After having taken a long hard look at GPUs last fall, it is time, once again, to see what the graphics card vendors have to offer the world. We wanted to get our hands on as many cards as possible in order to put together a roundup of epic proportions.
There is more to buying a graphics card than an IHV's chip, and the tender loving care that the OEMs give their cards is where much of the end user value comes from these days. In this review, we will take a look at each card and what it has under the hood, along with its GPU. Beyond the gaming experience, we will be looking at fan noise and heat dissipation as well as which memory modules the vendor decided to use. Some of the cards are even pre-overclocked for our gaming pleasure.
We will see everything from two-fan cooling solutions to LCD temperature and fan speed readouts. After looking at what each card has to offer above and beyond the performance, we will put them all to the test. Obviously, if the cards all run at reference speeds, they will all perform similarly to earlier tests that we have already completed. We've got something more special in mind.
We are going to overclock each card to its highest stable core and memory speed, and then run a few select benchmarks to cover the most recent few DirectX releases as well as OpenGL. All told, 23 cards were benchmarked, so we didn't have the ability to run a whole suite of tests on each one.
Keep in mind, when looking at overclocking, that every single card is different. We are only hoping to get an idea of what a particular vendor's cards can do. Heat sink and fan selection are very important to overclocking, as is RAM cooling and board layout. Though these issues impact overclocking a great deal, they, alone, do not dictate success or failure at any given clock speed. Each GPU, RAM module, capacitor, etc. has different limitations, which can only be determined experimentally.
Beyond understanding what the vendors are doing, we will be able to glean a little more information about the IHV's GPUs. Hopefully, our tests will tell us what kind of speed increases we can expect from each core, and what impact core and memory clock speed have on performance. But that's enough talk. Take a look at the cards in our lineup.
UPDATE: As the performance tests in this article place a heavy weight on overclockability, it is important to note that the Radeon 9800 (R350/R360) based cards are fabbed on a 150nm process and will have a little more trouble overclocking than the 130nm parts that make up the rest of the cards we tested. Also, NVIDIA GPUs underclock themselves when running in 2D mode, so their idle temperatures will be lower on average than the ATI based solutions. We appologize for having omitted this information from initial publication.
Facebook
Twitter
RSS
44 Comments
View All Comments
DerekWilson - Thursday, February 5, 2004 - link
We have added an update to the first page (and other relevant pages) of the article explaining a couple things that were omitted from the original publication of this article. One of these points being the fact that NVIDIA cards are underclocked when running 2D; the other being the fab process of ATI's R300/R350/R360 GPUs.I hope these points being included in the article help to explain some of the results we see here. Thanks again, everyone, for your feedback and input.
Indig0 - Thursday, February 5, 2004 - link
Derek, this article is great, so many questions answered, hooray for nVidia!!! I was very dissapointed by nVidia's previous lackluster preformance, and I am pleased to see that nVidia figured out how to close the gap.I have some questions though, Valve said that in HL2, the nVidia FX cards ran better if seen as DirectX 8 cards instead of DirectX 9, does this judgement still stand with the new nVidia drivers? What about with the new 5700 and 5950 cards?
DerekWilson - Thursday, February 5, 2004 - link
We have an article dedicated specifically to image quality, and all vendor's cards performed on par with the reference cards they extend from in terms of IQ.http://anandtech.com/video/showdoc.html?i=1931
impar - Thursday, February 5, 2004 - link
I think Pumpkinierre hit the jackpot!It is nice to read such a thorough review of graphic cards speed, noise and temperature but in my opinion the power required by each card should also be taken into consideration!
For example, in the mainstream we have 96XT, 57U and 59SE. Of those which one is the less/more power demanding?
It would be nice to see that kind of info displayed...
I imagine IQ wasnt a part of the review due to time constraints?
DerekWilson - Thursday, February 5, 2004 - link
Thanks for the linkage ..."As it stands, I'd say we've got ourselves a riddle wrapped in an enigma with hand-coded conundrum planes."
intensely complex proprietary hardware running equally intricate closed source software having to do with very involved mathematics will often hand you a situation that can be described as such. :-) ;-)
Pete - Thursday, February 5, 2004 - link
Thanks for the reply, Derek. I look forward to future articles and your Halo investigation.FYI, I posted a thread about that at Beyond3D. If it turns out that the limiting factor in Halo is not pixel shaders, then the default benchmark would seem to become less than useful in indicating PS2.0 performance. As it stands, I'd say we've got ourselves a riddle wrapped in an enigma with hand-coded conundrum planes.
http://www.beyond3d.com/forum/viewtopic.php?t=1014...
DerekWilson - Thursday, February 5, 2004 - link
qquizz: Thanks for the suggestion; we'll include that card next time.Pete: As always, your input and comments are appreciated.
The way we do our graphics makes showing multiple series for large datasets extremely difficult. We could have done this for the single card type graphs though, and will in the future if that's something that would help.
As for game choice. It would have been impossible to pick a PS2.0 game benchmark that was not part of TWIMTBP (Tomb Raider and Tron are both TWIMTBP games). The unreal engine is a widely used engine in the gaming industry, and we could not have left UT2K3 out. Your are, however, wrong about JK:JA. It is not an XBox port; the XBox version was released near the same time as the PC version from which it was ported. JK:JA is very much a sequel to the other titles in its series and is based on the OpenGL Q3A engine (Its called (Direct)XBox rather than GLBox for a reason).
In the end, games were chosen for poularity of engine or title only.
As for graphs, its a mixed bag. Some people prefer percentages while others perfer pure frame rates. Of course, this was really the first article of its kind in quite a long while, and we've got a few nice tricks up my sleeve for the next incarnation that will involve percentage graphs... The overall performance graphs were more of an afterthought as the goal of the review was to show the performance gains OEMs could make in their catagory.
On HIS: we actually have an IceQ in the lab, and we were very excited about including it in the roundup. Unfortunately, we plugged it in and it didn't work. Nothing but the fan came up. On the whole, we were very impressed by HIS, and we look forward to including them in future reviews and roundups.
Seeing that the NVIDIA cards reach higher high temps should tell more about the cooling than idle. But your point is well taken, and in future reviews we will include a page about the reference cards themselves. A page is necessary because idle clock speeds aren't the only caveat to which we must pay attention when looking at these cards.
As for your Halo specific question, that's a matter we are still researching. :-)
Thanks for your questions and thoughts, they are always appreciated.
jbirney - Thursday, February 5, 2004 - link
Bummer,1) no mention of lack of tri-linear on the FX cards.
2) no mention that Halo allows for partial precision vrs full precion for the ATI cards
"Many people crucified the NVIDIA cards with Tomb Raider, Aquamark3, ShaderMark and Tron benchmarks. The problem is that there are popular games out there that NVIDIA cards perform better in, and that info got lost in the shuffle."
No its that in these titles, they utilize more of the PS2.0 engine. We all know the ATI cards have a better PS2.0 engine. And while driver tricks on NV part will help to close the gap, the fact remains. I also agree with Pete I find it very strange that you hand picked DX9 games that NV was good in. Why not include some of the other ones that NV did not do so well?
Pete - Thursday, February 5, 2004 - link
Hey Derek,Thanks for the article. An impressive amount of data, as usual, and some surprises WRT to OCs. The noise data are especially useful.
A few criticisms and questions follow, if you're willing to suffer them. It's clear which is which, but if you feel I'm displaying an anti-nV bias (it's not intentional, and perhaps the pro-nV bias I perceive in this article is also unintentional on your part--or all in my head), at least consider responding to the questions.
1. Too many graphs! Could you show plain and AA+AF scores in the same graph, or are you locked into a default format?
2. I also found the choice of games somewhat curious, in that all three seem skewed toward nV. UT2K3 is a TWIMTBP title, and both Halo and JK:JA were developed for Xbox (exclusively among the consoles) in addition to the PC. It's not a big deal--I focused mainly on the heat, noise, and OC'ing results--but it doesn't seem entirely appropriate. The fact that you settled on a more "mainstream" CPU, included all the cards in one graph and listed framerates rather than percentages to me implies that you expect the benchmarks to be considered in and of themselves, rather than just as OC'ing yardsticks.
2a. The Halo bench, is that the game's built-in one that measures the intro cut-scene? Is that a really good test of PS2.0 effects and speed? I ask because I noticed in the huge recent THG roundup that the 5800U scored very closely to the 5900U, yet the latter was (IIRC) supposed to offer much-improved shaders. I thought it strange that the NV35's extra FP units didn't translate into performance gains in a title that used FP shaders. Perhaps NV3x's PS performance in Halo is register-limited?
3. Could you perhaps add comments on why the cards score as they do in the games (CPU-/bandwidth-/fillrate-/shader-limited?), or is that outside the scope of the article? If so, then why not just show OC'ing gains as percentages, to make it easier to compare with the gains over stock core and memory speeds?
4. In light of the HIS 9600XT's exemplary OC'ing, I was curious why you didn't also test an HIS 9800 Pro with the well-reviewed IceQ HSF? Did HIS not have one to ship to you, or is HIS not selling the latter card in the U.S.? It's curious that I can't find a single HIS card on Pricewatch or PriceGrabber, because I'm almost positive I found them there before.
5. I, too, forgot that nV cards are underclocked when idle. I think you'd be remiss not to remind your readers of that. If people are going to take the risk of OC'ing their cards, they'd be better off with all the info. At first, I thought the differences in idle temps indicated the FX cards have more effective HSF's and thus OC better, but that's not really something I can draw from the data. OTOH, the IceQ Cooler seems to allow for much greater OC's of 9800 Pro's precisely because it cools the GPU so much better than the stock HSF.
Thanks again.
BTW, I think 12x9/10 +/- AA&AF is a good compromise res to test cards with in the future (high-end 4xAA, mainstream 2xAA).
Pumpkinierre - Thursday, February 5, 2004 - link
The temperature measurements are a welcome change to a graphics card review. For too long reviews have gotten by with statements such as 'it runs warm'. When I built my system I probably wouldnt have bought the 9800pro had i known the levels of heat produced- 70C on the heatsink at load. I had to remodify my cooling to target the GPU and now keep it within 60C.Presently, with justifiable criticism levelled at Prescott P4 regarding excessive heat, it would be nice to know the max. heat dissipation, max. case temp., current draw and wattage of these GPUs.