580m vs 6990m Benches

ARGH are those stock 6990 clocks? My Metro was run with stock CPU/GPU settings. The 6990 is a beast no doubt. I think i'm gonna wait to 7 series. By then I can justify an upgrade drop in...plus a nice Thermal repaste by then

I only run 800/1000 when gaming as that is a very safe OC that keeps my MemIO in check. My overall temps are 71-77 but the MemIO can get to about 90. I always watch each diode because overall temps I am not concerned about..if one diode goes....bye bye GPU. Even at stock the CF system runs Crysis 1/warhead/metro beautifully. So I really can't complain..I do wish I waited until the "outlet was flooded" as i could have spent 1800 instead of 2400 for my system. Oh well; this life and tis technology..the train that never stops

 

my cards are bios-flashed overclocked to 825 / 1000. i thought you were at 810/1025 clocks.

 

Nope I ran 680/900. I am not bios-flashed. I just use trixx. I will run my 800/1000. I can be stable at about 830/1050 but the temps are too high for me. I will go run metro at 800/1000; not quite your speeds but will post results.... How do you just post the image to the screen....going through image shack is a PITA!

 

http://imageshack.us/photo/my-images/225/metro2x.png/

Sorry again for the cut off. I need to know how you are posting your images as I am using mspaint! lol. But at 800/1000 it gives just under 33FPS. Again, these are very close in performance real world gaming. I am surprised all the extra shaders are not really being utilized in these games....maybe next gen ones?

 

bios-flashed back to stock clocks.

heaven benchmark;

[/URL][/IMG]

metro run like first one to compare with yours, 4x AF;

[/URL][/IMG]

 

ya..very close. See my post above at 800/1000.

edit: At stock again Metro2033 "average results" was 29FPS to your 31.

 

yea and again yur using FRAPs which does (ironically) cut down on your frame rates.

I know every card is diferent but its weird that I can run at 850/1150 no problem for gaming ... temps hit high 70s low 80s ... and I am ok with that. Seems other people take a more conservative approach. I mean for most games totally not needed, but i like the extra FPS and for SC2 on some of the crazy pvp custom maps its needed

 

So it appears at stock and overclocked (to close to the same settings) is about an almost 10% difference which was what I was expecting. But the 6990 seems to have a nicer OC headroom than the 6970. However, it appears that the results of OC 6970 tend to yield better results at lower clocks. The reason I am saying this (relatively based on a small sample) was that on average, the stock clocks on the same bench at same settings yielded a 2FPS difference which is just under 10% difference. Now at your bios-flashed 810/1025 to my 800/1000 it yielded again about a 2FPS difference. Interesting......

 

Shaden can you post your MemIO after gaming one session. I can't believe that is high 70's low 80s? What thermal paste did you use to repaste? Those are VERY LOW temps if that really is including every diode? My temps are high 70s low 80s as well but my MemIO is always around 89.5 at 800/1000. If I go to 830/1050 it goes to 94!!! (I am on stock paste ATM).

 

Hey dave maybe i am looking at overall temps .... i will check. its my bachelor party tonight on the way to the hotel now... but i will doublr check and post. stock paste ... oddly enough. I must have the best stovk paste job and heat sink placement going though luckily .... and randomly. my 6970 system must have had the worst though

 

don't go get into trouble either buddy!!

have major fun tonight.

 

Congrats my friend. Wishing a happy marriage! I sure wish mine lasted:/ Probably the fault of AW gaming lol. Take care and keep it between the ditches tonite my friend!

 

So are we concluding that the 6990m is better? I havnt seen much defense from the 580m people. I was really thinking about getting the 580m because of adveratised better tessellation reasons but does the 6990m do the job just as good or even better? Is this because of the 11.8 catalyst driver? Because I recall 2 weeks ago everyone was claiming the 580m was better.

 

How do you raise the 580m to 0.92 volts

 

No chance, we're just having too much fun with games It's already well documented that 580M are the king of the mobile performance in most games

 

bah, disagree, turn on overdrive and 6990s would beat the 580s at almost every game ...

 

PS thanks all, had a great time last night and have lived to tell about it ... it was like something out of the hangover ... went skydiving at 14000 Feet, drank on the hotel roof, hit the casino and won some money, hit the club and had a VIP booth with bottle service ... it was ... crazzzzy awesome.

 

If it makes you feel good man, I'll agree. I also believe my Jeep Grand Cherokee is better than the Audi Q7, makes me feel very good about paying half the price too

 

nothing personal..but until an aliwneware m18x with 580's beats me. my 6990's are king. haha. now if i decide to grab 580m's. and hold the records..my speech will be exactly the same as with amd..

 

hahaha John, nice.

 

yeah shaden...had to lighten the mood around here...reminds of the 9800/3870/4870 days.

 

lol well since my last system is a 8600GT I think I missed those days

 

So what do you think will run battlefield 3 better. 580M or 6990M?

 

had that as well. 8600gt/gs 8800gts/ 9800gt/gtx 5870 6990

 

4MB AGP card
Voodoo3 3000
Radeon 8500
Radeon 9800pro
Radeon X700pro
Radeon X600 mobility
Geforce 6800U
Radeon x1900GT
Radeon HD3870
Radeon HD4850 (then crossfire)
Radeon HD4770 crossfire
Geforce GTX280
Radeon HD4870 crossfire
Radeon HD5870
Radeon HD5830
Radeon HD5850 crossfire
Radeon HD5670 crossfire (getting married)
Geforce 9600M GT
Radeon 4670 mobility
Geforce 240M GT
Radeon 5730 mobility
Radeon 6770M
Geforce 560M
Radeon 5870M

I think I've missed a couple out...

 

I think it's a "played with Nvida" type game so that should yield slightly better results.

 

So is one faster overall?

Is Nvidia worth the bucks or not?

Is it splitting hairs?

Be nice if the overall results of 38 pages could be put in the first post by
OP.

To call getting information from this thread unwieldy is to do a dis service to unwieldy.

 

the thread has been summarized

 

I know you don't want to hear this, but with the Bar-Code Bios and the extra juice, it is absolutely worth the extra bucks, as it stands they are dead even but all of the 6990 run OC while 580M have not.... I will be updating this thread with actual numbers, but if Bar-Code is correct and the cards run stable at 900/1800, then we're looking at a big % increase...

At Stock the 580M are better... (but don't take my word for it)

Review GeForce GTX 580M SLI vs. Radeon HD 6990M CF - Notebookcheck.net Reviews

The 580M is more expensive but overall a better package and without OC!

 

hummm...so when does barcode plan on posting this stuff?

side note:
and those are clevo 580m's not dell 580m's. sure hope these 580m's aren't broke (so to speak)

edit:
ok, found it.
looks impressive, but if you can't use the clocks...then that sucks big time.

 

well first of all you yourself have a pair of these 580m's and did not post a single bench with proof.

second of all, a 45% overclock on the core for 580m, lol?! i dont how much voltage or cooling you apply to it, whoever is talking numbers like that doesn't understand what stability means.

how about providing a crysis 1080p bench with 8x aa?

 

He seems to indicate that he can game with them no issues, the problem that he's having is with 3DMark/Furbar which 6990 have no issues, howevber those games kinda suck in my books

 

im a bencher/over clocker. so everything is open to interpretation. i bought this machine to bench and have the ultimate mobile tech machine...playing games with it all day is fun and all...but not really paying for itself.

side note: crysis 1 66 fps no gpu over clock very high.


but on a serious note...did you flash your cards? and what sort of fps are you doing all around? thanks.

 

Well to each it's own I guess.... I'm definately not a bencher, I like to play games with my M18x, benches are fun to me for the first couple of hours .... but hey more power to you, it's good to see someone pushing the limits of thes machine and we can all learn a thing or two from people like you

That said, looking at your numbers and mine, I'm very happy with the performance that I've seen with my machine without coming close to all the time you put into OCing.

 

66 fps? can't be. i scored much lower with overclocked gpu's. maybe it was your cpu overclock? please only use the in-game 32bit version benchmark gpu batch file and not 3rd party program. first run the game in 32bit mode and ensure all are very high spec and no AA. the 64bit version runs the game slower for some reason so just stick with 32bit.

[/URL][/IMG]

 

for sure the cpu over over clock. i ran the bench in dual monitor mode so i could watch what was going on.

crysis 1 was racking in a 50W on cpu usage and all 8 threads...
while warhead was using much more gpu processing and 35w cpu across less than 8 threads.

 

Hey Hey Johnksss,

Thats the problem we/i cant provide bench test to prove the performance gains as it gets throttled by nvidia, but in house gaming is currently not throttled for me running at 900mhz, because the actual game/s is not extreme testing loading the card through thats much power/load (no game would load the gpu as much as the bench tests do) This 900mhz was performed with gpu fans on full bolt just to be safe (using hw). Everything runs smoothly no artifacts or glitches (with 0.92v).
But as i said in an earlier post to me this is not worth running 24/7 just to gain a fps in games.

Its a shame this cant be completed in benches, but soon as i figure this throttle issue i will post benches at 900mhz eventually. lol

Until then this should be classified as speculation only. No proof no results! This shouldn't even be discussed in the bench thread full stop! I hope this helps.

 

hey hey bar-code...was wondering where you were at.

what game you got that can run that in game? and outside of games...what is the highest you can bench?
and did you trying benching in balanced mode...some times it tricks the system....some times.

 

Anything above 740mhz in any bench test throttles full stop.

I ran the 900mhz gpus on a in RTS game called starwars Empire at War (dont laugh). Hey hey im a RTS fan

The fans had to be full bolt to stop the temp throttle kicking in but my gpu never reached above 72c

Balanced mode doesnt work nor does renaming furmark etc.

Currently working on an older set of drivers that dont contain the drivers throttle commands for any stress test program as soon as i have them working i'll post some results, but nvidia has changed lots of stuff in the newer drivers and its making it difficult to get the 580m to work on these older drivers.

 

nothing wrong with that. to each his own.

is there throttling in the clevo cards? i haven't been over to investigate in a while.

 

Try running a modern title and the power consumption can get MUCH closer to those benchmarks. You might find it throttles then.

45% overclock and over-volt could easily mean a 60% max power increase.

Can your machine handle 320W GPU power consumption? Add a 55W CPU and you will want a 400W power brick. At least.

 

Dell oem gtx580m would oc better since Dell uses full mlcc on mos circuit while the clevo to keep costs down only users 4 poscap.

But clevo's dont have the throttle, which i haven't confirmed yet!! Clevo have been able to run benches at 860mhz with no throttle.

So the question remains what is causing the throttle. Could be a hardware component as well.

I know my sli desktop computer is running the asus 560ti sli configuration standard clocks at 830mhz no throttle overclocked to 960mhz again no throttle. So go figure!!

 

Your probably correct, i beleive the clevo users where running double power supplies or above 330w anyway, so thats probably why Dell have throttle power load issues, until this is figured out, this is all rumors and unproven tests.

And you know what, i'm so determined to figure this throttle out and will blow the 580m up on purpose! Just to see how far these can go!! ( so there nvidia) lol and yes you will replace them free of charge ^o^

 

i would get throttle on my evga gtx580 when over clocked it high. had to put the ovp/ocp on to run 1200/1200. same with my pny 580 and my direct cuII 580...

anyone hook up a second monitor with all the other software running on it...and just sit and watch to see exactly when and where things go bad?

 

How do you run the GPU Benchmark with 8xaa? is there a parameter?

 

Crysis 1 only utilizes 2 threads. You must have had something else running in background

 

Launch the 32bit version of crysis from bin32 folder. Ser the in-game settings and exit. Then from this same bin32 folder launch the benchmark gpu file. It will benchmark the game with the in-game settings you set. Afyer benchmark is complete it will show the results. Take a snapshot of that.

Avoid the bin64 folder as the game will run slower in 64bit mode for whatrver reason.

 

Gotcha, thx.

 

run a dual monitor display and you'll have your answer..

and you can use the 1.5 version of the benchmark. we use to use the in game bench because the earlier versions of the bench tool would not force the settings on the bench. instead it would run what the game was set to. and people would set it to 800x600 and then claim they had some out of this world gpu's. to damn funny that was when they found out their runs were garbage...haha

 

Hey Xen, I am just not sure what you are saying here and I don't want you mislead people asking for clarity on a post which I thought I provided quite well and without bias.

"as it stands they are dead even but all of the 6990 run OC while 580M have not"

I believe they are dead even at stock (and again without running Overdrive on the 6990s) They are according to the FPS posts in the review you point to, only 1.1% difference (edge going to the 580s to be sure)

IF Bar-codes stuff works out, then awesome, those 580s should be beast, (Which even Bar has been careful to note is an IF) but I just don't understand how you can use Stock stats when you want (while discounting 6990s for OCing) and then use (Potential) OC'd stats when you want. Because if you were to be truthful with the person asking, then the 6990s are OC'ing like beasts ... right now ..... and with insane ease, DL Trixx, Chose your clocks, click. Done.

I thought I was pretty fair at pegging them dead even, and actually giving the 1.1% to the 580m although I disagree with the methodological approach used in the testing. Except with a huge price differential, that I didn't even mention....

**and I don't mean any of that as fightin words, just wondering what you meant and if you disagree'd with the over all assessment, since people were asking for guidance.