Multi-GPU Scaling - Impressive Results!

Technical Support
One of the great flaws in the test Kodiack ran here, and a mistake I've seen others do, is to run a dual-GPU card and then compare it to running the same card with only 1 GPU enabled ("Crossfire off"). This is not the same as running a separate single-GPU card and comparing it to a dual-GPU card fully enabled. Ideally, there shouldn't be a huge difference, but this is the real world, not the one we think it should be.

The Tomshardware article posts results from DX9 and DX11 runs averaged together. What we need is for Kodiack to make the same /timetest run as Chris made in his tests to compare more directly, but it looks to me like the results Chris got were compatible with the results Kodiack got for a HD 5970. here's the question, though - if it was working properly, wouldn't it scale to be some large percentage over the single-GPU HD 5850 results, rather than being less than those results, still? (The 5970 is essentially 2 5850s on the same board.)

Kodiack's HD 5970 isn't giving the performance of a single-card HD 5850, by his own test results.

;)
I'm...confused. I didn't run the same test as Tom's Hardware. I used a flight path in Northrend that's particularly taxing on the GPU. It should be scaling quite well over a 5850, but that 5850 would also have to run the same processor, drivers, Catalyst Control Center settings, resolution, video quality, and everything else. Basically, the tests have to be identical to see. By disabling CrossFire, I did essentially run a 5850.

If you don't find my results conclusive enough, Elecman got similar scaling on an Nvidia-based system:

http://us.battle.net/wow/en/forum/topic/1305770941?page=1#4

That's using SLI. Just like my results, SLI on vs. SLI off gives a significant performance boost. He has two individual 9800 GT graphics cards. My 5970 is largely two graphics cards as well; the chips and memory is just placed on one PCB instead of two.

01/08/2011 1:26 PMPosted by Zyr
if it was working properly, wouldn't it scale to be some large percentage over the single-GPU HD 5850 results, rather than being less than those results, still?


Yes, yes it would. If I ran the exact same processor, motherboard, memory, resolution and graphics settings as Chris, I'd likely see the same sort of scaling. The only difference would be the updated drivers and the fact that I run the latest CrossFireX application profiles. If I go run the same benchmark they did at 1920x1080, full ultra quality, I'd likely squeeze more FPS than even their 5870. In fact, I think I'll go run that exact same test now, just to see the results I do get.
________________________________________________
One does not demand respect. One earns it.
CORE I7 3.4GHz | 12GB RAM | RADEON 5970 | F120 SSD
'Your processor is under that hockey puck thing.' -Dat
http://www.wowpedia.org/Portal:Technical_support
For live support, click the #wowtech link
My results using the same test that Tom's Hardware performed:

Minimum FPS: 11.429
Average FPS: 93.694
Maximum FPS: 241.972


Better than even their 5870 performed, and leagues beyond their 5970.

01/08/2011 1:43 PMPosted by Zkrieg
Nobody cares how well a system performs flying around the game world.

Believe it or not a majority of people raid and run heroics.

But hey if think you got a kick ass system because your locked at 60 FPS flying from zone to zone than i guess thats your business.

A majority of the population actually play the game and don't care about 60fps flying around. What they do care about is whether their systems can play end game content or not.

You keep on rocking your kick ass system that can play fly zone-to-zone better than anyone else's.


This isn't about raw performance, but showing that the game scales with multiple GPUs. If you're running a single graphics card and are GPU-bottlenecked, you can get better performance by using multi-GPU technology. I don't think you understand that dungeons and raids are normally CPU-bottlenecked. If a player is indeed still GPU-bottlenecked in a dungeon/raid, they can still see a performance improvement by throwing in another graphics card.

Could you please stop trying to undermine what I'm saying? You're not arguing about the game not scaling with multiple GPUs, you're arguing about not getting "enough" performance in end-game content. If you want to complain about that, you'd be better off arguing in a thread about WoW's scaling with multiple processors.
________________________________________________
One does not demand respect. One earns it.
CORE I7 3.4GHz | 12GB RAM | RADEON 5970 | F120 SSD
'Your processor is under that hockey puck thing.' -Dat
http://www.wowpedia.org/Portal:Technical_support
For live support, click the #wowtech link
01/08/2011 1:43 PMPosted by Zkrieg
Nobody cares how well a system performs flying around the game world.

Believe it or not a majority of people raid and run heroics.

But hey if think you got a kick ass system because your locked at 60 FPS flying from zone to zone than i guess thats your business.

A majority of the population actually play the game and don't care about 60fps flying around. What they do care about is whether their systems can play end game content or not.

You keep on rocking your kick ass system that can play fly zone-to-zone better than anyone else's.


I think you're missing the point... This isn't to compare performance at all. Quit being a moron and understand the purpose of this thread. It was intended to show Multi-GPU Scaling, which if you read the dang title is exactly what its named too. Please if you can't comprehend this or don't have multi-gpu's to even ask questions or properly comment don't even bother posting in here.

Once again, THIS IS NOT ABOUT HOW MUCH A 5970 PERFORMS, THIS IS ABOUT HOW WELL MULTI-GPU'S SCALE IN WORLD OF WARCRAFT. You can't test GPU's in a CPU Benchmark, so why would you test a Raid?

Do you even know what Scaling is?
No you guys are missing the point.

How often do you see posts about sli not working while flying around the world? ZERO.

How many posts do you see about sli running like crap during raids, moving around major cities during peak hours or in other highly populated areas like BG's and TB. to many to count.

So your little theory of "SLI and Crossfire work just fine because my machine works great flying around the world" is pointless.

I'll say it again. nobody cares that can rock 60 frames per second idling for a few minutes flying from zone to zone. Everybody but you cares how well the game runs while doing the things mentioned above.

AGAIN: NOBODY CARES THAT YOU CAN ROCK SLI AT 60FPS IDLING AND STARING AT YOUR SCREEN.
01/08/2011 2:17 PMPosted by Zkrieg
No you guys are missing the point.

How often do you see posts about sli not working while flying around the world? ZERO.

How many posts do you see about sli running like crap during raids, moving around major cities during peak hours or in other highly populated areas like BG's and TB. to many to count.

So your little theory of "SLI and Crossfire work just fine because my machine works great flying around the world" is pointless.

I'll say it again. nobody cares that can rock 60 frames per second idling for a few minutes flying from zone to zone. Everybody but you cares how well the game runs while doing the things mentioned above.

AGAIN: NOBODY CARES THAT YOU CAN ROCK SLI AT 60FPS IDLING AND STARING AT YOUR SCREEN.


...This isn't a troubleshooting thread either. There's more Eyefinity or Nvidia Surround posts than SLi or Crossfire posts. You're still not understanding either, this isn't a basis of what FPS you get where. I'm quite sure Kodiack understands exactly what HIS thread was intended for. This was simply intended to show GPU Scaling in GPU Dependent locations and NOT GPU Scaling in CPU Dependent Locations.

So how about this, go somewhere else and enjoy your prebuilt PC while the big boys who understand what Scaling is stay here.

Reported for Trolling. /thread

AGAIN: NOT A BENCHMARK SHOWCASE THREAD, NOR IS THIS A E-PEEN CONTEST BECAUSE IF IT WAS I'D HAVE POSTED MY RESULTS.

If you REALLY want Kodiack to go benchmark his 5970 in a inconstant environment with ever changing variables then why don't you go buy a 5970 or dual GTX 480/580s and try this yourself.
Zkrieg, I'm afraid you're the one "missing the point". Some players are held back by the graphics cards, particularly if they're lower-end. If you're graphics card-bound, an upgrade will net you more playability. This topic's to show that people that are being held back by their graphics cards can upgrade with another of the same manufacture to increase FPS.

Sure, most players are largely graphics-bound in the open world, but some players, particularly on lower-end cards like the GeForce 8500 or Radeon 5450, can get better FPS in dungeons and raids with more graphics power too.

Populated areas are CPU-bound. You can't measure SLI/CrossFire capabilities in something that's only going to be held up by a processor. It's just illogical and doesn't help come to any conclusions.
________________________________________________
One does not demand respect. One earns it.
CORE I7 3.4GHz | 12GB RAM | RADEON 5970 | F120 SSD
'Your processor is under that hockey puck thing.' -Dat
http://www.wowpedia.org/Portal:Technical_support
For live support, click the #wowtech link
Awesome. you have a system that kicks some serious butt while idling.

Now go benchmark you SLI while staring at your desktop.
01/08/2011 2:27 PMPosted by Zkrieg
Awesome. you have a system that kicks some serious butt while idling.

Now go benchmark you SLI while staring at your desktop.



He has Crossfire, you failed.
01/08/2011 2:27 PMPosted by Zkrieg
Awesome. you have a system that kicks some serious butt while idling.

Now go benchmark you SLI while staring at your desktop.


It's not idling when the GPU load's 100%. >.>

Both open-ended zones and populated zones push systems. The graphics card is pushed when the CPU load is light, and the CPU's pushed when there's a lot going on. They're two ENTIRELY different things. Neither one of them is "idle" since they both push hardware quite a bit.
________________________________________________
One does not demand respect. One earns it.
CORE I7 3.4GHz | 12GB RAM | RADEON 5970 | F120 SSD
'Your processor is under that hockey puck thing.' -Dat
http://www.wowpedia.org/Portal:Technical_support
For live support, click the #wowtech link
I think you both made your point.

Kodiak you have an awesome system that can rock max FPS while flying from point A to point B.

I'm not sure how this is relevant considering there are no support request's for poor performance while flying around the world staring at ones computer monitor.

I am afraid i have to go with Zkrieg on this one. This thread is pointless.
This thread was never about FPS to begin with, its original intent was to 1, show that Crossfire does indeed have a place in this game. Then 2, that SCALING(which for you dumb morons who can't comprehend this is the load distribution between processors.) is quite good for Crossfire. The 3rd intent is to hopefully get other Multi-GPU users responses on how theirs performs.


Everyone keeps being stupid and reading this as a benchmark comparison, which it isn't. If you read through everything kodiack says in the first thread he specifically states the percentage increase(or scaling) of how much performance he gains from enabling his second GPU.

If it were about FPS and comparing different hardware based on performance, it wouldn't have SCALING in the title and SCALING wouldn't be the main topic of what Kodiack discusses.
No you guys are missing the point.

How often do you see posts about sli not working while flying around the world? ZERO.

How many posts do you see about sli running like crap during raids, moving around major cities during peak hours or in other highly populated areas like BG's and TB. to many to count.

So your little theory of "SLI and Crossfire work just fine because my machine works great flying around the world" is pointless.

I'll say it again. nobody cares that can rock 60 frames per second idling for a few minutes flying from zone to zone. Everybody but you cares how well the game runs while doing the things mentioned above.

AGAIN: NOBODY CARES THAT YOU CAN ROCK SLI AT 60FPS IDLING AND STARING AT YOUR SCREEN.


...This isn't a troubleshooting thread either. There's more Eyefinity or Nvidia Surround posts than SLi or Crossfire posts. You're still not understanding either, this isn't a basis of what FPS you get where. I'm quite sure Kodiack understands exactly what HIS thread was intended for. This was simply intended to show GPU Scaling in GPU Dependent locations and NOT GPU Scaling in CPU Dependent Locations.

So how about this, go somewhere else and enjoy your prebuilt PC while the big boys who understand what Scaling is stay here.

Reported for Trolling. /thread

AGAIN: NOT A BENCHMARK SHOWCASE THREAD, NOR IS THIS A E-PEEN CONTEST BECAUSE IF IT WAS I'D HAVE POSTED MY RESULTS.

If you REALLY want Kodiack to go benchmark his 5970 in a inconstant environment with ever changing variables then why don't you go buy a 5970 or dual GTX 480/580s and try this yourself.


(reported for trolling a troll)

i'll post my result's if you want. But i dont think Koidak cares much for a quad 580 system backed by dual CPU's pushing triple 55 inch TV's.

I am locked @ 240 FPS across a triple monitor system. How's that show for e-peen?

Some people have both money and time to waste.
01/08/2011 2:34 PMPosted by Eleeginz
I'm not sure how this is relevant considering there are no support request's for poor performance while flying around the world staring at ones computer monitor.


This thread isn't made for those requests. This thread is for people to have some information available on scaling in the case that they start looking at multi-GPU systems. Players can cheaply get more juice from their system by plopping in another graphics card, contrary to what some people say. CrossFire/SLI scales nicely in WoW, and this thread is to show just that, with hard numbers to back it up.

If I had another dual GPU system to test with, I'd do that. As Ast said, this isn't a benchmark comparison thread or a topic made for bragging. It's to show how well the game scales with extra graphics cards when graphics hardware is pushed. If a player's only getting 30 FPS some place and they want a bit more with the same visual quality, they could toss in another graphics card and get roughly 45-55 FPS. A 60% increase in performance is absolutely nothing to scoff at.

________________________________________________
One does not demand respect. One earns it.
CORE I7 3.4GHz | 12GB RAM | RADEON 5970 | F120 SSD
'Your processor is under that hockey puck thing.' -Dat
http://www.wowpedia.org/Portal:Technical_support
For live support, click the #wowtech link


(reported for trolling a troll)

i'll post my result's if you want. But i dont think Koidak cares much for a quad 580 system pushing triple 55 inch TV's.

I am locked @ 240 FPS across a triple monitor system. How's that show for e-peen?

Some people have both money and time to waste.



lmao I'll believe you have 240 fps in Stormwind or Orgrimmer if you post a GPU-Z Validation with your username in it or a screenshot with MSI Afterburner's OSD up but until then you're full of it ;)
I guess any idiot can buy hardware nowadays and google how to install it =P
i'll post my result's if you want. But i dont think Koidak cares much for a quad 580 system pushing triple 55 inch TV's.


Actually, I do. Post your results so I can see the scaling with one GPU, two GPUs, and so on. I would like to have some extra scaling results available. Going from a single GPU setup to a dual GPU setup shows roughly a 60% increase in performance. I'd love to see numbers beyond that as well.

01/08/2011 2:42 PMPosted by Zkrieg
I am locked @ 240 FPS across a triple monitor system. How's that show for e-peen?


I really don't care if my FPS is above 60. I have vertical sync enabled for a reason. "Epeen" means nothing to me. I have this computer built this way because I love technology. Working with and building computers is my hobby; I'm a bit of an enthusiast. I like the pretty colors with mind-boggling amounts of AA, but I very rarely benchmark stuff and post results, and when I do, it's largely just to share some results.
________________________________________________
One does not demand respect. One earns it.
CORE I7 3.4GHz | 12GB RAM | RADEON 5970 | F120 SSD
'Your processor is under that hockey puck thing.' -Dat
http://www.wowpedia.org/Portal:Technical_support
For live support, click the #wowtech link
01/08/2011 2:47 PMPosted by Ast


(reported for trolling a troll)

i'll post my result's if you want. But i dont think Koidak cares much for a quad 580 system pushing triple 55 inch TV's.

I am locked @ 240 FPS across a triple monitor system. How's that show for e-peen?

Some people have both money and time to waste.



lmao I'll believe you have 240 fps in Stormwind or Orgrimmer if you post a GPU-Z Validation with your username in it or a screenshot with MSI Afterburner's OSD up but until then you're full of it ;)
I guess any idiot can buy hardware nowadays and google how to install it =P


I have been building systems long before you were born.

Clearly your narrow mind fails to grasp the concept that some people have been around a lot longer than you and actually finished high school only to move on to college and graduate school. Some people have multiple PhD's is this field of work.

If you had stayed in school you might be able to understand this. That or you're too young to understand what it means to hold a college degree and earn more than minimum wage.
The point of this thread seems to be testing GPU-limited scenarios so I'm not sure why raids are even brought up here. The others have stated already that raid scenarios shift the game from being anything remotely from GPU-limited to CPU-limited so SLI/Crossfire setups won't see benefits there. For everything else though (questing, dailies, dungeons, things that make you fill-rate bound on one card) you're able to get a boost.

This is a term from other forums but please don't thread-crap. You're not really adding anything to researched results and bloated the post into a two-page post.
________________________________________________
Account and Technical Services
Can't find a resolution on the forums? Contact a support representative directly: http://us.blizzard.com/en-us/company/about/contact.html

Great minds discuss ideas; Average minds discuss events; Small minds discuss people.
Vial of Tears http://www.surveymk.com/s/RHXXKHK
The point of this thread seems to be testing GPU-limited scenarios so I'm not sure why raids are even brought up here. The others have stated already that raid scenarios shift the game from being anything remotely from GPU-limited to CPU-limited so SLI/Crossfire setups won't see benefits there. For everything else though (questing, dailies, dungeons, things that make you fill-rate bound on one card) you're able to get a boost.

This is a term from other forums but please don't thread-crap. You're not really adding anything to researched results and bloated the post into a two-page post.
________________________________________________
Account and Technical Services
Can't find a resolution on the forums? Contact a support representative directly: http://us.blizzard.com/en-us/company/about/contact.html

Great minds discuss ideas; Average minds discuss events; Small minds discuss people.
Vial of Tears http://www.surveymk.com/s/RHXXKHK

\
Its not thread crapping .

You cant properly test a game when the scenario invloves systems that cant even push the game properly.

On a side note. Kodiak why do you need my results?

I've already stated that i'm locked @240 FPS. Yes my TV's are of the 240Hz variety. I have 3 TV's each one is connected to a different GPU and the fourth is a dedicated physX card for when i need it or do some real benchmarking with LN2.

I have yet to find a game that can make my FPS dip below 240.

There is no scaling when i'm locked at 240fps. I wont apologize for you being unable to understand this concept. It is clearly out of your league.

Good luck with your low end system that cant even play a 6 year old game with out flopping over and wanting to commit suicide.

Join the Conversation

Return to Forum