580m vs 6990m Benches

What has highend 4 card scaling got to do with 2 card mid range scaling (which is what the mobile cards are).

Seriously, if you are going to make comparisons, try at least picking ones that however biased actually have a point.

Every well informed person agrees that this generation AMD have an edge when it comes to Xfire scaling at 2-3 GPUs.

 

Post some benchmarks from your favorite games and I'll show you how my 580 scales

 

lmao, thats exactly what it is. relax guys. if it can play every game at max, i think were fine.

 

Quoted for truth, just settle everyone, well i was going to say this isn't a pisssing contest, but being a benchmark thread it is , but at this point its turning almost as bad as that other fecking debate that rages on forums (console vs PC), both are getting so old ...

 

damn, where is batboy when you need him...

 

Are people even serious regarding these debates?? I find it to be fun and that's it....

 

funny thing is..they aren't debates...

 

alright lets end this once and for all. just take a look at the hard numbers of these cards.

compared to the 485m, the gtx 580m has a 7.8% higher clock speed at 620 vs. 575. the memory speed and shader cores are the same.

compared to the 6790m, the 6990m has 16.7% more shader cores and these cores are also clocked 5% higher. doing simple linear math you can easily say the 6990m is 20% faster than 6970m.

the 485m is how much faster than 6970m, 5%-10%? on a pure technical level the 6990m is the stongest mobile card and AMD has a right to claim this.

on a realistic level the 6990m and 580m should be considered equal cards with the exception of driver support in particular games. this exception can be significant but at the end of the day these cards will be trading punches.

if pricing were equal i would choose nvidia for their better driver support. when you factor in current pricing the clear winner is 6990m. it would be almost foolish to go with nvidia's 580m right now.

 

+1 to this, if you don't need mobile cude based computing and can't convert it to use DirectCompute or OpenCL.

 

Here're my benches, everything stock:





Dirt 3

1920x1080
4xMSAA
Vsync OFF
Detail ULTRA

Stock GPU's: 620/1500



OC'ed GPU's: 750/1750



The Witcher 2:

Ultra Spec:
Vsync OFF
Ubersampling OFF

Frames: Min-19, Max- 62, Avg-46.983

Crysis 2

Ultra specs: Patch 1.9/DX11/Hi-Res Textures

Frames: Min-33, Max- 41, Avg- 38.533

Deus Ex : Human Revolution

never seen it below 60fps with Vsync ON



Stock GPU's: 620/1500



OC'ed GPU's: 750/1750



Metro 2033

Resolution: 1920 x 1080
DirectX: DirectX 11
Quality: Very High
Antialiasing: MSAA 4X
Texture filtering: AF 16X
Advanced PhysX: Disabled
Tesselation: Enabled
DOF: Enabled

Stock GPU's: 620/1500



OC'ed GPU's: 750/1750

 

updated my post both with stock and oc'ed GPU's in heaven benchmark 2.5) i experience no throttling in this tool...

 

which bios are you using?

 

original for the GPU's and unlocked A00 for the system (but it doesn't help)!

the good: i was able to run heaven benchmark 2.5 tool at 750/1800

the bad: in 3dmark11/Vantage it throttles at 680/1500

the situation is different in games, most of them run at 720/1600 for quite a long time with no issue.

i have to grab somthing like this watt meter to check a power draw. But i still consider this as a drivers protection which will be improved with the next official M18x 580M SLI nvidia release set

 

+ rep thx bro

 

6990m xfire overclocked to 825 core / 1000 mem and 2720qm cpu. dell 11.7 drivers.

edit; i want to make a point that my gpu overclocks are 100% stable and i can take them higher, to 850 core but they start artifacting only in crysis 2 so i found the sweetspot at 825 core. they do pass 3dmark11 and heaven benchmarks withotu artifacts or issues at 850 core but to me it is pointless to posting non-stable overclock in general.

[/URL][/IMG]

edit again, i noticed a gpu bug where the 2nd gpu was not being utilized fully so i simply rebooted the machine and got a better score. the reboot pegged the 2nd gpu like it normally should have in msi afterburner. guess it's worth showing the second pic;

[/URL][/IMG]

 

Hey argh ... same clocks ... 2630 OC'd cpu

 

Shaden, could u make a run at 870/1210 + OC'ed CPU with no messing in CCC, please?

 

lol I will give it shot tonight Red.

And by messing you mean leave everying with run application settings ?

 

hows that ???

 

really, same clocks? Cpu has no bearing on score. Care to explain how you got a score 16% higher with the same 825/1000 clocks?

 

prob disabled tesslation or you are using Dell drivers.

 

well i am also using dell 11.7 drivers. but messing around with the visual fidelity is meaningless.

 

I am on 11.8 drivers. tess is on btw

 

well then 11.8 drivers must have really upped the performance at least in this benchmark. hopefully you did not tamper with the catalyst settings.

 

To be the "same" you HAVE to have "AMD Optimized" ticked and "enable surface whatever" ticked. If you untick those, benches SHOOT through the roof. I can make benches look really good with 3dmark11 at almost 8000k (GPU score) with OC 6970CF and nothing ticked. That's about 1K difference than very high 6990 OC benches with the same settings. All of this translates to a "couple" FPS in game (minus Crysis 2) as AMD and Nvidia are really pushing drivers to claim superiority. The 6990 Murders the 6970 in that game despite being very close in most other game benches.

Again, the 4870CF and 5870CF still run everything on Max (mostly) and the 6970 was a true successor to the 5870. It's "orders of magnitude" higher due to different architecture (much more efficient) higher shaders, etc. It's absolutely retarded to argue about laptop GPU's because new ones are right around the corner that will "outbench" the current top class. BUT....while gaming it will hardly have a difference until current GPU's cannot play games at high settings.

We have seen people post in the "gaming" section very detailed pics of a SINGLE 460M in Crysis at "High" detail and the difference between that and Ultra (as he showed in very clear pics) was almost nothing and that is a still shot. When moving, you can't tell the difference. So if you can play games with eye candy that is very good; you have nothing to worry about.

These m18x systems are upgradable to an extent and the PS 330watt will probably allow 7 series upgrades in 2012 for anyone getting their hands dirty.

However most gamers buy a system, play for 2 to 3 years and then just buy a new one. It's really that simple. Benchmarking and going crazy with OC is really just fun in the beginning of a new system to show off how big of an E-peen you have and then it's best to just play games because there is very little you can do unless you are a johnkiss or extreme OC'er. Watch how the forums shift when 7 series come out. VERY FEW people with the m18x with 6970/6990 will be on the forums posting benches. They will either leave the forums or upgrade and keep going. That's how it has ALWAYS been and I have been on the forum for 3.5 years

I have no desire to put my system on ice or DICE or anything to that extent. Benchmarking really is just a tool to see where your system lies and allows you to understand your system. Learn, if you have AMD, your CCC, find your "quality" settings you want to play each game at, and keep your system up to date. You will get 2 to 3 years of High image gaming out of it. This goes for CF 4870/5870/6790/6990. If BF 3 kills the system (doubt it) then upgrade to the 7 series (or whatever minimum card plays at settings you like).

People at Hforum would probably be laughing at this because it's really arguing over about 1000K difference in 3dmark11 between stock systems. that's peanuts in FPS in games.

1900 by 1080 is hardly "high def" by PC standards these days. So any CF system that equals a relatively powerful single card desktop GPU mid-top tier....will game EASY at 1900 by 1080. They wouldn't even recommend higher because it's just overkill. Once you up resolution to really high levels you NEED two 590s. But for us notebook gamers....a CF system or SLI system will LAST. Even 460M SLI is "good." That's my take anyway.

 

nope all set to use application settings... I have a video actually will post, shows the FPS screen and then I exit and you can see the CCC settings, will post. But thats it. Might be some drivers, might be CPU, CPU must have -some- effect, it even effects the gpu score in vantage and 3d mark ...

 

cpu clock wont have any effect on heaven benchmark. it is strictly a gpu-based benchmark and that is what makes it great.

 

indeed, I hate how dependant vantage is on CPU ... must be the driver.

 

No, it's the application due to the heavy implementation of CPU tests. For GPU definitely Heaven for sure.

When you say you run "application settings" are your "amd optimized" ticked and "enable surface format optimizations" ticked? Because if not, you basically will NOT have DX11 enabled. That is because you will see your slider bar to "off." I THINK. I am not sure. Actually, that would be good to clarify as I THOUGHT I needed "AMD optimized" ticked in Crysis 2 to get Ultra to work. But come to think of it, if I use application settings and "untick" AMD optimized I may start to get that playable framerate on Ultra I wasn't getting as it will "Default" to application settings and disregard that the slider is OFF. Could this be the case Shaden?

Edit: I am going to go test it because AMD optimized really does kill the framerates.....

 

looks like i was running with a driver bug on the last heaven benchmark i posted. i noticed that the 2nd gpu was not fully pegged in gpu utilization. i rebooted the machine and gpu utilization was back to noramal.

[/URL][/IMG]

 

maybe shaden is right. with my simple reboot the driver bug cleared up and gpu utilization on the 2nd gpu went back up. his score now is only 8% faster than mine so maybe the 11.8 drivers give higher performance in heaven benchmark.

would still be nice if shaden would take the time to show proof

 

Well I did confirm that DX11 is indeed utilized without AMD "optimized" enabled. It's much MUCH smoother with just application settings in Crysis 2. The ONLY level that seems to drop FPS to dismal rates is the tessellated water level that cannot be seen. Overall it is smooth, but if you carefully go around the level where you "drop down in Amrour Mode" after you kill the enemies before dropping to the lower levels....it's really bad frame rates. That simply was just poor poor use of tessellation that isn't seen. However the game runs in the 30s-80s now depending on what's going on. Complete error on my part assuming you needed that 'ticked' to have DX11 effects. Application settings works fine.

 

Seems like all Crysis have certain areas that are just really poorly optimized. The ONLY game that I think was done right was Warhead. Crysis 1 is fine for the most part minus some of that last level in the warship and fighting the final boss. Oh well...price we pay to be PC gamers playing very very VERY advanced games much more difficult than consoles. However, I guess this is way off topic considering it's about 580 Vs 6990 so I will bow out now. But ya....at least it helped me realize an error I have been making when playing DX11 games. Best to have "application settings" ONLY instead of ADDING AMD Optimized. Peace out.

 

Yea I have application settings NOT AMD optimized clicked.

 

ok i just installed the new 11.8 underground drivers along with CAP3 for these and they did not make a difference. normal variation, i would say. maybe shaden you got lucky with your run or something.

but clearly the 6990m is the stronger gpu. i am waiting for voltage unlock in the future so i can bios-flash these cards to 900 core with stability.

[/URL][/IMG]

 

i agree...this **** is hella damn expensive!!!after doing the 9800m gtx's i chilled on laptop stuff for a bit..then grabbed a g73jh...chilled, then grabbed this one. against desktop benching...500 bucks a card 1000 cpus 300 dollar mother boards and other various over clocking stuff. EXPENSIVE! and upgrading cards and cpus everytime they make a change...it's crazy. best to just be happy...

argh...is that a stock run?

your going to be waiting along time for that bios flash...1.1 is the highest voltage in the vbios. so no going over that.

 

I just ran it with a relatively mild OC on my CF 6970s with a relatively safe and mild OC of 810/1025 and got 62.5. I see you are running FRAPS. Disable that for an extra frame or two I would imagine. I see it in the top left of your screen. Our scores are too close I would imagine since you are running 6990 Vs my 6970???? I used the exact same settings?

 

825 core / 1000 mem on the gpu. i wonder why it's taking so long for someone to come out with a voltage unlock. these 6990m's are going to be very popular because theya re half the price and as strong as 580m.

 

there wont be a software/bios voltage unlock.

 

on your sig i notice you have the bus-clock overclocked. i wonder if that forces everything in the system to run faster, similar to the old days of overclocking the front side bus. if so might explain the performance numbers.

i say let's settle this once and for all, lol. let's run strict gaming benchmarks at 1080p and with vsync disabled! fraps enabled don't make a difference since that runs on cpu only and we got 4 cores to play with. take a snapshot of the results and post them up.

1) crysis; there is a dedicated gpu benchmark file in the bin64 folder of that game. make sure you put all in-game settings to very high with vsync disabled. exit the game and go back into the game to verify all the settings remained at very high. i think there is a bug where one of the important settings may not stick to very high for some reason and you have to set it back. patch game to patch 1.2 as well. do not use the 3rd party benchmark program floaring around the community, this is made by madboris. the settings you apply in that benchmark may not stick.

after you are done with the benchmark the 64 bit gpu application has a bug where it might close prematurely after the run so you wont get to see your results. you can simply run the batch file again and wait for it to end or you can end the process prematurely and it will spit you back out to the command prompt screen where you can view your previous run.

EDIT; just run the 32bit version as it benches much better and does not bug out.

2) metro 2033 also has a built-in benchmark utility. set all settings to max, dx11, very high, advanced DoF enabled and the 4x AA option. keep physix to off since we dont have nvidia cards. do only 2 runs as you wont need more than one to load everything into memory.

3) crysis 2; this game does not have a built-in benchmark utility but there is an excellent section at the begining of the game that i use to check for fps and gpu stability. since this game is the most demanding out there to date with ultra-settings dx11 and high res texture pack, it should be used to test for gpu overclock stability. if the overclock is not stable (especially with artifacting) you should consider lower the clocks until you achieve stability otherwise the overclocks are meaningless.

the game or the gpu drivers have a bug (or at least it does for my system and dell 11.7 drivers) where keeping the settings to ultra will cause the xfire to bug out and wont run properly. set the game to ultra then set it back to extreme. now set it back to ultra again to ensure xfire works. have fraps running.

in the very begining of the game in the 1st level you after you performce the high jump for the first time, go open the door. there will be a cut scene where you grap an axe to break open the door and the camera will pan slightly to reveal NYC under attack and in chaos. at this very point do not touch anything, not the keyboard or mouse. simply observe the frame rate provided with fraps. the fps counter will be a little jumpy so make an honest attempt to take the fraps screen shot when the fps becomes stable.

i will get these up shortly.

 

can you explain why? nvidia cards have this.

 

560M can't. It's just a happy accident that the Nvidia cards do.

 

Meh, this is too hard to really coordinate in Crysis 2 since the FPS will be all over the place and hard to shoot a shot at the exact same time, render frame in which other's shoot. Variance will be too high.

There shouldn't be anything to settle with 6990 and 6970 as 6990 IS more powerful. I don't observe 20% "on average" across the boards in games and no review or benches are proving that either really. But it is a tangible difference and I would rather have gotten the 6990 if that makes a difference. However, as others have said, it's much more important to enjoy what you have. Honestly, by the time "580M vs 6990" is settled...the 7 series will be here lol. I am playing every game at settings perfect for me. Heck, most games I play would only require ONE 5870 LOL as I play a lot of CS:S and L4D2. 580M SLI I don't think I would want (and I don't have either that this thread is about) due to the throttling issues. I do like the reported temps though.

I remember back in the 6800 x800 days of ATI/Nvidia and Nvidia had the "faster" cards but after a long time, it was revealed they were sacrificing IQ for performance. It's really a never ending war. So debating a "war" is pointless because it can't be settled...wars don't end

 

ok well you can skip crysis 2 and do crysis and metro's built in benchmarks. i am curious to see the difference in real games between the cards.

 

I'll run Metro 2033 right now at settings you said. It's the most fair benchmark at DX11 and new. Crysis will probably be very similar and I think other benches show that because it's so old. Some games like black ops and starcraft 2 actually show higher FPS for some reason on the 6970 so again I don't rely too much on benchmarks because I think the proof is in the architecture and the 6990 IS a better card than my "lowly" 6970s Edit...I will NOT have OC on my cards either for the bench.

 

Imageshack - metrol.png

Sorry for the bad IQ. Basically it averaged 29.5 (loop1)-31.5FPS (2). I do hate running through steam as friends kept popping up and L4D2 must have had some update downloading as that was updating too. But you can see the minimum and max. Hope this helps.

 

here is my metro run. i forgot to mention that AF should be at 16x but thats ok. i left it at 4x to match your settings. your second run is cut off so just going off the first run to compare.

looks like there is about a 15% difference in fps between 6790 and 6990. i would say that is abotu in line with what to expect in real world gaming, 15-20% difference.

[/URL][/IMG]

 

here it is again but this time with 16x AF;

[/URL][/IMG]



and this is with physix enabled;

[/URL][/IMG]


again all these are with gpu at 825 core / 1000 mem. 2720qm cpu or bclocks are not touched so no other overclocking. ram is kingston 1600mhz cas 9.

 

ah what the heck, another one in case someone wants to compare.

no phsyix and no AA;

[/URL][/IMG]

 

in crysis benchmark i noticed something strange. for some reason the 64 bit version was getting significantly lower results and it also was not utilizing the gpu's at all times properly. the 32bit version did much better and it always utilized the gpu's so i assume that is the better version to play. here are both 32bit and 64bit versions maxed with 8x aa.

32bit;
[/URL][/IMG]



64bit got lower results for some reason;

[/URL][/IMG]