Nvidia GeForce GTX 590 3 GB Review: Firing Back With 1024 CUDA Cores

[digitalw]

it is importand review, it would be nice to see with new AMD pre-release drivers Catalist 11.4?

>>>> from AMD site

 

[digitalw]

new drivers, somehow, missed in previous post.

http://support.amd.com/us/kbarticles/Pages/Catalyst114earlypreview.aspx

Up to 23% performance improvement in Aliens vs. Predator
Up to 22% performance improvement in Batman: Arkham Asylum
Up to 15% performance improvement in Battleforge
Up to 72% performance improvement in Civilization V
Up to 59% performance improvement in Call of Duty: Black Ops
Up to 26% performance improvement in FarCry 2
Up to 35% performance improvement in Lost Planet 2
Up to 30% performance improvement in Metro 2033
Up to 11% performance improvement in S.T.A.L.K.E.R.: Call of Pripyat
Up to 19% performance improvement in Chronicles of Riddick: Assault on Dark Athena

 

[cangelini]

[citation][nom]kwezzy[/nom]not impressed at all by the conclusion tom have drawn. the review at hardocp was insightful and truly reflective of what these cards should do. lower res for these cards is just plain dumb. anyone who buys these cards has a multi monitor config or 1600p monitors! nvidia is not the king of the hill. thats a fact![/citation]

Resolutions from 1680x1050 to 5760x1080 were explored, including 2560x1600. If there were additional configurations you missed, please feel free to leave constructive feedback!
Thanks,
Chris

 

[cangelini]

[citation][nom]masterofevil22[/nom]"..it wins because I can put a GeForce GTX 590 in my workstation, which doubles as my gaming machine, and not hear it. It’s quiet enough to use."Contradicting yourself in the same sentence; in the conclusion no less... wow. I bet two 590's are significantly louder than my two 6850's and trust me people, "quiet enough to use" means someone either has there hearing aid turned down or the music/game/movie turned UP.[/citation]

There's no contradiction there, unless you cited the wrong text? Please check out the videos of noise! Two 590s probably are louder than your 6850s. What's your point?
Best, Chris

 

[cangelini]

[citation][nom]dlagrew3899[/nom]So why use benchmarks without maxing AA and AF at max?? If I spend that much money on 'the world's fastest graphics cards' I don't want to have to turn any settings down. This is just another case of tomshardware favoring benchmarks to make Nvidia products look better than they are. Hats off to Nvidia and ATI though for two amazing graphics cards! Can't wait for the shrink to 28nm![/citation]

Benchmarks are included with AND without AA/AF. Enjoy the copious amount of data that took a week straight of benchmarking to generate! =)
Best,
Chris

 

[lemlo]

I really enjoyed this article as well. Something I would have like to see was an overclocking section. AMD didn't downclock their gpu's nearly as much as the gf110's on this board and it would interesting to see with fan speeds ramped up what both these dual gpu bad boys can do. I bet the 590 has some massive oc overhead.

 

[Marcus52]

After their little video thing, I frankly expected more. I expected a 6990 killer. Kudos to them for using complete 110s though, maybe it will give the serious liquid cooling enthusiasts enough to work with so they can crank up the clocks to "decent" speeds. Full 580 power times 2 in a single slot would be something to talk about.

Nice job Nvidia, thanks for the thorough article with WoW testing, Chris. Would have been nice to see minimum frame rates.

 

[Rizlla]

So the best setup to keep both parties quiet and happy would be a hybrid setup of 1 HD6990 and 1 GTX590.
Then you have the performance of both cards for all games.
It would be very interesting to see the setup, that is if it could be done.

 

[Guest]

[citation][nom]foogoo[/nom]
I'd have to do some more research on weather or not the tech is there to make a single chip twice as powerful as the 580 or 6970 but i couldn't see why not. They already put 2 gpus on the same board. Why couldn't they just make a single gpu twice the size of these? If someone could explain why they couldn't id like to know.[/citation]

Sure. I'll use Nvidia as example, due to their woes with Fermi, but everything also applies to AMD as well.

A single wafer coming off TSMC has 416 "candidates" for a single GF110 chip. There are 3 billion transistors in a GF110 die. Which means that some of the transistors will be defective, in one way or another - not work at all, not able to work at a wanted frequency, or that specific transistor uses too much power and therefore overheats, etc. Not every chip will even exhibit problems, either. If you go back to GTX480/470/465, they all used the same GF100 chip, but, due to massive parallelism of video cards, the defective portion can be turned off, exactly what we see in those cards (coupled with the reduction in operating frequency). Now, to create a dual-580 chip, you need double the transistors needed originally, b/c all functional parts are doubled. But with doubling of the complexity of the chip, the chances of something going wrong quadruples, if it is 4 times as complex then the chances of failure are 16 times higher, etc. So, let say you are yielding 300 workable (as in "not completely unusable", some of these become GTX580, some become GTX570, depending on what dies are working 100% as intended) dies from those 416. Now the size of new chip also doubles. Wafer doesn't change, and we are left with a 213 new candidates, but instead of getting 150 workable dies (1/2 of 300) we are down to ~40 workable dies, far cry from 150 pairs of 580 dies. Cost of wafer is the same, but the yields drop from 70% down to 18%. Nvidia is making unprofitable GPU which still has the same power consumption/heat as 2 separate GPUs. End result: Gotta wait for a new manufacturing node to address power/heat issues, do a LOT of engineering, and then release a single-chip card with double performance. If you remember, HD5870 is almost exactly double the HD4870 in shaders, ROPs, performance (same 256-bit bus did not help performance, hence "almost"), but it took a new node to achieve.

Peace, hope it's not too much to read!

 

[silverblue]

[citation][nom]silverblue[/nom]Ryan Smith was using a 2600 clocked at 4GHz[/citation]

Sorry... it was an i7-920 at 3.33GHz. Not sure why I thought it was a Sandy Bridge unless it was originally typed as such or I just made an assumption based on another review.

 

[renz496]

to me the result is draw for both camp each with their own advantage. but of course fanboys from each camp will put their own argument to defend their preferred brand . alright that sums it up. waiting the next bout in 28nm realm

 

[srgess]

[citation][nom]rizlla[/nom]So the best setup to keep both parties quiet and happy would be a hybrid setup of 1 HD6990 and 1 GTX590. Then you have the performance of both cards for all games.It would be very interesting to see the setup, that is if it could be done.[/citation]
actually you can, gtx590 for phyx and 6990 for video lol.
also gtx590 is a good future proof for 3d gaming 3d vision > 3DHD ? but yeah 2 card are high epeen flex card to buy.

 

[silverblue]

Lucid FTW in that case. However, imagine the price of the PSU needed to keep that lot fed.

 

[masterofevil22]

Anandtech has it right. The 590 and 6990 trade blows. The 6990 will get better fps in some games and the 590 will get better fps in others.

While Nvidia gets the nod with respect to air flow "volume" (level of sound, not the actual amount of air being moved), Ati has forged a top performer with lower power draw. Both cards offer compelling features and performance. In the end it will come down to consumer preference, case air flow and the ability to power these bad boys.

Still though, if your case, wallet and PSU can handle it, it looks like a "true" quad SLI or CrossfireX remains the best route for ultimate performance. Giving each GPU room to breath and plenty of power means fewer sacrifices.. period

 

[verbalizer]

as stated earlier, I'm cool with a GTX580 and then SLI them; sit back and game for a year or so and then see what happens..

 

[cburke82]

Benchmarks are included with AND without AA/AF. Enjoy the copious amount of data that took a week straight of benchmarking to generate! =)
Best,
Chris

We all saw the AA/AF numbers. I think the point is that toms basicly says the 590 wins by just a little. BUT if you look at the 1080p and up number with AA/AF on its not close to winning as it looses most benchmarks at those settings. So I can only speak for myself but when I read the review and then really look at the numbers and see that at 1600p the 6990 wins hands down and at 1080p with AA/AF on it also wins overall (a lot closer) and then toms says that the Nvidia card is a little better. Its just a bit confusing when you look at it that way. And im sure there is a reason but mind if I ask why there would be benchmarks at anything lower than 1080p? For a single 6970 ok, but why spend $700 on a duel GPU card to run at less than 1080p when a much less expensive card will do fine?

 

[ubercake]

Gotta try it again with today's driver update...

 

[cangelini]

[citation][nom]cburke82[/nom]We all saw the AA/AF numbers. I think the point is that toms basicly says the 590 wins by just a little. BUT if you look at the 1080p and up number with AA/AF on its not close to winning as it looses most benchmarks at those settings. So I can only speak for myself but when I read the review and then really look at the numbers and see that at 1600p the 6990 wins hands down and at 1080p with AA/AF on it also wins overall (a lot closer) and then toms says that the Nvidia card is a little better. Its just a bit confusing when you look at it that way. And im sure there is a reason but mind if I ask why there would be benchmarks at anything lower than 1080p? For a single 6970 ok, but why spend $700 on a duel GPU card to run at less than 1080p when a much less expensive card will do fine?[/citation]

I absolutely agree that a less expensive card does fine at 1680. However, that'd been a standard resolution in the benchmark suite. There is no harm in covering all of the usual testing, plus adding things like 5760x1080 to complement the single-display settings like 2560x1600.

It's all data--you use what's important. In the case of the GTX 590, I'm very clear that it is not hard at all to overwhelm its available memory. If the settings I demonstrate that in are what you'd be interested in using yourself *of course* the GTX 590 is going to be the totally wrong card for you.

The 590 wins, in my opinion, because I wouldn't tolerate gaming with the 6990 in the machine next to my foot. If I were using settings that made the GTX 590 slow down to 3 FPS, of course it'd be a big fat loser for me too, in which case I'd follow my OWN advice and grab a pair of 6970s.

The continued irony in all of this is that I don't really care about EITHER the 590 or 6990 unless we're talking about quad-GPU. I favor two 570/580s or two 6970s, depending on your needs. Someone who reads the story top-to-bottom should come away with the same conclusion.

Best,
Chris

 

[Crashman]

[citation][nom]cangelini[/nom]The continued irony in all of this is that I don't really care about EITHER the 590 or 6990 unless we're talking about quad-GPU. I favor two 570/580s or two 6970s, depending on your needs. Someone who reads the story top-to-bottom should come away with the same conclusion.Best,Chris[/citation]Wow, you know there's a good article on this site that shows what an outstanding value a pair of 6950s is as well?

And the fun part is that there are some STANDARD ATX boards out there already that support quad-SLI and quad-CrossFire using single-GPU cards! Cards that aren't crippled by underclocking! Cards that exhaust the heat out the back of the case!

I think someone should do a Quad GPU article using single GPU cards on a 4-slot standard-ATX motherboard. And, since the bottom card hangs 1-space below the case's bottom slot, maybe even an 8-slot case roundup! Wouldn't that be killer?

 

[silverblue]

Looking forward to it. If it's anything like the 3-way comparison though, we may not see a real big boost going from three to four cards, but it's always worth finding out.

 

[cburke82]

I absolutely agree that a less expensive card does fine at 1680. However, that'd been a standard resolution in the benchmark suite. There is no harm in covering all of the usual testing, plus adding things like 5760x1080 to complement the single-display settings like 2560x1600.

It's all data--you use what's important. In the case of the GTX 590, I'm very clear that it is not hard at all to overwhelm its available memory. If the settings I demonstrate that in are what you'd be interested in using yourself *of course* the GTX 590 is going to be the totally wrong card for you.

The 590 wins, in my opinion, because I wouldn't tolerate gaming with the 6990 in the machine next to my foot. If I were using settings that made the GTX 590 slow down to 3 FPS, of course it'd be a big fat loser for me too, in which case I'd follow my OWN advice and grab a pair of 6970s.

The continued irony in all of this is that I don't really care about EITHER the 590 or 6990 unless we're talking about quad-GPU. I favor two 570/580s or two 6970s, depending on your needs. Someone who reads the story top-to-bottom should come away with the same conclusion.

Best,
Chris

I read the whole thing and it is a very good article. I just re-read the conclusion page. you say.

"And it wins because I can put a GeForce GTX 590 in my workstation, which doubles as my gaming machine, and not hear it. It’s quiet enough to use. And that’s a requisite."

Now as a work PC I would see your point. But does that mean you would use the AMD card for the better performance if you were building a PC that would be used mostly for gaming? In other words if my speakers/headphones make noise a non issue would you then call the 6990 the winner? Also If you see my system in the sig I will be going the 2 card rout lol. Just found it a little odd that a GPU would win based on sound is all.

 

[aaron88_7]

Those of you saying the 6990 is better, or the 590 is better must have skipped the conclusion page where Tom's stated: "buy two Radeon HD 6970s or two GeForce GTX 570s and enjoy all of the performance without having to balance thermals and acoustics."

Doesn't sound like a win for either with statements like that!

I'm not so disappointed now knowing that EVGA won't let me step up from my 580 to the 590. Looks like I'm better off with a 2nd 580. Of course I need some more monitors to make such a purchase be worthwhile. Triple monitor display here I come

 

[cangelini]

I read the whole thing and it is a very good article. I just re-read the conclusion page. you say.

"And it wins because I can put a GeForce GTX 590 in my workstation, which doubles as my gaming machine, and not hear it. It’s quiet enough to use. And that’s a requisite."

Now as a work PC I would see your point. But does that mean you would use the AMD card for the better performance if you were building a PC that would be used mostly for gaming? In other words if my speakers/headphones make noise a non issue would you then call the 6990 the winner? Also If you see my system in the sig I will be going the 2 card rout lol. Just found it a little odd that a GPU would win based on sound is all.

Ah, but you miss the last graf:

"If you’re only concerned with the performance of an individual dual-GPU card, then I happily repeat the recommendation I made in the 6990 review: buy two Radeon HD 6970s or two GeForce GTX 570s and enjoy all of the performance without having to balance thermals and acoustics. Unless you’re 1) worried about plugging up four expansion slots, 2) you really need the 6990’s fifth display output, or 3) the GTX 590’s sweet light-up logo won you over, don’t worry about which company is selling the “fastest graphics card in the world.” Rather, take advantage of the fantastic CrossFire and SLI scaling we now see from AMD and Nvidia (respectively), and save yourself a few bucks in the process."

Now, if we're talking about quad-GPU to make this relevant to the 590/6990, then no, I still wouldn't use the 6990. The noise that card makes is not insignificant. Play back the video I embedded on the Noise page. It's not the kind of din you want to try to drown out with headphones or speakers. It's f'ing *loud.*

But I see you're going to go dual-card anyway. So again, who even cares which of these cards is slightly faster? The noise makes the 6990 unusable to me. If I have to declare a winner, it's the 590. But I wouldn't spend $700 on EITHER card when I could get by for $6xx-something and get better performance and more elegant cooling! You're doing it right.

Have a great weekend
Chris

 

[silverblue]

Haha ignore me, I misread.

*shuffles off*

 

[cburke82]

Those of you saying the 6990 is better, or the 590 is better must have skipped the conclusion page where Tom's stated: "buy two Radeon HD 6970s or two GeForce GTX 570s and enjoy all of the performance without having to balance thermals and acoustics."

Doesn't sound like a win for either with statements like that!

I'm not so disappointed now knowing that EVGA won't let me step up from my 580 to the 590. Looks like I'm better off with a 2nd 580. Of course I need some more monitors to make such a purchase be worthwhile. Triple monitor display here I come

Yes ill still end up using 2 flashed 6950's. But if your comparing 2 cards you may not recommend one or the other but one will still be better than the other. So in this case I would say with noise out of the equation the 6990 wins just by a hair. Now would I buy one? No lol but it would be the better ( by only a few FPS but still) card if one were to for some reason just really want a duel GPU card and wasn't using it in a situation were noise was a big concern.