 1 gigahertz, all the way to 1.4 gigahertz...until Intel realized it was slaughtering their xeon line. A Dual 1.4 Tualatin-M can and will outperform a Dual 2.0 Xeon in many server tasks...and most definitely in price. So Intel cut the Tualatin off at 1.4 gigahertz (It could be at 2.0 gigahertz easily now) to prevent it from outcompeting the P4/Xeon line both in performance and cost. Excellent example of good engineering vs bad engineering within the same company. PR saved the day though :).

Tualatin only has a 7-stage pipeline... I seriously doubt that even Intel could have really pushed it to 2 GHz, let alone past it, at least not without moving to an even smaller than .13 micron process.

Still, look at the facts. The P4 has obviously shown that it can perform in the real world; look at how close it's coming to Power4 performance... and Power4 is an 8-issue superscalar design. Sure, it's ahead of the P4 with 1/2 the clock rate, but look at the price, the size of the 2nd level cache, and so on. Intel's engineers felt that they could get better performance by gunning the clock speed, so they did... and they designed it for simultaneous multithreading so that they could improve its utilization (i.e. IPC) down the road.

The fact that they're outperforming everything on the desktop, and most of the upscale server stuff is rather difficult evidence to ignore.

GregHess
01-14-2003, 05:20 PM
Btw excellent discussion. You points on the pipeline are very interesting. And I was pretty sure my analogy was flawed. Its hard to keep tabs on all the hardware, and learning about the inner engineering behind everything, and not go completely insane.

The fact that they're outperforming everything on the desktop, and most of the upscale server stuff is rather difficult evidence to ignore.

See this is where we disagree. I don't see this as fact. Thats the whole point of my discussion. They aren't winning 100% of the time. You've got cpu's almost 800 megahertz slower, still beating the flagship P4's. If you put a 1.0 P4, vs a 1.0 P3, vs a 1.0 Tbird, vs a 1.0 XP. You'll end up with this...

1.0 XP, 1.0 Tbird, 1.0 P3, 1.0 P4.

What really sucks is when you see holes in your own argument, so you try to type really fast to fix them, but someone has already read/replied before you can keep typing. Ha!

What would be really interesting would be a 2.0 tbird a 2.0 XP, a 2.0 P3, and a 2.0 P4. If bandwidth wasn't an issue. (All chips having their maximum available bandwidth)

In fact, thats a great idea. Just need an intel engineering sample. That would make a great baseline and put a stop to this discussion. hehe.

Thalaxis
01-14-2003, 05:33 PM
Originally posted by GregHess

But this isn't a fact!!!!


Actually it IS a fact...


Thats the whole point of my discussion. They aren't winning 100% of the time. You've got cpu's almost 800 megahertz slower, still beating the flagship P4's. And I still see the pipeline as intel's biggest fault. If you put a 1.0 P4, vs a 1.0 P3, vs a 1.0 Tbird, vs a 1.0 XP. You'll end up with this...


Actually, the real point of your discussion is that the P4 doesn't have enough of a performance advantage to justify the price tag, especially when you can get a dual Athlon for not much more (possibly less) than the price of a flagship P4.


In fact, thats a great idea. Just need an intel engineering sample. That would make a great baseline and put a stop to this argument :).


No, it wouldn't... what I'm getting at is that clock speed is nothing more than a design parameter. Intel's marketing droids attached a lot of false significance to it, but that doesn't change the fact that in the real world, the P4 is still the king of the desktop.

However, as you noted, the gap isn't particularly large. That's why the Athlon is still able to command a better price/performance.

And I agree, stability is no longer a sticking point. It hasn't been for a while.


And btw...Intel can make the tualatin faster. They just choose not too.

While they probably end-of-lifed it sooner than technology dictated, it's pretty certain that there is no possible way that it could have hoped to reach high enough speeds to match the current Northwood iteration without a significant redesign.

Carnifex
01-14-2003, 05:36 PM
When the first P4's came out and we got them in in the store that I work at, I decided to run a little experiment, since I was looking for a new upgrade to my then current rig.
I brought in the demo for Cinema 4D XL 7 on a disc and proceeded to install it on the following machines.

HP P4 1.5 Ghz 128 Meg of PC800 RDRAM

HP P3 933 128 Meg of PC133 SDRAM

Compaq Duron 800 64 Meg PC100

Compaq Athlon Tbird 1 Ghz 128 Meg of PC100

Apple G4 Cube 450 Mhz with 64 Meg of PC100 (Had to enable virtual memory since it wouldn't run the demo render without it)


I set all these machines at the exact same settings, disabled any other software running in the background and then rendered the same scene on all machines at the exact same settings.

The results were as follows:

HP P4 1.5 : 48 seconds
Apple G4 450 : 48 Seconds
HP P3 933 : 36 Seconds
Compaq Duron 800 : 36 seconds
Compaq Tbird 1 Ghz : 18 seconds.

I call this my real world test and while it may not apply for all things, due to individual apps being optimised better for the P4, like Lightwave, it applied excellently to me, since my primary 3D app is Cinema 4D.
I had in a way sorta made up my mind already about what kind of processor I was gonna buy when I upgraded, but seeing results like this, totally convinced me that AMD was the way to go.
I think what this particular test shows beautifully is how ridiculously under powered the P4' s FPU performance is.
So much for 50% more FPU performance than other processors :)

GregHess
01-14-2003, 05:39 PM
I tried to edit my message fast enough, but you replied too quickly :).

I wasn't comparing a 3.06 HT to a Dual athlon...I'm saying the 2800+ XP beats it in a variety of situations. Thats why I am disputing the fact that Intel currently has the performance crown. IT has most of the crown, but is still missing a few rather large jewels.

And when we speak of architecture, anytime you see an 800 megahertz gap, and the slower proc is STILL competing, I'd say its pretty damn impressive, and does showcase problems with the P4's architecture.

Thalaxis
01-14-2003, 05:57 PM
Originally posted by GregHess
I tried to edit my message fast enough, but you replied too quickly :).


I thought it was just because of a stalled network connection, but that's the disadvantage of asynchronous communication :)


And when we speak of architecture, anytime you see an 800 megahertz gap, and the slower proc is STILL competing, I'd say its pretty damn impressive, and does showcase problems with the P4's architecture.

Well, given the stellar performance of the P4, plus the fact that it's specifically designed to combine massive bandwidth and high clock speed, I would instead have said that the fact that the Athlon can compete while being so thoroughly outclassed in fabrication technology and memory performance says a lot about the phenomenal engineering that AMD put into the Athlon.

GregHess
01-15-2003, 03:29 PM
I agree :).

puu
01-16-2003, 11:43 AM
...wait for Opteron/sledgehammer later on this year, the first ones are sposed to be on par with Intel P4 3.06ghz, but will only clock in at 2.0ghz...BUT with 64bit computing power you can have more RAM, waaayyy more RAM! :bounce:

Thalaxis
01-16-2003, 04:08 PM
The wait is almost over... the first Hammers (for enterprise customers) are due to start shipping in just another 2-3 months...

CGTalk Moderation
01-14-2006, 02:00 AM
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.