Workstation Videocard Woes....please help!!

[erazal]

Nobody has ever used an ATI FireGL X2 for professional use before?!

Since the FireGL X2 is significantly cheaper than the Quadro FX 3000, i wanna know for sure if the price is worth Quadro.

[dvornik]

Whatever you use for max is going to be slow. Check the Maxtreme 6 post here somewhere - maxtreme is what makes Quadros look faster in those tests. In ATI case it's supposedly the Maximum drivers - I've never used ATI though.

That article shows firegl lagging behind Quadros in "optimized driver" obviously. Which part has blown your mind?

I don't think FX 3000 is worth the price for max. It's going to be a little less slow than a 1000.

[erazal]

What blew me away was that the FireGL X2 actually beat the Quadros in a few of the benchmarks. That's why i was asking about how the FireGL X2 handles max.

I've used ATI with 3d Max before and have never used Nvidia myself, and i must say, i sure hate it when Max crashes and shuts down for no good reason, whenever the intensity of my animations gets a little bit on the high side.

So your saying right now that the FX 3000 is slower than the FX 1000 when it comes to using 3d applications like Max and Maya?!

I find that a little hard to believe, anybody else can back this up?!

[3Dfx_Sage]

I think he said the 3000 was *less* slow than the 1000.... implying that either one is going to be slow but the 3000 will be faster than the 1000.

[loop29]

The problem is that most of the benchmarks are not showing application performance. I experienced this by myself with the Quadro FX 2000, for example SpecviewPerf 7.1. shows massive performance increase. But application performance is not going up like the benchmarks would suggest, for example ProE and UGS scores. Ok, my PC is somewhat slow but the tendency would be the same for other systems. I don´t think that you can say that the Quadro FX series is slow, but for the price/performance ratio the Quadro 3000 is not worth the money. I think that was what dvornik tried to say, you will have close application performance with a Quadro FX 3000/2000/1100. Well, I should mention that I use a Geforce FX 5800 with modified drivers so I can use this card as a full working Quadro FX 2000. What you can do is grab a FX 5800 (Ultra), you can now get them pretty cheap of ebay or a store, and see how it performs with your application and than you can give it back after testing and get the prof. card you think you need. My personal opinion is ATI=trouble under OpenGL, so I wouldn´t consider it, but that is more from the MCAD viewpoint.

regards

[3Dfx_Sage]

Originally posted by loop29
The problem is that most of the benchmarks are not showing application performance. I experienced this by myself with the Quadro FX 2000, for example SpecviewPerf 7.1. shows massive performance increase. But application performance is not going up like the benchmarks would suggest, for example ProE and UGS scores.
there's a very simple explanation for that- they key is that everyone uses the same benchmarks which means everyone is using the exact same scene over and over. It's very easy to hand-code driver detections and optimisations for a specific scene if you know exactly what's going to be shown. Workstation vendors have been doing this for a very long time and nVidia has been doing it with games since as far back as the original Quake, and ATi was caught doing it with Quake3 a few years back. It's even happened in the CPU world with design choices made only because they would give higher benchmark numbers (like the guy that hand-coded a 30 instruction loop and used it to represent overall performance of the IA-64).

Benchmarks are used by marketers to convince people to buy their product and, as such, end up nothing more than a product of marketing and not real technical referenes.

[GregHess]

Here are a few "main" schools of thought on upgrading video cards.

1) Buy the fastest accelerator you can afford, and upgrade every other generation. (Basically skip a gen or two)

2) Buy a midrange variant of whatever the current generation of cards offers.

#2 is sometimes cheaper in the long range.

#1 usually works pretty well, because it takes a good year or so for software to catch up to the hardware capabilities of a card. Case in point, a Quadro 900 XGL. Still a VERY fast card.

A good question to start out asking is...

What card are you currently using? What in your opinion is the downfall of your card? Under what conditions does it slow down? What processor is powering the card? What max version are you using?

3dsmax is unique in that it has a program known as maxtreme available to it. Maxtreme is nvidia's 3dsmax optimization software, which in some cases, can GREATLY speed up the viewports.

Without maxtreme however, virtually any nvidia will work in max ogl, the program's just not very picky when it comes to nvidia cards, whether they be geforce's or quadro's.

Depending on the sys config, I'd get a Quadro 1100 FX and spend the cash I saved on ram and cpu upgrades :).

[erazal]

Well Greg, here's the thing...

This card i'm getting is for the brand new system i'm building. Right now, i am using:

P3 733Ghz
512mb of ram
30gb HDD
AND the videocard is a Radeon 64mb.

With those specs, you draw your own conclusions to how fast this baby can fly;)

I'm one of those types to get the newest things now, rarely upgrade and lasts me about 4 years till i get a new one.....which brings me to this, my new system. I'm planning on getting this setup:

P4 3.2ghz 800Fsb /w hyperthreading
Corsair TWINX Dual Channel 1024MB PC3200 Registered ECC Memory (2 x 512MB)
2 S-ATA 80gig drives
videocard? tba.

I'll be using 3D stuido Max 6 and Maya 5. I'll also be doing a lot of crazy, heavily complicated, lots of characters onscreen type of animations, so keep that in mind.

So what can you make of my specs and which videocard to get? The reason i was opting for the 3000 was because i wanted it to last, however if the $$$ is not worth the actual card, then forget it because the last thing i want is to waste the $$$ i have.

Right now Greg, i can get the Quadro 1100FX for $914 which is a very good deal here in Canada. The Quadro 3000FX for $1800 is also an amazing deal, but as i said before, $2000 for a videocard that wont make much difference than buying the $914 one is kind of a crazy thing to think of.

[3Dfx_Sage]

if you're wanting to last for 4+ years then yeah I'd say the 3000 is worth the money. However, since you're planning on this being a long-term commitment I'd say get an NV40 if you can wait a few months. For people who upgrade every year ot year and a half waiting a few months may not be worht it. However, the speed improvement that the NV40 will likely extend the lifetime of the system 1-2 years. The only problem is that I have heard that nVidia is facing delays again and likely won't launch in April as was previously expencted. It could well be May or June, however at the least it will make the existing cards go down. I think you would actually save money by buying a GF4 now (perfectly fine card, last night I took out my FX and put in my old GF4; my system just feels so much better) and then waiting until the next gen quadro's hit the market. At that point you can either get a mid-range next-gen Quadro or an FX3000 for much cheaper.

[GregHess]

Thats an excellent idea by Sage.

Buy a cheap inexpensive video card for your new system, then upgrade to the new quadro line when they come out later this year.

If your looking for something thats fast AND cheap...see how much a Ti 4600 will set you back. That card goes for a bit under 150-160 USD in the states...and in non crazy textured operations, it'll hand most of the FX line a beating.

[Neison]

Originally posted by GregHess Buy a cheap inexpensive video card for your new system, then upgrade to the new quadro line when they come out later this year.

Just on a side note, will the NV40 also be applied to the QuadroFX Go mobile series? The FX Go 1000 was just released and I was debating about upgrading my M60 sometime in the next couple months. However, if this new chipset will go across the board, I'll obviously wait it out :)

[erazal]

That's a very good idea actually, i'm going to try and get something cheap and then get the newer cards later this year. It'll give me something else to look forward to. It'll sure save me a lot of $$$ for now.

Greg, what brand is the Ti 4600? i'm going to look for prices now.

Also, what do you think about the GeForce FX line? should i invest in that for now if it's cheap? I'm willing to spend like $400CAD for a card now and in around August/September get the NV40 at that time.

[loop29]

If you go Geforce I would consider a FX 5800 if you can grab them cheap. I bought one last week for 147 Euro´s and that is pretty low price. The point is that you will be able to use it as a Geforce and as a Quadro FX 2000 with 44.xx and 45.xx drivers until you get NV40 or Quadro FX 3000. In my opinion it is on the same level as a Geforce 4 and the advantage of testing Quadro capabilities is a plus.

regards

[3Dfx_Sage]

well you can use a GeForce4 as a Quadro4 as well. Just last night I finally got fed up with my accursed GFFX 5800 and ploped my GF4 back in. Hey, want to buy a very slightly used ASUS 5800? :p

[erazal]

Okay, so technically the FX1100 is the equivallent of a GeForce FX what exactly?

and if i buy the GeForce equivallent, will it be good enough for now? cuz i'm pretty sure the NV40 will be for gaming also, and not professional work.

I heard that a workstation card, even an older one, will beat the crap out of any gaming card in viewport speed mostly because of the optomization....if that's true, i sure as hell would rather get a workstation card cuz that's exactly what i need....and currently i'm unfamilliar with the NV40.

Well tomorrow is the big day and i'm gonna go out there and buy my new PC. I'm thinking of using my older videocard...the one i'm using right now....for now. Anyone have any issues against that? cuz i figure if i'm going to upgrade anyways in the next few months, then i might as well just use my older one for now. The Radeon 64mb piece of crappp lol

[3Dfx_Sage]

the 1100 = FX 5700.

Yes, the NV40 will be a gaming chip. However, there will also be an NV40GL which is the workstation version. But I think that even a gaming NV40 will slaughter a QuadroFX 1100 and 2000 in almost all scenarios.

[erazal]

agreed.....BRING ON THE NV40....

One question though....will it support BOTH AGP 8X AND PCI Express? cuz the current board i'm getting wont have PCI express.....at least i dont think it does

:shrug:

[3Dfx_Sage]

yeah, the NV40 is natively an AGP 8x chip. There will be a PCI-E version of it however it will still only run at AGP8x speeds because it merely uses an AGP-to-PCI-E bridge, the chip will still be functioning as an AGP8x chip.

oh and this is interesting, although I do have my doubts as to it's validity... http://www.theinquirer.net/?article=14373

[GregHess]

Sounds like Pr². (Pre-Release X Public Relations).

Either way looks like a kickin card. I think I'll try to make it through 2004 without buying any computer hardware...think I can do it? Aw crap I just bought that scanner.

Well the rest of 2004...

[loop29]

Lol @Greg, the scanner doesn´t count : )
3dfx, why did you plugged in your GF 4 again?
I´m in the same situation than you and I have to say the FX 5800 beats my GF 4 Ti 4400 in all fields. If I only could grab in my case without fear because of these superhot memory heatsinks. I did that mistake yesterday again, I clocked the memory down to 400 MHz and still when I touched it, it had 70 or 80° .... (ouch)!

regards

[CgFX]

Originally posted by 3Dfx_Sage
yeah, the NV40 is natively an AGP 8x chip. There will be a PCI-E version of it however it will still only run at AGP8x speeds because it merely uses an AGP-to-PCI-E bridge, the chip will still be functioning as an AGP8x chip.
This is not 100% accurate. nVidia will be running its AGP on the private side of the HSI chip at AGP 16x speeds. You loose the full-duplex nature of PCI-E but it will be the same bandwidth (4 GB/s vs. 4 GB/s) to the GPU.

All apps and games currently assume half-duplex access and performance to the GPU anyway so it is arguable if this will be a factor until applications start showing up that do significant amounts of reading back from the GPU.

It should be interesting to see witch strategy proves to be the better one.

[3Dfx_Sage]

Originally posted by loop29
3dfx, why did you plugged in your GF 4 again?
I´m in the same situation than you and I have to say the FX 5800 beats my GF 4 Ti 4400 in all fields. If I only could grab in my case without fear because of these superhot memory heatsinks. I did that mistake yesterday again, I clocked the memory down to 400 MHz and still when I touched it, it had 70 or 80° .... (ouch)!

regards
well because it was giving me lots of troubbles. I tried changing the drivers and got BSOD after BSOD. One time it just refused to reboot at all (could get the BIOS ot POST, but everything powered up) and I know what I would have to do it fix it- reset the motherboard BIOS (not the first time this happened). When I put the GF4 in it booted up but I found my BIOS had a checksum error and reset itself to default (like it often would with teh FX in it). I havn't had any problems since moving back to the GF4.

As for speed: the main game I play (Lineage2, uses a modified UT2 engine) was totally unplayable at times on the FX, often dropping down to 0.x frames per second in crowded areas. Also, whenever water came into view it slowed down tremendously (water is shaded using PS 1.2). My GF4 Ti4200 (overclocked to 4600 speeds) never slows down even with 2xFSAA enabled, whereas I couldnt use AA or AF on the FX. The only difference in quality I have noticed is that the FX was showing better shadows on the Ti Mi Tran's (treant-like things). The GF4 only shows a filled in outline, where the FX shows all the little branches and leaves. However, there were also numerous bugs that appeared when using the FX- often, when a 2d plane mapped with an alpha map (like a collage leaves) was in front of the camera things would just disappear. My character would poof, tree's would poof, bushes and plants would poof, mob's would poof, and spell effects would poof. If I moved that camera slightly everything would reappear. Also, I noticed that water was not being shaded at great distances with the FX but it is with the GF4, so I don't think it's an engine optimisation. There were various other z-related issues when alpha was used as well.

[3Dfx_Sage]

just out of curiosity I read your post...

Originally posted by CgFX
This is not 100% accurate. nVidia will be running its AGP on the private side of the HSI chip at AGP 16x speeds. You loose the full-duplex nature of PCI-E but it will be the same bandwidth (4 GB/s vs. 4 GB/s) to the GPU.

All apps and games currently assume half-duplex access and performance to the GPU anyway so it is arguable if this will be a factor until applications start showing up that do significant amounts of reading back from the GPU.

It should be interesting to see witch strategy proves to be the better one.

Yeah, I forgot about that AGP16x thing. Of course, it won't really matter if you'r using AGP8x, 16x, or PCI-E 16x for quite some time. If you look at the performance of the GF4 4x version and look at the 8x version you'll see little-to-no improvement. AGP bandwidth has already reached what developers are pushing through it. That will change, however, when developers start using the high GPU-t_CPU bandwidth to offloead non-traditional (even non-rendering) work to the GPU. GPU's are very quickly becomming highly parallelized vector processing units which can be a great boon when it comes to things like collision detection and can perform some tasks many times faster than the CPU can. However, PCI-E probably won't even be hitting the market in volume until much later this year or even 2005. It will likely be even longer before programmers start to really take advantage of it. The NV40 generation shouldnt even need PCI-E bandwidth, and designing them for AGP is, I think, a better idea because most people are giung to be buying the AGP version. This saves both money and also you don't have to worry about any latency induced by a PCI-E-to-AGP bridge (like ATi will likely have to face).

[erazal]

Sooooo......i can use my AGP8X mobo with the new NV40 coming out?

[3Dfx_Sage]

yes. the NV40 was designed primarily for AGP 8x.

[wshorten]

I would love to know where I can get a Quadro 1100FX for $914. I live in Canada too, and can't find it for less than $1,100.

I would appreciate the tip.

[erazal]

Originally posted by wshorten
I would love to know where I can get a Quadro 1100FX for $914. I live in Canada too, and can't find it for less than $1,100.

I would appreciate the tip.

There really isnt a "Tip" i can give you, other than the fact that you should find and be best friends with a distributor. That's how i'm getting those prices.

Now in my case though, i have one distributor friend i grew up with that's just starting out and i'm not gonna ask him for any help just yet.....so this deal is coming from another distributor who isnt my close friend, BUT is the brother of the guy who is gonna marry my sister:p

Sure helps

[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.