Early Fermi Benchmarks

[GotNoRice]

Thats generally how it works... like the guy that you quoted said "Given that graphics power about doubles every generation (18 months)" You're comparing SLI (so basically about double the graphics power) to a single card.

Guess what, my 8800GTS640's in SLI were extremely competitive with the GTX200 series, omg!

Good thing you cut off the part of the quote that explained my reasoning

It's not that the 5870 isn't a good card but it simply did not represent the level of improvement over the previous generation that we have seen in the past. When the 4870 came out it could best dual 3870x2's, whereas a 5870 is only as fast as a single 4870x2. If Nvidia hadn't been stumbling over Fermi for the last year things would have been different. It's easy to have a "great" product when it exists mostly in a vacuum.

Not complaining, I think it's great my old tech is still competitive. Just makes me wonder, what if ATI never released the 9700 Pro but Nvidia managed to get their FX5800 out of the door? Would people have considered the FX5800 a good card also simply due to lack of competition?

 

[maxius]

at 275 watts i will pass on the dust buster 2.0

 

[Mr. K6]

Good thing you cut off the part of the quote that explained my reasoning

It's not that the 5870 isn't a good card but it simply did not represent the level of improvement over the previous generation that we have seen in the past. When the 4870 came out it could best dual 3870x2's, whereas a 5870 is only as fast as a single 4870x2. If Nvidia hadn't been stumbling over Fermi for the last year things would have been different. It's easy to have a "great" product when it exists mostly in a vacuum.

Not complaining, I think it's great my old tech is still competitive. Just makes me wonder, what if ATI never released the 9700 Pro but Nvidia managed to get their FX5800 out of the door? Would people have considered the FX5800 a good card also simply due to lack of competition?

That's because your reasoning was ridiculous in the first place. The increase in performance of the 5870 vs. the 4870 was even higher than the 4870 vs. the 3870, never mind the fact that they cut power consumption, improved cooler performance/noise, and added some great new features (DX11, Eyefinity etc.). You clearly live in a dream world that doesn't coincide with reality. The fact that multi-GPU drivers have improved in the last year and a half (wow, who would have thought?) doesn't negate the quality of released parts. To clearly point it out:

There are quite a few cases where you can best the current gen even using old tech. Something like a pair of GTX260's is still extremely competitive with a 5870 and even my quad crossfire setup using cards nearly 2 years old gives me 5970 level performance in games that scale well.

And compared to a 5970, your 4870X2's use anywhere from 50-60% more power, don't overclock nearly as well, lack all of those great new features like DX11 and Eyefinity, and cost nearly twice as much. I also bolded the greatest part of that statement - games that scale well with four GPUs; good luck with that.

Compare that to when the 4870 was released. In many cases a single 4870 was faster than dual 3870x2's in quad crossfire.

And there were just as many cases where a single 3870X2 stomped the 4870, much more so than the 4870X2 ever did the 5870. Multi-GPU support was spotty then, such is life.

 

[GotNoRice]

That's because your reasoning was ridiculous in the first place. The increase in performance of the 5870 vs. the 4870 was even higher than the 4870 vs. the 3870.

That simply is not the case. The 4870 was easily more than twice as fast as the 3870. A 3870x2 will be beat by a 4850 in most cases. That makes sense when you look at the architecture. 320 stream processors per 3870 means that even a 3870x2 has less stream processors than a single 4870. On the other hand the 4870x2 and 5870 have the same number of stream processors.

And compared to a 5970, your 4870X2's use anywhere from 50-60% more power, don't overclock nearly as well, lack all of those great new features like DX11 and Eyefinity, and cost nearly twice as much.

I can still play DX11 games on the 4870x2 as discussed before, and not sure what you mean by cost twice as much; I got both of my cards for about the price of a single 5870, which is certainly cheaper than a 5970.

I also bolded the greatest part of that statement - games that scale well with four GPUs; good luck with that.

lol you say that like it's hard to find games that scale well. Most games scale well these days, especially new ones (which is where I want the best performance). Games such as Battlefield: Bad Company 2 had nearly flawless crossfire support with amazing scaling from day 1. Beyond that, the recent changes ATI made to the drivers to allow crossfire profile updates outside of driver updates will only make things better. Pretty much the only games I've found that don't scale well with 4 GPUs are old or niche games where I probably wouldn't notice any difference past 2 GPUs anyway.

And there were just as many cases where a single 3870X2 stomped the 4870, much more so than the 4870X2 ever did the 5870.

Not from the benchmarks I've seen.

 

[akotlar]

I don't honestly think the 5870's were that impressive from a performance point of view. Lack of competition along with DX11/eyefinity hype is what made the card a success. There are quite a few cases where you can best the current gen even using old tech. Something like a pair of GTX260's is still extremely competitive with a 5870 and even my quad crossfire setup using cards nearly 2 years old gives me 5970 level performance in games that scale well. Compare that to when the 4870 was released. In many cases a single 4870 was faster than dual 3870x2's in quad crossfire.

I agree. 5870 is a good card, but it is not a massive leap over the previous generation in terms of performance. It looks fantastic compared to Fermi though, so now opinion is heavily in its favor.

 

[heatlesssun]

It looks fantastic compared to Fermi though, so now opinion is heavily in its favor.

I think that's a question that's still up for debate. I'd say as it stands now Fermi does seem to be the card that's going to have a bit more longevity to it, it is overall faster than a 5870.

Not saying that the 5870 won't still be a good card but if the GTX 400s are priced competitively the 5870 may not look quite fantastic is all I'm saying.

 

[pasta4u]

I agree. 5870 is a good card, but it is not a massive leap over the previous generation in terms of performance. It looks fantastic compared to Fermi though, so now opinion is heavily in its favor.

I disagree. Its as fast as the dual chip cards from last generation , it uses much less power than those cards and it supports a whole new api. Also its brand new and drivers need to mature. The gtx 295 is based on hardware that has been around since the gtx 8800 which is what 3 years old now? So drivers are very mature.

The 5870 will become faster as time comes on while the gtx 295 will largely stay the same.

Since SLI / Crossfire came around its very hard for a new generation to increase performance over the last generation easily. You'll see the same problem with the gf100. It wil lbe faster than the 5870 in its gtx 480 guise , but will use much more power and produce much more heat whle being more expensive.

But i'm more interested in seeing how the gtx 480 competes with a 2 gig 5870. It seems in many games higher resolutions are purely framebuffer limitedon the 5870

 

[NKD]

I don't honestly think the 5870's were that impressive from a performance point of view. Lack of competition along with DX11/eyefinity hype is what made the card a success. There are quite a few cases where you can best the current gen even using old tech. Something like a pair of GTX260's is still extremely competitive with a 5870 and even my quad crossfire setup using cards nearly 2 years old gives me 5970 level performance in games that scale well. Compare that to when the 4870 was released. In many cases a single 4870 was faster than dual 3870x2's in quad crossfire.

oh man why do I hear stuff like this over and over, where you compare 4 cards vs. 2 and 2 cards to 1. 2 times the last gen performance only applies to one card. now when the 6870 comes out and lets say it is time and a half faster, and are you gonna say oh i get the same performance with three hd 5870's. Fail.

 

[NKD]

by the way 3870's sucked period. they were the worst of ati's cards basically a shrunk hd 2900, and 4870 was basically hd 3870 should have been, and quad crossfire never scaled that well with the 3870's, the cards just didn't have it. and ofcourse you can play dx11 games on dx10 hardware, does that mean that your hardware is using dx11 features, didn't think so. How can someone justify playing dx11 games on dx10 hardware and somehow think it is going to look the same.

 

[pasta4u]

by the way 3870's sucked period. they were the worst of ati's cards basically a shrunk hd 2900, and 4870 was basically hd 3870 should have been, and quad crossfire never scaled that well with the 3870's, the cards just didn't have it. and ofcourse you can play dx11 games on dx10 hardware, does that mean that your hardware is using dx11 features, didn't think so. How can someone justify playing dx11 games on dx10 hardware and somehow think it is going to look the same.

The 3870 was priced correctly. The 3870x2 was a good car on par with other $400 cards.

The 4x00 series came back with a vengance and now the 5x000 series is top of the line and it seems like nvidia will need an even bigger , more power hungry chip to beat it.

 

[sweetreddyd]

oh man why do I hear stuff like this over and over, where you compare 4 cards vs. 2 and 2 cards to 1. 2 times the last gen performance only applies to one card. now when the 6870 comes out and lets say it is time and a half faster, and are you gonna say oh i get the same performance with three hd 5870's. Fail.

I wouldn't pay much mind to that guy, trying real real hard to justify his quad-fire 4870s. He's in another thread somewhere trying to convince a guy to buy 4870x2 instead of a 5870.

 

[Wenglish]

I think that's a question that's still up for debate. I'd say as it stands now Fermi does seem to be the card that's going to have a bit more longevity to it, it is overall faster than a 5870.

Not saying that the 5870 won't still be a good card but if the GTX 400s are priced competitively the 5870 may not look quite fantastic is all I'm saying.

The 5870 might not look fantastic right now, but what if ATI lops $100 off its MSRP? They've been unopposed and make a much smaller chip than the Fermi. They have to have better yields.

Do I think that will happen? No, not really. But if nVidia is somehow aggressive enough with their pricing and better enough with their performance (two very big ifs), then ATI could simply slot their cards down a price-notch or two. And then the 5870 looks fantastic again, at least from a price/performance viewpoint (and those are the glasses most users wear).

 

[DualOwn]

idk why some people think that ati is going to drop prices of 5000 series when Fermi comes out, unless the 470 sells cheaper than the 5870 but that's very unlikely.

ati is selling their cards pretty good right now even with prices that are over the MSRP, the 5870 from Saphire on newegg is going for $440 right now and it keeps selling out.

 

[heatlesssun]

The 5870 might not look fantastic right now, but what if ATI lops $100 off its MSRP?

Not saying that the 5870 won't still be a good card but if the GTX 400s are priced competitively the 5870 may not look quite fantastic is all I'm saying.

I agree with you. If ATI adjusted their prices as you suggest nVidia could respond though I understand that Fermi is more exspensive than Cypress and nVidia can't afford to get into a price war with ATI at the moment with Fermi.

 

[Wenglish]

idk why some people think that ati is going to drop prices of 5000 series when Fermi comes out, unless the 470 sells cheaper than the 5870 but that's very unlikely.

ati is selling their cards pretty good right now even with prices that are over the MSRP, the 5870 from Saphire on newegg is going for $440 right now and it keeps selling out.

Like I said, I don't think that it will happen. ATI is making a killing and they have been and unless nVidia forces them to lower their prices, they won't. I was just saying that if Nvidia pressured ATI, it wouldn't be unreasonable for ATI to pressure right back.

And, given the sizes of the chips and the last few months of non-competition, ATI is in a much better position to get into a price war.

 

[rocketr2]

Fermi launches on the 26th right? Thats only 2 weeks away.

 

[YeuEmMaiMai]

by the way 3870's sucked period. they were the worst of ati's cards basically a shrunk hd 2900, and 4870 was basically hd 3870 should have been, and quad crossfire never scaled that well with the 3870's, the cards just didn't have it. and ofcourse you can play dx11 games on dx10 hardware, does that mean that your hardware is using dx11 features, didn't think so. How can someone justify playing dx11 games on dx10 hardware and somehow think it is going to look the same.

u got it wrong

the 2900 series was the worst one. The 3870 was the correction to the 2900 series everything since then has been correct

 

[akotlar]

I disagree. Its as fast as the dual chip cards from last generation , it uses much less power than those cards and it supports a whole new api. Also its brand new and drivers need to mature. The gtx 295 is based on hardware that has been around since the gtx 8800 which is what 3 years old now? So drivers are very mature.

The 5870 will become faster as time comes on while the gtx 295 will largely stay the same.

Since SLI / Crossfire came around its very hard for a new generation to increase performance over the last generation easily. You'll see the same problem with the gf100. It wil lbe faster than the 5870 in its gtx 480 guise , but will use much more power and produce much more heat whle being more expensive.

But i'm more interested in seeing how the gtx 480 competes with a 2 gig 5870. It seems in many games higher resolutions are purely framebuffer limitedon the 5870

Unfortunately for Fermi, the 5870 is roughly 2x as fast as the previous generation as well (even the 295). So Fermi showing up 6 months later, and giving a 10-15% average benefit over that (ok, when the 5870 ISN'T running out of framebuffer, a situation Nvidia will heavily promote) just isn't good enough, unless it is at a lower price point. I'm not going to purchase Nvidia just because they have PhysX. That is worth 0 dollars to me. It would be the extra gram that tips the scale in Nvidia's favor for a part that is at the same price point with greater performance (significantly greater, like 30-40% faster, a difference I'll notice without a frame counter). But at $500 + and ~10% in meaningful situations? No thanks.

And from an engineering standpoint, it's a disaster. It's worse performance/transistor than GT200 was, and that was already a step back for Nvidia. 275W, 3.1B transistors, and what, 10% performance improvement over a part that's 50% smaller and runs at a TDP 85W less? And came out 6 months ago? Yowsers, that's a big loss for a company that holds all the cards. Oh great, it offers a 4x faster setup rate, and 4x faster tessellation. Too bad it doesn't have the shading power to process any of the additional math required to shade those polygons! Never mind that no sane dev is going to boost tess rates to the level where Nvidia shows a marked improvement. You have far more important targets this gen that take up all of your ALU's +++. Namely lighting, G.I, indirect shadows, multiple bounces, convincing subsurface effects, etc. They're trying to make increased polygon budgets this panacea for modern graphics, but they're not. We could have gone for high triangle rates before this time. There was nothing stopping ATI or Nvidia from implementing a 1+ triangle setup when they were at half rates, or going 2+ when they were at 1, or going overboard with Vertex shaders to increase triangle rendering capability. They didn't because it's pointless. The far more important target is the quality of each pixel.

 

[Mr. K6]

That simply is not the case. The 4870 was easily more than twice as fast as the 3870. A 3870x2 will be beat by a 4850 in most cases. That makes sense when you look at the architecture. 320 stream processors per 3870 means that even a 3870x2 has less stream processors than a single 4870. On the other hand the 4870x2 and 5870 have the same number of stream processors.

No, it simply was the case and you need to go read and clue yourself in:
http://www.anandtech.com/video/showdoc.aspx?i=3341&p=1
http://www.anandtech.com/video/showdoc.aspx?i=3643&p=1

Note that the 5870 is actually compared to a 4870 1GB, which appeared later after the initial 4870 launched, although it still bests the 4870 by the same or greater margins as the 4870 did over the 3870. Although you should have done the research yourself instead of continuing to be ignorant.

I can still play DX11 games on the 4870x2 as discussed before, and not sure what you mean by cost twice as much; I got both of my cards for about the price of a single 5870, which is certainly cheaper than a 5970.

You're not playing any of those games in DX11 though, your loss. And where's the link to the store that sells 4870X2's for $200 each?

lol you say that like it's hard to find games that scale well. Most games scale well these days, especially new ones (which is where I want the best performance). Games such as Battlefield: Bad Company 2 had nearly flawless crossfire support with amazing scaling from day 1. Beyond that, the recent changes ATI made to the drivers to allow crossfire profile updates outside of driver updates will only make things better. Pretty much the only games I've found that don't scale well with 4 GPUs are old or niche games where I probably wouldn't notice any difference past 2 GPUs anyway.

And that's exactly my point, they aren't going to scale nearly as well as two 5870's. So enjoy your driver dependencies and all the headaches that come with it.

I wouldn't pay much mind to that guy, trying real real hard to justify his quad-fire 4870s. He's in another thread somewhere trying to convince a guy to buy 4870x2 instead of a 5870.

Which is why I'm stopping that shit here; not in my thread.

I think that's a question that's still up for debate. I'd say as it stands now Fermi does seem to be the card that's going to have a bit more longevity to it, it is overall faster than a 5870.

Not saying that the 5870 won't still be a good card but if the GTX 400s are priced competitively the 5870 may not look quite fantastic is all I'm saying.

Exactly. It's really going to come down to the details on this one. We have an idea about performance, and the cards are competitive. But the overclockability, power consumption, noise/aesthetics, and features as well as their implementation are all important factors. Most importantly however, the price is going to have the greatest weight in establishing Fermi's worth.

 

[DualOwn]

Unfortunately for Fermi, the 5870 is roughly 2x as fast as the previous generation as well (even the 295). So Fermi showing up 6 months later, and giving a 10-15% average benefit over that (ok, when the 5870 ISN'T running out of framebuffer, a situation Nvidia will heavily promote) just isn't good enough, unless it is at a lower price point. I'm not going to purchase Nvidia just because they have PhysX. That is worth 0 dollars to me. It would be the extra gram that tips the scale in Nvidia's favor for a part that is at the same price point with greater performance (significantly greater, like 30-40% faster, a difference I'll notice without a frame counter). But at $500 + and ~10% in meaningful situations? No thanks.

And from an engineering standpoint, it's a disaster. It's worse performance/transistor than GT200 was, and that was already a step back for Nvidia. 275W, 3.1B transistors, and what, 10% performance improvement over a part that's 50% smaller and runs at a TDP 85W less? And came out 6 months ago? Yowsers, that's a big loss for a company that holds all the cards. Oh great, it offers a 4x faster setup rate, and 4x faster tessellation. Too bad it doesn't have the shading power to process any of the additional math required to shade those polygons! Never mind that no sane dev is going to boost tess rates to the level where Nvidia shows a marked improvement. You have far more important targets this gen that take up all of your ALU's +++. Namely lighting, G.I, indirect shadows, multiple bounces, convincing subsurface effects, etc. They're trying to make increased polygon budgets this panacea for modern graphics, but they're not. We could have gone for high triangle rates before this time. There was nothing stopping ATI or Nvidia from implementing a 1+ triangle setup when they were at half rates, or going 2+ when they were at 1, or going overboard with Vertex shaders to increase triangle rendering capability. They didn't because it's pointless. The far more important target is the quality of each pixel.

so what are you saying? that Fermi will be slower than 5870 with dx11 games in the future?

imo, i think you got it all backwards. I believe that Fermi has a perfect hardware to run heavy tessellation with high quality shading.

 

[shansoft]

so what are you saying? that Fermi will be slower than 5870 with dx11 games in the future?

imo, i think you got it all backwards. I believe that Fermi has a perfect hardware to run heavy tessellation with high quality shading.

its all your claim, nothing else....

 

[akotlar]

so what are you saying? that Fermi will be slower than 5870 with dx11 games in the future?

imo, i think you got it all backwards. I believe that Fermi has a perfect hardware to run heavy tessellation with high quality shading.

No, what I'm saying is that indiscriminately raising your triangle throughput rate, and your tessellation rate, while not taking into account the extra processing power you'll need to perform calculations on those triangles ends up being an exercise in futility. Maybe it will turn out that Fermi has a better balance between math capability and tessellation + setup than Cypress does, I don't know. But at this point it seems like they're blowing the issue up because it was convenient to them to do 4 setup and 4 tessellation stages when they changed the triangle execution pipeline to an out of order design. And the reason for that was most certainly for GPGPU performance.

From what I recall, out of order execution for the polymorph engine (tessellator and setup) doesn't really confer many benefits for graphics rendering, with the possible exception of small/overlapping triangles (not sure on this last bit). It was clearly done with motivations other than consumer graphics, since the rest of the pipeline is in order and DX11 is written for in-order execution (meaning most execution happens in order regardless, so it doesn't help much to have the end piece execute out of order; scheduling still needs to happen the same way).

 

[DualOwn]

No, what I'm saying is that indiscriminately raising your triangle throughput rate, and your tessellation rate, while not taking into account the extra processing power you'll need to perform calculations on those triangles ends up being an exercise in futility. Maybe it will turn out that Fermi has a better balance between math capability and tessellation + setup than Cypress does, I don't know. But at this point it seems like they're blowing the issue up because it was convenient to them to do 4 setup and 4 tessellation stages when they changed the triangle execution pipeline to an out of order design. And the reason for that was most certainly for GPGPU performance.

From what I recall, out of order execution for the polymorph engine (tessellator and setup) doesn't really confer many benefits for graphics rendering, with the possible exception of small/overlapping triangles (not sure on this last bit). It was clearly done with motivations other than consumer graphics, since the rest of the pipeline is in order (meaning most execution happens in order regardless, so it doesn't help much to have the end piece execute out of order; scheduling still needs to happen the same way).

i personally believe that 5870 will be just as fast if not faster than Fermi in dx9 and dx10 games but not in dx11 games. But this is mostly due to the higher core clock on 5870, i will be doing water cooling when it comes out on Fermi, so i should be good in all dx modes.

 

[akotlar]

i personally believe that 5870 will be just as fast if not faster than Fermi in dx9 and dx10 games but not in dx11 games. But this is mostly due to the higher core clock on 5870, i will be doing water cooling when it comes out on Fermi, so i should be good in all dx modes.

I don't see why the 5870 would be slower except in heavy tessallation frames. In fact, on paper, it should be a lot faster for anything math related, so I'd be surprised if the increased tessellation & setup throughput did much more than offset the disparity in math performance.

 

[NKD]

u got it wrong

the 2900 series was the worst one. The 3870 was the correction to the 2900 series everything since then has been correct

correct, but 4870 was the real deal, while 3870 was a correction it was still was not enough to compete with the nvidia when it comes to performance.

 

[Deleted member 174368]

i personally believe that 5870 will be just as fast if not faster than Fermi in dx9 and dx10 games but not in dx11 games. But this is mostly due to the higher core clock on 5870, i will be doing water cooling when it comes out on Fermi, so i should be good in all dx modes.

You do realize Tessellation isn't the only DX11 feature, right.

 

[DualOwn]

You do realize Tessellation isn't the only DX11 feature, right.

i know, but its the best feature.

 

[NKD]

lol, it is funny when people compare clock speeds with ati and nvidia, what's fair is fair, I have had nvidia and both ati, and who knows I might buy fermi for another build if it is priced right, but when you compare clock speeds you can't just go out and say oh ati's clock speed is way higher than nvidia's, what about nvidia's shader clock, it is independent and runs at over 1.2ghz for the shader clock,and ati's shader clock is same as core clock speed, so ati could counter someone's claim by simply sayin that oh our shader clock runs slower then theirs and we can still match the performance. It is totally irrelevent.

 

[RamonGTP]

i know, but its the best feature.

Would it be the best feature if Fermi sucked at it?

 

[Brackle]

Would it be the best feature if Fermi sucked at it?

No it would be how well it scores in 3dmark over an ATI product.

 

[Michaelius]

Heavy Tessellation:
GTX 480: 43 fps / HD 5970: 29 fps

im very impressed, nice find.

it's like 9700 vs 5900 fx repeating itself but the other way.

5900 when released was just as fast if not faster in games that were available then it only started to be noticeably slower once real dx9 games started to be released.

 

[RamonGTP]

FX was never impressive... In anything.

 

[maximus87]

If it (Fermi) resembles anything it's R600. I'm sure ATi will gain some ground then lose it again to something else when Nvidia shows up with their next "G80".

It's a vicious cycle that will continue to repeat.

 

[XMAN245]

If it (Fermi) resembles anything it's R600. I'm sure ATi will gain some ground then lose it again to something else when Nvidia shows up with their next "G80".

It's a vicious cycle that will continue to repeat.

About the only thing it has in common with R600 is being late to the party. aside from that, performance is there unlike the R600.

 

[Brackle]

About the only thing it has in common with R600 is being late to the party. aside from that, performance is there in "uniengine".

Fixed your Typo

 

[XMAN245]

Fixed your Typo

R600 stunk up the joint so bad at launch X1950XTs could beat the card and the best it could do performance wise was match the 8800GTS 640. GTX470 is on par with the 5870 and the GTX480 is faster. So again, the only thing it has in common with R600 is being late.

 

[Brackle]

R600 stunk up the joint so bad at launch X1950XTs could beat the card and the best it could do performance wise was match the 8800GTS 640. GTX470 is on par with the 5870 and the GTX480 is faster. So again, the only thing it has in common with R600 is being late.

Do you have proof that the GTX470 is as fast as the 5870?

Real reviews? If you got the card shows us man! show us all this stuff you talk about!!

Proof is what we ask, not your opinion

 

[XMAN245]

Do you have proof that the GTX470 is as fast as the 5870?

Real reviews? If you got the card shows us man! show us all this stuff you talk about!!

Proof is what we ask, not your opinion

You are just as likely to get solid numbers from Kyle while he is working on his review. Good luck with that.

 

[Brackle]

You are just as likely to get solid numbers from Kyle while he is working on his review. Good luck with that.

Well If you think Kyle isn't fair in reviews why even visit this website? To harrass people who know ATI has a really good video card?

Or to just bash ATI any chance you get?

Kyle has one of the best reviews methods in the industry....There is a reason [H]ardocp has been around for over 10+ years.

Again, how about those Nvidia Drivers that kill Video cards?....any comments? ATI's fault?

 

[XMAN245]

Well If you think Kyle isn't fair in reviews why even visit this website? To harrass people who know ATI has a really good video card?

Or to just bash ATI any chance you get?

Kyle has one of the best reviews methods in the industry....There is a reason [H]ardocp has been around for over 10+ years.

Again, how about those Nvidia Drivers that kill Video cards?....any comments? ATI's fault?

I never said anything in regards to Kyle's reviews, I actually prefer them to others as they play the games, I said good luck getting him to give you performance numbers WHILE he WAS WORKING on his review.

As to teh driver issue, it sucks for those it hit, but they stepped up to the plate to help those affected.