Say i wanted a really huge (but non pro-cards) setup, what should i do?

ranakor

Member
Aug 8, 2007
77
0
66
What's the absolute best setup (not full computer, just the graphic part) money can buy out of consumer level cards? is it possible to SLI up multiple GTX690? If so how many times? Am i going to saturate a but and end up with diminishing returns?
I'm interested in making an insane gaming ring to test video game builds against, i'm in the planning stages of a video game and want something "future proof ish" so that if something runs at 30fps on it i'll know it's about where i want it at the max quality, "decent on mainstream hardware in 3-4 years".
Any tips?
 

chimaxi83

Diamond Member
May 18, 2003
5,457
63
101
Two 690 = quad SLI, and that's as high as you can go with those. Powercolor released a 7970x2, which is their own custom version of what a 7990 would be. Same thing there, two 7970x2 = Quadfire/CrossfireX/whatever they call it. That's the highest you can go right now, and there are definitely diminishing returns after the second card. If you're set on absolute maximum, might be better to just go with 4 custom cards with voltage control so that they can be cranked as high as possible.
 

ranakor

Member
Aug 8, 2007
77
0
66
Thanks for the info, so 4X sli / CF is the max and dual cards already "count" as 2 in sli, good to know, so basically it's 2X 690 or 2X7970 or something custom.
I'm not set on maximum as in "microperf maximum", if it's to get 2-3% more, i'm not interested in more work, if it's to get 30-50% more, i am. Which of those would the third solution go toward?
Also i've never had SLI/CF setups, any mobos that are known to work well with 2 top tier consumer grade cards like those and still allow for decent (min 24GB) of ram?
 

cmdrdredd

Lifer
Dec 12, 2001
27,052
357
126
A couple x79 boards have quad x16 PCIe slots. Like the Asrock Extreme 11.

Honestly after 2 cards it all goes downhill unless you are doing surround gaming/eyefinity with high resolution screens. Basically when you go beyond just 2 cards they don't scale well and the fps jump is small and in some games negative fps. SLI doesn't scale well in 3-4 card configurations, crossfire is a little better but I don't think the return is worth the cost.
 

ranakor

Member
Aug 8, 2007
77
0
66
Oki so 2X seems to be the best i can do reasonably.
A side question i should have asked earlier, if i don't "need" to purchase immediately, is there something coming out on the radar at the top end that is likely to be worth waiting for "soonish" or should i just go ahead now?
And if i go with the nvidia path of 2X690, any specific mobo for this setup known to run it reliably (i'm not going to overclock or anything)?
 

cmdrdredd

Lifer
Dec 12, 2001
27,052
357
126
hrm...soonish is very hard to define. If you mean this year I'd say no. Next spring maybe.

2x 690 is roughly equivalent to 4x 670. So you will have all the scaling issues I mentioned earlier. A single 690 is two GPUs setup in SLI on a single card. If your goal is two GPUs then two GTX 670s (4GB models preferably) is going to be a bit faster than a single 690 if you overclock, a bit slower (a couple fps here and there) if you don't overclock, but cost almost $200 less. On the AMD side for crossfire 2x 7950s would be a little faster than 2x 670s in benchmarks (not accounting for any microstutter discussions), and if you do overclock them you can get a lot of extra performance. Two 7970s would be even faster than that and again overclocking offers more potential. If you want 4 GPUs then a motherboard with 4 PCIe x16 slots and 4 7950s or 670s would be the way to go IMO.

BTW: You cannot run 4 GTX 690s, that would be 8 GPUs and is not possible. You can only have 4 GPUs at a time. Just remember GTX 690 = two GPUs.

What resolution are you running? That makes this discussion a little more worthwhile. Quad SLI might sound cool, but if you're on 1080p resolution you are wasting your time.
 
Last edited:

blackened23

Diamond Member
Jul 26, 2011
8,548
2
0
Oki so 2X seems to be the best i can do reasonably.
A side question i should have asked earlier, if i don't "need" to purchase immediately, is there something coming out on the radar at the top end that is likely to be worth waiting for "soonish" or should i just go ahead now?
And if i go with the nvidia path of 2X690, any specific mobo for this setup known to run it reliably (i'm not going to overclock or anything)?

All you need for 2 690s is any motherboard that supports SLI. Quad SLI is a different story, but the 690 has a PLX chip built in so you don't have to get a fancy motherboard for them - just get any SLI motherboard and you're set.

I agree that past 2 GPUs, you'll have a strong case of diminishing returns. (690 is 4 GPUs). Personally, I would prefer 2 680s or 2 7970s. I don't think I would go quad, the scaling is just sketchy past 2 GPUs in most cases.
 

ranakor

Member
Aug 8, 2007
77
0
66
What would be the limiting factors for 4X vs 2X (or 1X 690 vs 2X690)
I'm not going to run games on it as i said, i want to test for future proofness, so probably going to render at 2500 30" style
Not really interested in benchmarks so unless there is a strong perf insentive i'm going with nvidia as they seem to fare better in DX11 tesseleation, and i'm definately targeting that.
So i guess it's trimed down to
- I can target 2560 x 1600 resolution
- Price is not an issue (but i'm not paying 10X the price for 10% more perf)
- Preferably nvidia (unless my tesselation info is wrong/biased)
- Not for benchmarking/overclocking/Gaming, want to run as stock to test on something "future proof", as close as possible to a mid range gaming pc in a few years to project performance
With that in mind, should i go for 1X or 2X 690? Or something else? And while i understand that any (1 or 2 depending on choice of two or quad sli) slots pciX16 would do i'd like a recommandation on something known to "work well and stable", overclocking not mattering.
Specific recommandation on maker is welcome, stability > performance here again unless we're talking about major performance (i'm going to have this build run at max settings a lot of time and for long periods of time)
 

blackened23

Diamond Member
Jul 26, 2011
8,548
2
0
I would get quad xeons on a supermicro server motherboard with 4x GTX 690s, also mineral oil cooled.
 

RussianSensation

Elite Member
Sep 5, 2003
19,458
765
126
What would be the limiting factors for 4X vs 2X (or 1X 690 vs 2X690)

- Crossfire scaling
- 2GB of VRAM limitation in some games
- Memory bandwidth limitation
- GPU speed limitation

With that in mind, should i go for 1X or 2X 690? Or something else?

GTX670 4GB SLI, GTX680 MSI Lightning SLI, HD7970 GE Vapor-X CF.

The next step is 3 cards but it has huge diminishing returns.

GTX690 x2 is a total waste of $ as the SLI scaling for 4 cards is very poor on a single 2560x1600 monitor. Also, if you in the planning stages of making and game and want to see how this hardware performs in 3-4 years, I don't understand why you wouldn't just buy a GTX670 SLI now and then in 6 months sell them and get GTX780 SLI. Getting GTX690 SLI is not indicative of anything, other than how well SLI scaling works because it's not the same as 4x GTX680 mathematically as the 3rd and 4th card will have 50% and probably 25% scaling in games.
 
Last edited:

cmdrdredd

Lifer
Dec 12, 2001
27,052
357
126
AMD is just as good as Nvidia for tessellation in current cards.

I think what you want is what we can't give you. Nobody knows what cards next year or 2 years from now will do so trying to buy 4 GPUs of today's generation doesn't necessarily translate into tomorrow's. We don't know about compute strength on Nvidia's next parts for example. Direct Compute is used on some games and AMD is currently a ton faster in that regard. Nvidia might decide to put compute back into their desktop GPUs again but we don't know that.

Lets say you build an engine based on a GTX 690's strengths which we know does not include Direct Compute. Later you might find that DC will allow you to do something in a different way and free up resources. You wouldn't be able to test that with Nvidia cards at this time. However, we aren't sure if the next Nvidia cards will feature this like AMD's current 7970/7950 etc does. On the other hand you might be able to integrate Physx on the game or engine you're working on. That can only be done on Nvidia hardware. However, you do limit the target userbase slightly since you know before hand that nobody with an AMD GPU can use physx without hacked drivers. That's a small market, but might be something you wish to pursue anyway. Batman and Borderlands use Physx and it works rather well except for the slight performance hit.

There is no benefit with testing your game or game engine with 4 GPUs. That market segment is so miniscule that you would be wasting lots of time and resources for what would amount to 1% or less of the total market. I'd say regular SLI or crossfire is enough test and if the game or game engine takes off and hits it big then Nvidia and AMD will develop better drivers with profiles targeted to that title.
 
Last edited:

ranakor

Member
Aug 8, 2007
77
0
66
I'm not building an engine, but using one, what i want is to target next gen in assets, so having something that seems like "the best something can buy" today, is usually a good comparison of what average ish will be a few years later, the plan is to target assets to run "ok ish" in such hardware, which makes sense as games take quite a while to make.
Just to be clear, i'm interested in what could run the best today, not what today's consumer pc look like, so i "do" want to spend time and ressource on 1% or less of the total market, as it may represent a sizeable amount of market in 3+ years so that's definately where i'd like to be for top settings.
 

RussianSensation

Elite Member
Sep 5, 2003
19,458
765
126
I'm not building an engine, but using one, what i want is to target next gen in assets, so having something that seems like "the best something can buy" today, is usually a good comparison of what average ish will be a few years later,

See but I think it changed since the increase in GPU development has slowed down somewhat. For example, the best in Fall of 2009 was HD5870 CF. It's been 3 years now and HD5870 CF is very high-end for the majority of the PC gaming market outside of these enthusiast forums. The average gaming PC on Steam probably has a GTX560 or HD5770 in it, way slower than HD5870 CF.

Even if you look at the best system from Fall 2010, that's GTX580 SLI. Again 2 years since that point and you'd are looking at 40% faster than that with 670/680/7970 OC SLI/CF.

So your analogy of GTX690 and especially GTX690 SLI representing the mainstream market in 3 years sounds optimistic. The mainstream market performance in 3 years will probably be GTX680/7970 level of performance.

Just to be clear, i'm interested in what could run the best today, not what today's consumer pc look like, so i "do" want to spend time and ressource on 1% or less of the total market, as it may represent a sizeable amount of market in 3+ years so that's definately where i'd like to be for top settings.

Yes, that's another point I don't understand. Why don't you get GTX670 SLI now and then in 6 months reassess the market. I don't understand why you want to buy GTX690/690 SLI and keep it for 3 years for "market estimation purposes". The market is dynamic and as GPUs release, you can sell your old cards and get what you feel is an estimate of the market then. Your idea of estimating the market is wasting $2000 on GPUs that may or may not represent the mainstream GPU performance in 3-4 years? How does that make sense? I don't really understand what you are trying to do. Are you trying to future proof for yourself or seeing how a top of the line system today will fair in 3-4 years? If it's the former, you are wasting $ since future-proofing never really worked. If it's the latter, why not keep upgrading all these years and then see what the mainstream performance is in 3-4 years and then make graphical adjustments accordingly?

If you want to target next generation in assets, you could make those assets as advanced as possible (assuming timeline and budget constraints) since GPUs will only continue to grow and that would make the game look better for even longer (ala Crysis 1). If you want to target the next generation mainstream gamer in 3-4 years, then GTX690 SLI is still too optimistic I feel.

The other difficulty is we do not know how much faster graphics will improve after next generation of consoles launch. For example, if next generation GPUs are 5-10x faster in tessellation and next generation game engines use a lot of tessellation, then may be even GTX690 SLI would be slower than a $200 GPU in 2015. However, if next generation engines are hardly better looking than BF3 is today, then GTX690 SLI would represent something far exceeding the power of a mainstream GPU in 2015 because in that case most people would probably be happy with a GTX680 level of performance. You are basically trying to estimate how fast a $200-250 GPU will be in 3-4 years from now and trying to get a system today that most closely approximates that level of performance? This is very difficult to answer but if look at the last 3 years:

HD5870 ---> HD7970 GE on the AMD side in 2.5 years (that's about 70-75% faster)
GTX480 ---> GTX680 on the NV side in 2.5 years (that's about 50% faster).
^ So you can see it has been very slow progression in the last 2-2.5 years actually.

It's probably expected that GTX880 (Maxwell) will be as fast as GTX690 by 2H 2014 and then let's say GTX1080 is 70-80% faster than Maxwell by 2H 2016. So probably by 2016, a $200-250 GPU will be as fast as a GTX690/GTX670 SLI. But it could be faster depending on what technologies next generation games use that benefit from virtual memory, pre-emption, direct compute, geometry engines of newer GPUs (i.e., they may be exponentially faster with those next generation graphical features).

To show you what I mean HD5870 is exponentially slower in some newer DX11 games but overall it's not 2-3x slower than modern GPUs. That's why your question is so difficult to answer. If in 3-4 years all games use global illumination via DirectCompute lighting effects, then GTX690 SLI would be dog slow even compared to a $200 GPU that's because current GPU architectures are not really built for next generation games 3-4 years from now which would mean an exponential performance hit in the more advanced graphical features such as tessellation, depth of field, global illumination, multiple area lights, etc.

45143.png


And then you have the other uknown: Will 1080P still be the mainstream resolution, or will it be 2560x1440/1600 or 4K or some other resolution by 2015-2016?

I feel like GTX670 SLI 4GB sounds like a good balance. It gives you plenty of VRAM to test out your assets with without crippling the GPU (such as would be the case with the GTX690) and should we be wrong about GTX670 SLI 4GB as representative of a mainstream GPU in 3-4 years, you can quickly go out and add a 3rd such card. So it's very flexible in this regard. Going with GTX690 SLI costs more and you don't get the 4GB VRAM flexibility down the line to test your art assets. Also, you still get some benefit of Tri-SLI scaling which is not great but still much better than GTX690 SLI. Plus, you save $700-800 going this path vs. 690 SLI.

You can also use the GPU performance increases in the past to try and approximate where we'll be in 3-4 years (although this is pretty hard but you can ballpark):

Graphics Evolution from 2005 to 2011
Recent Graphics Evolution - 2012
Here is how GTX690 compares to some of those cards

And here is a GPU database for all the launch dates to give you an idea of how many years it took to get what level of GPU performance increase. In the past it took 2 years to get 70-100% GPU speed increase, but lately this has levelled off.
 
Last edited:

Face2Face

Diamond Member
Jun 6, 2001
4,100
215
106
RussianSensation, dude you own these forums. Your Posts are some of the most informative I have seen. Keep it up!
 

boxleitnerb

Platinum Member
Nov 1, 2011
2,605
6
81
A couple x79 boards have quad x16 PCIe slots. Like the Asrock Extreme 11.

Honestly after 2 cards it all goes downhill unless you are doing surround gaming/eyefinity with high resolution screens. Basically when you go beyond just 2 cards they don't scale well and the fps jump is small and in some games negative fps. SLI doesn't scale well in 3-4 card configurations, crossfire is a little better but I don't think the return is worth the cost.

No matter how often people repeat that, it is not true. If you chose the appropriate settings (SSAA, AO etc.) and have a really really fast CPU, 3-4 cards scale reasonably well. CF scales badly with 3+ cards on average (meaning not only optimized titles that are benchmarked by major sites).

@OP:
If you really want the most insane setup, get 4x GTX680 4GB, a 3960X and overclock the latter as high as it goes. That should last you for a while.
If that is a good choice, that is another question altogether.
 
Last edited:

cmdrdredd

Lifer
Dec 12, 2001
27,052
357
126
No matter how often people repeat that, it is not true. If you chose the appropriate settings (SSAA, AO etc.) and have a really really fast CPU, 3-4 cards scale reasonably well. CF scales badly with 3+ cards on average (meaning not only optimized titles that are benchmarked by major sites).

@OP:
If you really want the most insane setup, get 4x GTX680 4GB, a 3960X and overclock the latter as high as it goes. That should last you for a while.
If that is a good choice, that is another question altogether.

Sli scales worse than crossfire when both are available in a particular title usually, but sli for the most part works in more games with less issues.

You need to read this thread I linked below.


Annisman made a thread showcasing the performance difference between 2way and 3way sli using 4GB GTX 670s and the scaling was not impressive. Conclusion was it was a waste beyond two cards.
http://forums.anandtech.com/showthread.php?t=2260081&highlight=
 

boxleitnerb

Platinum Member
Nov 1, 2011
2,605
6
81
Well, I said "on average", including many games, not just popular titles.

Annisman didn't really put his cards to good use.
3DMark11 Performance preset?? That is 720p...
Many of the games he tested are CPU bottlenecked, at least partially. Sometimes, custom SLI bits can improve scaling as well.

When GK110 arrives, I will get 3 of them and then I can show you how well SLI can scale.
 

cmdrdredd

Lifer
Dec 12, 2001
27,052
357
126
Well, I said "on average", including many games, not just popular titles.

Annisman didn't really put his cards to good use.
3DMark11 Performance preset?? That is 720p...
Many of the games he tested are CPU bottlenecked, at least partially. Sometimes, custom SLI bits can improve scaling as well.

When GK110 arrives, I will get 3 of them and then I can show you how well SLI can scale.

Keep telling yourself that...The drivers won't magically change and the way SLI works will still be exactly 100% the same.

Metro is CPU bottleneck? Not even remotely accurate.
 

jacktesterson

Diamond Member
Sep 28, 2001
5,493
3
81
No matter how often people repeat that, it is not true. If you chose the appropriate settings (SSAA, AO etc.) and have a really really fast CPU, 3-4 cards scale reasonably well. CF scales badly with 3+ cards on average (meaning not only optimized titles that are benchmarked by major sites).

@OP:
If you really want the most insane setup, get 4x GTX680 4GB, a 3960X and overclock the latter as high as it goes. That should last you for a while.
If that is a good choice, that is another question altogether.

Can't agree here.

Crossfire has scaled better in both the 6000 and 7000 series in my opinion. Whether SLI works better, I can't answer that. I have used different Crossfire setups that I was more than happy with performance wise though. (basically referring to Microstuttering. I did have an issue with 7850 Crossfire but that was close to launch)

For 680/670 4 way crossfire reviews I've seen, the 3rd GPU adds about 25-35% and the fourth about 15-20%. Not anywhere near worth it.
 
Last edited:

boxleitnerb

Platinum Member
Nov 1, 2011
2,605
6
81
Keep telling yourself that...The drivers won't magically change and the way SLI works will still be exactly 100% the same.

Metro is CPU bottleneck? Not even remotely accurate.

Please read more carefully. I said "many" not "all".

Can't agree here.

Crossfire has scaled better in both the 6000 and 7000 series in my opinion. Whether SLI works better, I can't answer that. I have used different Crossfire setups that I was more than happy with performance wise though. (basically referring to Microstuttering. I did have an issue with 7850 Crossfire but that was close to launch)

For 680/670 4 way crossfire reviews I've seen, the 3rd GPU adds about 25-35% and the fourth about 15-20%. Not anywhere near worth it.

Depends on the number of the games you are looking at. Once you leave the mainstream a bit, it gets ugly quickly. Especially for 3+ GPUs. There is a nice review with many many games from AlienBabletech that shows exactly that.

4-way SLI please ;)
Problem is with those reviews that they quickly become CPU bottlenecked. As I've explained many times already, 5760x1080 is 3x more pixels, but with 4-way SLI you have 4x the GPU power. So the problem begins already here. Then with a doubling of pixels, you don't get half the performance, but a bit more. 400% scaling is unrealistic, that much is true. But 370% compared to one GPU is possible. But for that you would probably have to use downsampling (up to 3840x2160 - more than 5760x1080) and/or SGSSAA.

And remember:
It is possible for a scene to be CPU- and GPU bottlenecked during the course of the benchmark. If you are not 100% GPU bound during 100% of the time, the scaling, which is based on the average fps value, obviously suffers. If one were to test more demanding settings or only focus on the demanding part of the scene, the results would be quite different.

Look at 2-way SLI. Most reviews don't show the 90%+ scaling that I observe every day with my setup. Especially ratings can be very misleading when one or two titles are CPU bottlenecked our when the profile is not optimal. The latter often can be fixed with new compatibility bits (either through Nvidia update or custom bits found by experienced users), restoring scaling to 90+. Of course, these are things that most people don't know. For example GTA 4 scales horribly in SLI (if at all), but with custom bits, performance nearly doubles and both my GPUs are fully used. Last week a friend provided me with custom bits for Mirror's Edge, giving me full GPU usage and a 15-20% fps boost. Same with Sleeping Dogs where SLI was basically broken in conjunction with ambient occlusion. The new bits he provided there completely fixed the issue and will be used by Nvidia in one of the next driver updates. This list could go on for quite a while.
 
Last edited: