ATI X1900XTX vs ATI X1800XT

[halduemilauno]

http://www.hothardware.com/viewarticle.aspx?articleid=785&cid=2
:)

[fatal1ty]

That system has only had ATI cards in it.

If you look at my FEAR results, the XTX was almost double the single GTX 512 (edit: max fps that is, plus the average is a good deal higher) and the single XTX was close to the SLI results. I'm not sure what you're complaining about there.


Right - My point was that your FEAR results are the only place the X1900XTX beat your single GTX 512 - take alook at the Hexus link I pasted - the XTX beats the GTX in more than JUST FEAR. My GTX512 was not even close to the X1900XTX in most the games that I had benchmarked using them both.

[OWA]

After looking at other people's results for HL2 (for example) the results are usually pretty close between the XTX and GTX. Doom 3 results between them are also close. So, the only thing I can think of is that there are a few things possibly tipping the balance toward my nvidia system. I already mentioned the memory timings being a little different and tonight while trying to think of other reasons that might tip the scale, I realized I hadn't installed the dual-core hotfix on the 2nd system yet. Also, if Hexus or other sites are using a reference GTX 512 my results would also have the added clock speed advantage. So, maybe the memory timings + hotfix + factory overclocked differences are enough to tilt the results toward Nvidia.

I did run the optimizations on cases but it mostly stayed the same. The only ground the XTX gained was in Doom 3 but it was still behind (56.6 vs 63.2). So, the result of optimizations on was that the XTX handily won in FEAR, 3DMark05@1920x1200 (7796 vs 7074), and 3DMark06@1920x1200 (4414 vs 3876).

So, I'm not going to post the chart until I install the dual-core hotfix, try relaxing the memory settings, and re-run the benches.

[fatal1ty]

hrmm...
Hexus Review...
Processor(s) AMD Athlon 64 FX-60
2.4GHz, single-core, 1MiB L2 per core, Toledo

Hexus used a FX60, I have an opteron 148 @ 3.1 Ghz...

Maybe Nvidia off loading to the Dual Core CPU is gaining some performance ATI hasn't yet tapped....

Using single core cpu and seeing bigger gains using the X1900XTX that would be my belief.


Scrap that - ok I am using a single core CPU ;)

Still maybe that has something to do with it...

[OWA]

Dual cores do help with Nvidia in some games and 3DMark (gained over a 1000pts in 3DMark05 with the dual-core drivers). Doom 3 gained around 10fps. Other games only gained a few fps. I don't believe I've checked recently to see how much dual-cores help with ATI. I know they added that to their drivers but if I remember correctly, it helped in some and hurt in others initially. The same sort of thing has happened with Nvidia. The dual-core support by both is coming along but I don't think it's fully arrived yet. Even so, dual-cores vs single-core do often produce different results. If Hexus is using a FX-60, I think it's actually a dual-core CPU though.

I've got the memory settings the same now and I've installed the dual-core hotfix on the 2nd system. I'm re-running things again to see how it goes this time.

The one difference I can't eliminate is that the motherboards are different. They're both nforce 4 SLI boards but one is an Asus and one is an Abit. In motherboard reviews, the results usually show a few fps difference so that's something else that could cause slight differences. I really just meant this as a quick check though. A recent thread mentioned gaming with a single card on the Dell 24" LCD (1920x1200) and I didn't think it would be very plausible without changing settings. So, I wanted to actually test that and then see how much SLI improves the situation over a single ATI or Nvidia card.

[fatal1ty]

Dual cores do help with Nvidia in some games and 3DMark (gained over a 1000pts in 3DMark05 with the dual-core drivers). Doom 3 gained around 10fps. Other games only gained a few fps. I don't believe I've checked recently to see how much dual-cores help with ATI. I know they added that to their drivers but if I remember correctly, it helped in some and hurt in others initially. The same sort of thing has happened with Nvidia. The dual-core support by both is coming along but I don't think it's fully arrived yet. Even so, dual-cores vs single-core do often produce different results. If Hexus is using a FX-60, I think it's actually a dual-core CPU though.

I've got the memory settings the same now and I've installed the dual-core hotfix on the 2nd system. I'm re-running things again to see how it goes this time.

The one difference I can't eliminate is that the motherboards are different. They're both nforce 4 SLI boards but one is an Asus and one is an Abit. In motherboard reviews, the results usually show a few fps difference so that's something else that could cause slight differences. I really just meant this as a quick check though. A recent thread mentioned gaming with a single card on the Dell 24" LCD (1920x1200) and I didn't think it would be very plausible without changing settings. So, I wanted to actually test that and then see how much SLI improves the situation over a single ATI or Nvidia card.

Yes the FX60 is DC, my mistake. I'm really wanting to try a DC chip but I realize the only gain would be theoretical anyways.. ie: 3dmark 06, etc...

[OWA]

Some games see a boost with DC as well but the improvement is usually minor. 10fps is the biggest gain I've had. But, the thing that is great is that Windows is really snappy with DC. I had the single core 4000+ in the second system and even though it was 2.4ghz as well, the system was like it had gotten a fresh install of the OS after I put the DC in. Everything is more responsive and snappy. I don't really do all that much multitasking but just the improvement to windows in general was enough for me to go DC in both systems.

...or the quick version. :) I highly recommend them especially if you use Windows a lot for things other than gaming.

[fatal1ty]

Some games see a boost with DC as well but the improvement is usually minor. 10fps is the biggest gain I've had. But, the thing that is great is that Windows is really snappy with DC. I had the single core 4000+ in the second system and even though it was 2.4ghz as well, the system was like it had gotten a fresh install of the OS after I put the DC in. Everything is more responsive and snappy. I don't really do all that much multitasking but just the improvement to windows in general was enough for me to go DC in both systems.

...or the quick version. :) I highly recommend them especially if you use Windows a lot for things other than gaming.



DAMN!! 2 x 4800+ on Newegg ~ $1300!!!! You must be rich man! ;)

[OWA]

Heh, I really wanted the FX-60. So, not rich, not even close but no car payment and no mortage payment does equal a lot of disposable income.

Okay, below is another table. Four things changed this time. I relaxed the memory timings on the SLI system so that they match the other system. I installed the Microsoft dual-core hotfix on the 2nd system. I set the CCC and Nvidia CP to app control for AF except for timeshift. With timeshift I forced 16xAF since there isn't an in-game setting for AF. I ran the tests with optimizations enabled (Quality mode for the Nvidia system, Cat AI mode for the ATI system).

With optimizations enabled, it usually just changed things slightly except for the HL2: LC single GPU score on the 7800. In the timedemo, it saw a big jump. I ran it 3 times and got essentially the same thing each time. So, I think I must have entered the wrong value yesterday, otherwise, I'm not sure why the jump would be so high. I'll try it again in HQ mode shortly to see if I get yesterday's results or not.

http://img156.imageshack.us/img156/736/xtxvsgtxsummaryoptson9iy.jpg (http://imageshack.us)

Maybe I should add SCCT to the mix as another double-check. I think I'll try disabling Adaptive AA and TAA:SS next though just to see how that affects things.

Edit: I may just wait until the official drivers for the XTX are released.

[OWA]

The official Cat 6.2s didn't really make any difference (vs the beta 6.2s). The results I posted earlier are in line with the official driver results. With the factory overclocked 7800 GTX 512, it seems hold up pretty well against the XTX (except in FEAR). I'll try to include more results later and try running the 7800s at reference speeds. Maybe most of the review sites used reference cards?

This time, it's a comparison of the driver defaults (i.e. opts on, no Adaptive AA or TAA). Res is still 1920x1200 and using 4xAA/16xAF.

It looks like the temperature gauge is flaky with the 6.2s. With the betas, it would sometimes show zero and sometimes it worked. With the official 6.2s, I seem to be showing 20C all the time. That's one way to resolve concern with high temps I guess. :)

Edit: My chart has an error. The Cat 6.2 beta should say Cat 6.2 official. I used the official drivers for the last set of tests.

http://img425.imageshack.us/img425/6631/xtxvsgtxsummaryoptsonnoextraaa.jpg (http://imageshack.us)

[OWA]

Okay, more high quality mode widescreen gaming comparisons. Using the official Cat 6.2s this time. If the game had lots of frames to spare, I then tried bumping up the AA level even more. A couple of games aren't quite apple-to-apple comparisons. In TrackMania Nations, the ATI cards do hard shadows while the Nvidia cards do soft shadows. I wasn't sure how to change that. In Pacific Fighters, I did two runs for the Nvidia cards. With the ATI cards, they only support up to a water quality level of 2. The Nvidia cards support up to level 4. This is with the patch released today (4.03) and they completely reworked the water renderer. The level 2 looks pretty good, very similiar to level 3. Also, I didn't notice much difference between it and the level 4 but I didn't compare screenshots or anything.

Keep in mind that since this is a high quality mode run, all optimizations are disabled. This hurts ATI's opengl results (at least with Doom 3) since they made up a lot of ground with the Cat AI optimization. I kept the HQ AF mode of ATI off for most of the tests since I was trying to make it a more apples-to-apples comparison. In one of the games that had a really high framerate (CS:S), I did turn it on seeing if would reduce the crawl effect that the stress test exhibits (with both cards). Maybe at some point I'll re-run all of those to see what the HQ AF impact is but most of the time, it doesn't cause much of a hit.

Edit 2: Chart deleted. See next post.... Old chart attached showing. Wrong part with yellow box around it.

Edit: A couple of things about TrackMania Nations. One, when playing around with the shadow options, I messed up the Nvidia results. Two, once I fixed that I don't see a difference in the shadows. The installer shows Nvidia cards having soft shadows while other cards aren't supposed to have that but when comparing screenshots from similiar areas, I'm not seeing that. I'll update the chart and include the Quality mode results later this evening.

[OWA]

Okay, I think I'm done. Here is hopefully the last update. I corrected the last chart and the 2nd chart is the Quality mode comparison although I still left Adaptive AA on for ATI and Gamma AA/Transparency AA Super Sampling for Nvidia. I never turn these off unless there is some issue.

http://img369.imageshack.us/img369/6338/widescreengaminghq9qn.jpg (http://imageshack.us)

http://img369.imageshack.us/img369/9662/widescreengamingq6dr.jpg (http://imageshack.us)

[rev.]

Hey OWA -

replaced the stock X1900XTX cooler on my card with the Zalman VF 700 and some arctic silver 5. Load temps tops about 63 deg c, idle is 36 c. I'm running max CCC of 690 / 800 with the XTX.

3dmark 03 = 20,242; 3dmark 05 = 12,484; 3dmark 06 = 5,590.

Just wondering how happy you are with yours!? Hopefully you got a good OC'r this time around!

(nana2)

I was wondering myself if the Zalman would do a better job. I my throw mine on as check it out myself. My card doesnt lik to OC with the stock cooler.

[OWA]

To go through to see the impact of HQ AF, do I just need to check the HQ AF option in CCC and leave the AF set to app control or do I need to control the AF with the CCC as well? I'd prefer to leave it at app control since some games have better performance that way.

[OWA]

I just ran them again with the HQ AF checkbox on and left the CCC AF to app control. It doesn't look like the HQ AF causes much of a hit in the games I checked so I'll just leave it always selected from now on.

Edit: Fixed typo and made it a little easier to compare.

http://img49.imageshack.us/img49/5395/wsgcomparisonuhq8ns.jpg (http://imageshack.us)]

[SH64]

Oops .. i did it aswell! :D

http://img102.imageshack.us/img102/6558/copyofdsc000650be.jpg

[saturnotaku]

Nice. :thumbsup:

[OWA]

Cool, 4th system or replacing one of the others?

[SH64]

Cool, 4th system or replacing one of the others?

For now its replacing the P4:3.4Ghz system.
so the P4 system (mobo+CPU+GPU) might form a 4th system in the near future! ;)

[fatal1ty]

'gratz - the X1900XTXs really rock!

[Lfctony]

Since everyone is posting theirs, here's mine ;)
http://www.cygaming.net/modules/PNphpBB2/files/x1900xt.jpg

[Mr_LoL]

Owa whats the image quality like on the Ati card compared to the GTX especially in AA and AF? Is there a noticeable difference as I can't tell from the images on the first page.

[OWA]

The IQ is very close most of the time so I pretty much go with the usual line on this. That is, IMO, the AA on the GTX is sometimes a little better (the TAA seems to do a better job than the Adaptive AA and the SS modes are better, if you can use them) and the AF on the ATI is sometimes better (due to the HQ AF). I almost exclusively game in High Quality mode on both systems and in the games that I play, I think they look very similiar. Of course, there are games/circumstances where you can make one or the other not look quite as good (or one of them look better depending on how you look at it :)).

In the pics that someone posted on the first page, in terms of AA/AF I think they're close. Nvidia has a sharper picture in both sets but depending on settings (AA) (maybe the game) being used, that is sometimes reversed.

I think the biggest complaint lodged over the top set is that ATI doesn't seem to be rendering all the objects in the top pic. Apparently, during the X1800XT reviews and now with the X1900XTX reviews, some of the reviewers had drivers (or maybe it's the game -- since I think it is always HL2) that would set the LOD differently on ATI cards. I didn't bother checking this time around but after that was brought up during the X1800XT reviews, I couldn't reproduce it myself. So, I don't know what is going on with that.

[fatal1ty]

I always had problems with my GTX and GTX512 not applying AA to everything in a scene. Someone had screenshots of that posted around here where a few things in the scene would not have AA applied to them.

The GTX had crisper IQ but that was what I found to cause most of my frustration because it made the edges shimmer even more. I now realize that I like the AA and IQ of the ATI over Nvidias. But to each his own.

[OWA]

Well, if you check the old threads on this you'll see most people don't agree with you. With AA, a lot of time you'll find one will apply more AA to one area of the scene while the other will apply more AA to a different area of the scene. That's why it's usually inconclusive. In one of the old threads, you were picking areas where AA looked better on the ATI side but ignoring areas in the same screenshot where Nvidia had better AA.

Then if you check the Far Cry screenshots I posted in one of the threads in the other graphics forums (a thread where a guy had switched to a nvidia card and was asking if things were worse now), you'll see where the TAA is clearly applying more AA on railings than the adaptive AA.