[CP5670]

Two GTSs would seem to be a good buy, but with SLI you need to deal with its vsync problems in the vast majority of games. That makes the framerates you see rather misleading, as games often appear to be choppy even when the framerate is high. This makes SLI good for getting okay performance with maxed out resolutions and AA settings (in situations where one card would be unusable), but useless for getting liquid smooth performance at somewhat lower settings.

[asyikarea51]

Right now I'm running a single 8600GT 256mb and I'm fairly happy with it (my previous computers were crappy, so you can say that this is the first time I've ever experienced the "maxed-out"... experience).

I'm just frustrated that I can't create a listen server in CS Source... damn "the memory could not be 'read'" error.

Wonder what would happen if someone had two 8800U's and somehow got the same problem. That'll be an expensive shocker... O_O

[Nuke]

i usually always maxed out my settings even if the computer couldnt handle it. for example i played the fs2 demo on a 120 mhz machine with 24 megs of ram and a banshee card, settings maxed

[Sandwich]

for sli performance the best config is actually a pair of gts cards. which will boost your 3dmark06 score by about 3000 - 4000 points. pretty much the largest boost of any of the 8800 sli configs when compaired to a single card setup. a pair of gtx or ultra cards will perform better, but not much better than a single card config. if i can im gonna try to swing for a secod gts card this summer.

Probably a bottleneck issue elsewhere, not the cards' fault. Even sole 8800 GTX's get noticeable performance increases when pairing them with high-end CPU's. I bet the rest of the computer's systems - any computer's systems - are unable to feed SLI 8800 GTX cards what they need to run flat out.

[Nuke]

most sli boards only run each card in 8x mode because of hardware limitations in the chipsets. however the board im gonna go with is fully capable of running in dual 16x and still havew enough left over to run a physx card in 8x as well. not going with a physics card, mainly because theyre starting to embed physics instructions into video hardware.

[Sandwich]

most sli boards only run each card in 8x mode because of hardware limitations in the chipsets. however the board im gonna go with is fully capable of running in dual 16x and still havew enough left over to run a physx card in 8x as well. not going with a physics card, mainly because theyre starting to embed physics instructions into video hardware.

Uhm, no, not most. There are some cheaper boards that have that issue, mainly the nForce 650i boards IIRC. But the 690i series supports full 2x (or 3x? I forget) PCIe x16. My board, which is an Asus P5N32-E SLI Plus hybrid of some sort, has 2 x16 and 1 x8.

[MetalDestroyer]

I think the test made by DailyTech is not a solid test. Did you wonder why the X2900XTX have the same or a little less performance than the X2900XT (where it should be less powerfull than a XTX version) ?

They didn't compare the image quality between a 8800 GTS/GTX with those X2900 XT/XTX, but just only pure performance. I recently have a 8800 GTX, it does his job perfectly except ONE thing, and this is about how the image/texture  is rendered, I mean the filtering parts. When I played GTR 2 with my old ATI 9800 Pro, the road were kicking ass. But when I receive my 8800 GTX, roads looks like crap. I'm not an ATI fanboy, but I'm just a little desappointed with nVidia. It's been a while they don't fix these filtering quality. (I don't know the right word in english) but the quality related with textures are crap in certain case and it's a little annoying when you taste the same game with an ATI card.

I think, X2900 XTX can do better performance. So, I will wait for further more test from other technical reviewer like guru3D, anandtech, tom'shardware, or hardware.fr.

[Nuke]

most sli boards only run each card in 8x mode because of hardware limitations in the chipsets. however the board im gonna go with is fully capable of running in dual 16x and still havew enough left over to run a physx card in 8x as well. not going with a physics card, mainly because theyre starting to embed physics instructions into video hardware.

Uhm, no, not most. There are some cheaper boards that have that issue, mainly the nForce 650i boards IIRC. But the 690i series supports full 2x (or 3x? I forget) PCIe x16. My board, which is an Asus P5N32-E SLI Plus hybrid of some sort, has 2 x16 and 1 x8.

oh,duh total brainfart there. the newer chipsets do all do the acceleration in dual16.

[CP5670]

I think the test made by DailyTech is not a solid test. Did you wonder why the X2900XTX have the same or a little less performance than the X2900XT (where it should be less powerfull than a XTX version) ?

They didn't compare the image quality between a 8800 GTS/GTX with those X2900 XT/XTX, but just only pure performance. I recently have a 8800 GTX, it does his job perfectly except ONE thing, and this is about how the image/texture  is rendered, I mean the filtering parts. When I played GTR 2 with my old ATI 9800 Pro, the road were kicking ass. But when I receive my 8800 GTX, roads looks like crap. I'm not an ATI fanboy, but I'm just a little desappointed with nVidia. It's been a while they don't fix these filtering quality. (I don't know the right word in english) but the quality related with textures are crap in certain case and it's a little annoying when you taste the same game with an ATI card.

I think, X2900 XTX can do better performance. So, I will wait for further more test from other technical reviewer like guru3D, anandtech, tom'shardware, or hardware.fr.

Chances are that this is some sort of driver bug in that particular game, of which there are currently many. The only real difference between the various cards has been their trilinear and AF quality (at the same settings), and the 8800 cards have better AF than anything since the GF4 series.

Although I (and many others) agree with you that the XTX benchmarks look dodgy and don't make any sense when compared to their own XT scores. Some sites are throwing around May 10 as a launch date, so hopefully we'll see some real results soon.

[IceFire]

At least they got rid of the first X. Although replacing it with that other overused acronym, HD, isn't much better.

That said...the last few reviews I read...especially the one from Hard OCP seems to recommend the BFG 8600GTS as it pulls out ahead of the X1950Pro in nearly all of their tests.

That review should be taken with a grain of salt, as its conclusions are contrary to those of almost every other site, and it has come in for a lot of criticism for that. It's not the first time that [H] has come up with something off the wall like that either.

In any case, the 8600GTS's competitor  (at the same price) is actually the X1950XT, not the Pro. The X1950XT essentially demolishes it. The Pro is priced lower, at the same level as the GT model.

HardOCP always takes flak for their reviews. Thats why I enjoy reading them the most

They have their own methods...which they have explained pretty clearly...and they seem to make a fair bit of sense to me.

[CP5670]

It's funny that you bring it up now, since they've just made another howler with their HD2900XT review. The card is generally a bit underwhelming, but some of [H]'s results are once again contrary to those of all the other major sites. Although it's nice to see that they included a proper comparison at equal settings this time, and their graphs are much more useful than the single number readings on most other sites.

I (and many others) think the best graphics reviews were actually at Rage3D, but their main reviewer left, so we won't be seeing much more of those. Apart from that, Techreport and Xbit are generally thorough and reliable in my experience, and Anandtech is pretty good as well.

[IceFire]

Yep I tend to read all of those too.  If you're thinking of buying a card like that...or any card...its worthwhile to check out all the sites that do testing and see where they fit into the grand scheme of things.  HardOCP just does their testing differently...they logically laid out why they do what they do...and I appreciate their going against the grain in not doing apples-to-apples testing.  Not everyone is going to like it...but its different than the other reviews you read.  Generally means I can form a decent opinion for myself based on everything I'm reading...in the end

[Scooby_Doo]

I think the phrase "too little too late" applies here.

[CP5670]

Yep I tend to read all of those too.  If you're thinking of buying a card like that...or any card...its worthwhile to check out all the sites that do testing and see where they fit into the grand scheme of things.  HardOCP just does their testing differently...they logically laid out why they do what they do...and I appreciate their going against the grain in not doing apples-to-apples testing.  Not everyone is going to like it...but its different than the other reviews you read.  Generally means I can form a decent opinion for myself based on everything I'm reading...in the end

In general, you're right, but in H's case, it's not so much their testing methods as their results that are drawing suspicion. Their 8600GTS review is a case in point, in which they have it beating the X1950 Pro and 7900GS across the board while practically every other site out there shows just the opposite. They have a history of doing this sort of thing too. Their Core 2 review was exactly like that and we're seeing it again with the HD2900 review. When a site is consistently publishing results and conclusions that are contrary to all the others, it shouldn't be given much credibility.

I think the phrase "too little too late" applies here.

Exactly. It's not that bad a card, but given that it's six months late and that the drivers are currently in a sorry state, there seems to be little reason to get it over the 640 GTS at this point.

[IceFire]

I'd have to look very closely to see what was happening but HardOCP doesn't do the same tests as most of the other sites do...its not apples to apples.  So its pretty hard to compare flat out.  Again...my philosophy is that if I want to buy a piece of hardware for myself or recommend something to someone else you should read as many reviews as possible with as many methodologies as possible.  The cards are pretty complex these days and its obvious that the newest budget/mainstream level cards have some neat features that sometimes put them on top and sometimes not and depending on the method involved and the features enabled you can get some different results.

Yes the 2900 XT is several months too late and simply not up to what I think was being expected of it.  Its still a top notch performer and when the die shrink comes I think we may see it jump up a few notches or at least drop the thermal requirements a bit.  That'd be better than anything else I think.  I'm still far more interested in the 2600 XT...the top level cards are for smaller numbers of people while the mainstream cards are what the majority are using so thats where the real battleground, even if its not acknowledged, is.