HardOCP 6990+6970 CF vs 580 Tri Sli

[Silverforce11]

1.5GB vram is not limited at the tested settings. It's nothing to do with vram.

What's amazing to consider from this is the fact you can do a 3x 6950 setup for ~$650 or less, unlock or not, but just OC to 6970 speeds (easily) and it will beat the 6990 + 6970 or TRI 580 setup.. for less than $700. That's just absurd perf/$.

The 6950 is essentially a top end single GPU, when have we ever had top end stuff so cheap? I wonder if AMD didn't make a mistake by making 6950 will normal 6970 dies.. it just eats into 6970 sales.

 

[Zap]

How about people just accept that AMD has better multi card scaling and catches up at higher resolutions with the 6000 series?

*sigh* I guess most of you guys here are too accustomed to fanboyism. Let me bold it so you won't misunderstand (though I fully expect half of you will refuse to comprehend or will consider this sacrilege).

#1 My name is Zap and I fully understand that AMD graphics cards can be super awesomeness and more value and/or performance than Nvidia graphics cards.

#2 My name is Zap and most of you don't know me very well. I am very likely the most hardware agnostic person on these forums. Nay, on the whole internet.

Sheesh! The most recent graphics card I took fresh out of its box (which I purchased with my own money and was brand new from Newegg) was a Radeon 4650 that I temporarily used in my new HTPC. Nvidia fanboy indeed. I was just pointing out some thoughts I had on the HardOCP article.

I'm typing this on a Core i7 975X, one of several that I used to have. I also have owned a number of Atoms, and still have six. I also have a Zacate. I also have an AM3 proc (recently sold off another) and used to be big with socket 939 and even AM2. I have not skipped any Intel or AMD CPU interface except for Slot A (I knew socket A was coming and waited for it). I have purchased and used memory from Corsair, Kingston, Kingmax, Viking, Crucial, Mushkin, Adata, G.Skill, Super Talent, Patriot and probably a bunch of others I can't think of right now. I have hard drives from WD, Seagate, Samsung and Hitachi. Heck, I think I even have a Toshiba 2.5" HDD somewhere. I have LCD monitors from Dell, Samsung, HP, Westinghouse, Acer, NEC, IZ3D and probably some others. I have used pretty much all the motherboard brands. Heck, I even happily use ECS, Biostar and Foxconn motherboards, along with the usual Asus, Gigabyte, MSI, etc. I have SSDs from Intel, Adata, OCZ, Kingston, Patriot covering several different controllers.

About the only brands I don't do are Apple (so far I haven't found any Apple branded parts I can use with my custom built PC other than maybe some peripherals) and brands that are exclusively in some non-consumer/enthusiast niche, like SuperMicro. Oh yeah, and I'm kind of picky about my mouse, but I have tried a lot and know what works for me because of shape/size/feel.

But, if you guys want to accuse me of being a fanboy defending some random company that you hate, so be it. I used to avoid VC&G because of all the bickering that goes on around here. I guess I'm not missing much.

As far as the low FPS goes, that's what [H] always shoots for. How much eye candy can they throw in and maintain minimum playable frame rates.

Sure, but is 35FPS really that playable? Some of the minimums got pretty low too.

So $2300 worth of non-reference GTX580 (and probably a second power supply?) beat $1360 worth of reference 6970's?

If the guy has a super overclocked 990X, I don't think money was an object for him.

 

[dac7nco]

*sigh* I guess most of you guys here are too accustomed to fanboyism.

I didn't see much of that in this thread at all. Bad Day?

Daimon

 

[MangoX]

I don't see much fanboy-ism at all in this thread, at least not from the Red Team. Most people are price conscience and look for the best bang for the buck. From this review AMD delivers that in spades. Winning in perf/$, perf/VRAM, perf/watt, perf/die-size and display connectivity.

 

[Zap]

I don't see much fanboy-ism at all in this thread

I was immediately accused of "defending Nvidia" when I had a couple of (what I considered) valid issues with the article.

 

[RussianSensation]

There is no argument here.

Keep in mind if you care about 3D vision surround gaming though, then NV's setup is far superior. If you are spending $1000-1500 on a videocard setup, trying 3D gaming is certainly a viable option. In that case, NV wins hands down.

What's amazing to consider from this is the fact you can do a 3x 6950 setup for ~$650 or less, unlock or not, but just OC to 6970 speeds (easily) and it will beat the 6990 + 6970 or TRI 580 setup.. for less than $700. That's just absurd perf/$.

Agreed. Although with shortage of reference designed HD6950s, the prices have come up to about $270. Reference 6950 2GB look like they are about to go extinct.

 

[Ferzerp]

I was immediately accused of "defending Nvidia" when I had a couple of (what I considered) valid issues with the article.

I don't really care that much on what crazy extreme setup performs the best, but your reaction is understandable when dealing with that site. It lost objectivity a very long time ago. After reading reviews from in on the 480's, one would think that no one in the world could use a 480 without going deaf from the noise.

I ended up getting one, and I never hear the fans. It's not like I'm not stressing the thing either. I drive a 2560x1600 monitor with it and it's paired with a 4Ghz i7.

The AMD/ATI analog card (the 5870) review they put up didn't have such an extremely over the top reaction to the noise. I have access to one of those as well, and you know what? It is, as the nvidia card is, quiet.

Run furmark on them, and the fans will finally spin up. Guess what? They're both equally loud in that situation.

I'd realized there was some bias before that with that site, but it became glaringly obvious at that point. So yes, your hesitancy to accept the findings at face value is understandable. I'm not saying they aren't right, but I am saying that sadly, we cannot really trust that they will provide an objective review. It goes far beyond game choice, resolution choice, whatever. It would be just as bad if it were blatant nvidia bias. Either situation creates for untrustworthy reviews, even if a particular is accurate, how am I supposed to know at this point?

 

[bryanW1995]

E-mail Brent with a polite question and you will get a polite response.

Yeah, I've never had a problem with kyle or brent.

Are you saying they should test $1500 worth of gear for a measly 25x16 resolution? What's the point? If you spend that much on so much and game on 25x16, no matter what any Amd solution will be better.

On point 2, try to read the review before you try to defend nVidia:

$1500 worth of nVidia is less worth than $1000 of Amd.

Maybe you should more carefully consider your words before attacking a long-standing member like Zap. I've never seen him push for either AMD or NV at all. If you had been here for more than 16 posts you'd already know that. Good thing he's not like some of those mods at H (allegedly) are or you'd be permabanned right now.

 

[formulav8]

I was immediately accused of "defending Nvidia" when I had a couple of (what I considered) valid issues with the article.

The only thing I remember wrong with your thread is that you were making ram limitation excuses for why nvidia might have performed so badly (I don't feel like reading your exact wording right now). Its like you didn't even read the article. They plainly said that they used scenarios that would NOT bottleneck nvidia's frame buffer.

 

[Revdarian]

I'd realized there was some bias before that with that site, but it became glaringly obvious at that point.

I guess that you missed all the vids of Kyle running his sli 480 gaming rig because at the time it was the faster thing? (he bitched about the heat it outputted, but still that is understandable).


So yup, waaaay biased. /rolleyes

 

[SickBeast]

I can see this article as being useful for people who want to game on three monitors, but for the rest of us it probably does not reflect reality.

AMD cards seem to do considerably better at higher resolutions. nVidia really should work on this, because people who drop money on a pair of GTX 580s probably want to game at 2560x1600 at the very least (otherwise it makes no sense IMO).

 

[BrentJ]

Thank you for all the replies, we aren't done yet. I have a new test system, very fast current gen CPU, and we are going to re-test all of this again. For those who want 2560x1600 I will do some ap2ap graphs for those that find value in that, however I did make sure the settings I used at 5760x1200 were not VRAM limited, still, we want to address that concern directly, and so I will do a lower single-display resolution as well in the follow-up. There is also a lot more being tested, so stay tuned.

 

[3DVagabond]

Thank you for all the replies, we aren't done yet. I have a new test system, very fast current gen CPU, and we are going to re-test all of this again. For those who want 2560x1600 I will do some ap2ap graphs for those that find value in that, however I did make sure the settings I used at 5760x1200 were not VRAM limited, still, we want to address that concern directly, and so I will do a lower single-display resolution as well in the follow-up. There is also a lot more being tested, so stay tuned.

You're never going to stop the complaining. Thanks for listening and trying to answer all of the requested comparisons, though.

 

[Lonyo]

Thank you for all the replies, we aren't done yet. I have a new test system, very fast current gen CPU, and we are going to re-test all of this again. For those who want 2560x1600 I will do some ap2ap graphs for those that find value in that, however I did make sure the settings I used at 5760x1200 were not VRAM limited, still, we want to address that concern directly, and so I will do a lower single-display resolution as well in the follow-up. There is also a lot more being tested, so stay tuned.

You have to go and get 3x3GB GTX580s (they sell them at Amazon.com, they don't seem to actually be available from anywhere else, but that's not a problem) if you really want to satisfy people.

If you don't compare $1800 of NV hardware that is not particularly widely available to $1000 of widely available AMD hardware, it's not a fair test and you might as well not bother ever doing any articles ever again because you're clearly super biased.

 

[betasub]

If you don't compare $1800 of NV hardware that is not particularly widely available to $1000 of widely available AMD hardware, it's not a fair test and you might as well not bother ever doing any articles ever again because you're clearly super biased.

You need a whole new emoticon for that level of sarcasm, Lonyo

 

[dac7nco]

You have to go and get 3x3GB GTX580s (they sell them at Amazon.com, they don't seem to actually be available from anywhere else, but that's not a problem)

NewEgg has the Palit 3GB cards again... same bullshit "reviews" from people who don't own the card. I wonder why NewEgg allows that.

http://www.newegg.com/Product/Produc...scrollFullInfo

Daimon

 

[Skurge]

NewEgg has the Palit 3GB cards again... same bullshit "reviews" from people who don't own the card. I wonder why NewEgg allows that.

http://www.newegg.com/Product/Produc...scrollFullInfo

Daimon

Yeah, it should only be people who bought the card, I wouldn't be surprised if its just the same guy with a different account.

 

[Adul]

You need a whole new emoticon for that level of sarcasm, Lonyo

Just use his avatar, the trout.

 

[SolMiester]

The only thing I remember wrong with your thread is that you were making ram limitation excuses for why nvidia might have performed so badly (I don't feel like reading your exact wording right now). Its like you didn't even read the article. They plainly said that they used scenarios that would NOT bottleneck nvidia's frame buffer.

Im not sure on this, there was a link here last month about a guy who updated his 3 x 1.5Gb 580 to the Palit 3GB and not only were they faster, but left the AMD setups behind...<shrug>, ill see if I can find it!

Edit - Nah, its was Quad SLI, not Tri!!

 

[wahdangun]

so if it was not vram limitation then what makes it slower than radeon counter part?

 

[3DVagabond]

so if it was not vram limitation then what makes it slower than radeon counter part?

Resolution. As it goes up Cayman gets faster relative to Fermi.

Now, there is/was an issue with bandwidth through the crossfire connection that seems to bottleneck crossfire as you add more cards. This might be avoided/limited by going 6990/6970, only 2 cards, instead of 3, or more, discrete cards. This might be why this particular combo competes so well against the nVidia solutions. We've seen comparisons of tri/quad crossfire against tri/quad SLI not be as dominant. If so, and I'm only speculating, then it's not an issue of architecture, simply hardware implementation.

There's an article @ Kitguru that talks about the next gen crossfire being "the fastest yet" by improving the interface between the cards. Maybe, the best is yet to come?

http://www.kitguru.net/components/graphic-cards/dragan/3rd-generation-crossfire-will-be-fastest-yet/

No matter. For now, the 6990/6970 trifire setup is the dominant setup in it's class. this latest article at [H], leaves no doubt about that.

 

[wahdangun]

Resolution. As it goes up Cayman gets faster relative to Fermi.

Now, there is/was an issue with bandwidth through the crossfire connection that seems to bottleneck crossfire as you add more cards. This might be avoided/limited by going 6990/6970, only 2 cards, instead of 3, or more, discrete cards. This might be why this particular combo competes so well against the nVidia solutions. We've seen comparisons of tri/quad crossfire against tri/quad SLI not be as dominant. If so, and I'm only speculating, then it's not an issue of architecture, simply hardware implementation.

There's an article @ Kitguru that talks about the next gen crossfire being "the fastest yet" by improving the interface between the cards. Maybe, the best is yet to come?

http://www.kitguru.net/components/graphic-cards/dragan/3rd-generation-crossfire-will-be-fastest-yet/

No matter. For now, the 6990/6970 trifire setup is the dominant setup in it's class. this latest article at [H], leaves no doubt about that.

no, I think it's have to do with nvdia driver that use more cpu power, and you can see it in anand review the sli have worse minimum frame rate.
But why its exclusive to 6xxxx series gpu? And isn't bart and cypress have same architecture ?

 

[betasub]

^ Cayman GPUs (69xx) have VLIW-4, whereas the others mentioned (5xxx/68xx) are the earlier VLIW-5.

 

[wahdangun]

^ Cayman GPUs (69xx) have VLIW-4, whereas the others mentioned (5xxx/68xx) are the earlier VLIW-5.

yeah I know cayman use newer VLIW4, but what I didn't understand is why bart have near perfect scaling while the cypress was like garbage

 

[formulav8]

No matter. For now, the 6990/6970 trifire setup is the dominant setup in it's class. this latest article at [H], leaves no doubt about that.

Can't argue with that.

Not only is nvidia the slower setup, it is WAY overpriced. You can not only get better performance with the AMD setup, you can also Save alot of money.