The GTX 280M Overclocking and Benchmarking Results Thread

Yeah, rename the Device ID to 0808 instead of 080A, reopen the BIOS and it should read your clocks properly, then once you are done, before saving, switch it back ot 080A, that is very important!

 

How do I actually rename the device ID? I did it via the Nibitor, but it's still not recognized. BTW, isn't the default device ID 060A?

EDIT:
Nvm, I renamed it to 0608 and it worked

EDIT 2:
Well, I've managed to flash the card with the modified bios. I'm currently clocked @ 570/1400/940, so I guess it's an improvement, regardless of how small it seems, I'll try out more benchmarks to see if the increase in temp is worth it.

EDIT 3:
Flashed my card to the default bios. I believe that the performance gain was too little to justify the increase in temps. I'll deal with 550\1350\920 for now.

 

yes, it is 060A. Just rename it and hit rescan bios or something like that. but always remember to change it back to it's original device Id. I didn't once and it took me half a day to blindflash my card in order to get it working again.

 

Anyone knows how long I should run furmark for a stability check? Using the stability test coupled with the extreme burning mode, I think my GPU was burning like it was practically thrown into a fireplace.

After 60 secs, it was in the 90s. That's saying alot since after reseating the GPU, I've never gotten anywhere near 90 after a few games.

 

games dont stress a GPU like a stress test .. they use like 70-90% .... its the best way to test for heat problems

 

Ah, I figured it was something like that, so... about the stability test, how long should I run it? And I've added a Vantage benchmark too, although, the GPU score is only 5400 points with the O/C. BTW, is there a spoiler tag or some sort ? this way I can wrap it around the pictures.

 

The norm for stressing on Furmark is 15 minutes, but in my book the GPU has to last for 20 minutes.

 

How I do it with furmurk is I let it heat up the GPU as close as possible to the throttling point. I also run AtiTools in parallel. If by the time the GPU is close to 4-5 C to the maximum throttling point and you haven't noticed any artifacts in AtiTools or no crashes then you should be safe. You should repeat the process a couple of times to make sure and just as Soviet said, try to run it for 15-20 minutes. If it overheats in less than 15 minutes, I would suggest to lower frequencies /undervolt etc. of the card.This is how I test my stable OC .

Also note that sometimes, even after running this test, if you do encounter problems in games (artifacts/crashes) it is most likely from the GDDR frequency, just lower that and you will be fine.

 

Well, I managed to do a full run of 20 minutes without crashes or artifact. The temp hovers around 101-102c. I thought it would crash no doubt with such high temp, but it didn't, kept on going until it hit 1200 sec. I would post a screenshot of it, but, I cant seem to find the shots I took using the F9 button, there's nothing about screenshots on the Furmark GUI.

 

Furmark stores the screenshots in it's program folder in program files.

Ouch, 000022. You can cook eggs on that beast. What clocks and voltage did you have flashed for the stress test?

 

Yea I know, but at least it didn't crash though. I was on stock voltage as the performance gained after flashing the card wasn't really worth the extra temp. My clocks're 550/1350/920. Even running ATItool, the max temp I got was 78, 81 when I was fooling around in Crysis Warhead. God, I don't want to run this test again, it's insane. The threshold for my card is 108c, that's like 6c away..

Anyway, Here's the pic of it.




As you can see, it's sizzling. if it helps, I'm not a person to use laptop coolers? Did every test without a cooler.

 

Goddamn man. Your card must be crying from second degree burns right now. What surprised me the most is that it didn't crash at that temperature. 108*C is the maximum tolerance temperature for the GPU core only. You're lucky that your card still works as the memory chips have a lower maximum tolerance temperature.

Here is the Furmark test I conducted not too long ago with the same voltage and slightly higher clocks, ( http://forum.notebookreview.com/showpost.php?p=4868651&postcount=370). I used different settings for my test, but it doesn't matter as Furmark will still stress the GPU at 100% for the entire duration. Even with my mods, the temperature gap between my results and yours is way too large. I need to find a logical explanation for this when I am not sleep deprived.

 

000022: Great work. Looks like you can safely use your card on the settings you ran furmark on. You might experience problems in very hot environments, but that is easily solved if you raise the back of your laptop a bit.

thumbs up!

 

With those temps you need to invest in a cooler or a new card soon

 

Really, it takes real big sweaty balls to pull off what you did tonight, 000022. +1 for you.

 

wow... The difference between my temp and yours is way big. I'm gonna get to the bottom of this though, I'll loop Framebuffer's crysis warhead benchmark @ ambush 80 times and see if the temps really do get that high.

I don't want to be gaming on anything above 95c. Well, there wasn't any artifact during or even after the furmark test, so I guess my memory chips were saved from being fried. I told myself during the test that if here was any hint of artifact, I'd be pressing the esc button asap.

 

Dont run it that high, at that temp its gna downclock and loose performance anyway knowing nvidia, i know ATI does.

 

Well my throttling speed on the 9800M GTX is 90C. And with the last drivers I used to get to the throttling speed on some of the very hot days that were here last week. I solved the issue by raising the bottom of the laptop and now the most I get is 86C.

 

Strangely enough, it seems that only furmark has the ability to push it past 95c. Heck, Any other games (Last remnant, Crysis, GTA 4, DMC and of course COD 4) wouldn't even go past 90c, it's constantly hovering around 78-85c.

I've never seen the card downclock at all, I don't know if that's a good thing or a bad one. I'm leaning towards good as those games don't run as hot as furmark. In fact, the games benchmarks scores are kinda similar to the ones on the first page, so I guess my GTX 280m is performing at its fullest.

Am i doing something wrong?

I do that too, with my wallet, most of the times. I travel around alot, that's why I'd never invest on notebook coolers, unless of course it fits in a pocket. But I'm not doing that during benchmarks just for the sake of benchmarking, I guess, others might want to place a GTX 280m in their NP8660, so I think they'd want a heads up on the temp.

Although I might be the one who's doing something wrong resulting in high temps.

 

No, I think your temps and performance are just fine. I've tried putting a 9800M GTX in a NP8660 but the temps were way out of scale and the heat sink was inappropriate for the 9800M GTX. Just so you know, I think your card is doing just fine in that 15inch chassis.

 

Just to let you know I installed a GTX 280M in my Alienware M15X also and my Furmark temps get to a high of 97-99C degrees after running it at 30 minutes. I have ran this test with stock GPU clocks and also with clock of 600/900/1500 and it still gets to a high of 97-99C dgrees. When I did the Original Crysis GPU test and modified the .cfg file to loop 60 times, my temps got to a high of about 92-94C degrees that is at the highest resolution I can do of 1440 x 900 with 8X AA, everything on the highest settings. I have heard that the original Crysis is much harder on the GPU than the WarHead version. I hope this helps. God Bless

 

Alright, 80 runs of Framebuffer's Crysis Warhead, I'd run Crysis, but I don't have it, Jdeluna. Max temp at 83c.



EDIT:

Guys, after running Crysis Warhead for 80 runs and maintained a temp of 83c, I find it hard to believe that furmark actually made it go up to 102c. So I ran it again @ 1680*1050 with 8x AA. Guess what? it stayed at 82c for 30mins. Here's a screenshot of it. I don't know why this had happen, I thought that the card as downclocking, but nope, according to GPU-Z's log, it maintained its O/Ced clocks. I think it's because of where I placed my laptop initially, a cloth was beneath it (hadn't noticed), but I didn't think that the piece of rug would've made 20c of difference?

This is great. I am now convinced that the GTX 280m can be supported by the NP8660 without fear of high temps (unless if you placed it under a cloth?)

 

You can try downloading the Original Crysis DEMO here.


http://www.crysisdemo.com/download.htm

 

Thank you. Although, I'm almost done with stressing the card. After the thing with furmark, I'm sure that I'll not get that RROD...oh wait..

 

Now that's more like it. Those temps are good to go with those clocks. The GPU fan is quite strong and is capable of sucking that piece of cloth into the intake grill.

 

I ran Furmark on my P9600 GTX280m NP5797 and hit 95 degrees at 600 950 1500 and my driver crashed and had to recover after about 5min of testing.

I can't figure out why I'm having these GPU issues.

185.85 Driver, Vista x64

EDIT: And I'm flashed to 1.05v

 

phifer. 600/1500/950 doesn't need the extra voltage to run stable. Also, it's not a GPU issue. Every card is different in terms of manufacturing quality and yours may have been made on a Friday. Stress testing allows us to find the limits for our cards and set the appropriate clocks in consideration to how we use them.

Generally, the rule of thumb when overclocking a GPU is to find the highest stable core/shader clocks at a given memory frequency. From my observation, everybody in the Clevo forum seems to be doing the opposite; cranking the memory first and then the core/shader, which results in an unbalanced core/shader:memory ratio, and ultimately gaining minimal performance compared to raising the core/shader first.

 

i do the shader and then what ever floats my boat. lol.

but my ati was core then memory if i did it the other way around it would render unstable and downclock.

 

Well, I'm at stock voltage and clocked @ 550/920/1350. But the GPU's core and shader is practically 600/1450 respectively after taking into the account of my increased FSB via setFSB.

So I don't think that the extra voltage is needed for you clock, maybe wrong though. But even so, that extra 30 mhz of memory clock isn't worth the temp increase when you raise the card's voltage, imo.

 

I always start the OC with the core, the shader clock is just the ration of the core so it's nothing to overclock there, as in I don't do a separate stable OC lookup for it. The memory is indeed the last one.

 

finallyyyy i got mi np5797 today, it was a long wait.

i've intalled windows 7 but my fingerprint sensor doesn't work, i dont know if is a compatibility issue.

i'll post some bench later.

 

The reason I was running a higher voltage flash is because I was running at 625 975 1600 and 1.05v. This was pretty stable for me when gaming, etc.

I decided to run furmark and my temps hit 99 degrees celcius and Furmark crashed so I downclocked to what the nvidia software said was my "factory shipped frequencies" of 600 950 1500. Ran furmark again and same results.

Interestingly enough, I can reflash overvolted and run Furmark at 625 975 1600 stable if I have 8x MSAA enabled. If I have 0x MSAA on I run 10-15 Degrees hotter and Furmark crashes.

The deciding factor for me is not what my voltage or clocks at but what my MSAA is set to. (Which doesn't make sense to me, but I've tested it multiple times always with the same results)

 

Try doing the same but keeping the memories at 800. See what happens.

 

ok running crysis with stock clocks and 186,03 drivers from nvidia. my tems reach 94 and i'm afraid because there is no oc and no overvolt. what do you think guys

 

What is the temperature in the room? Is your laptop sitting on a desk or a notebook cooler? Have you replaced the huge amount of thermal paste that Sager splooged onto your GPU?

Also, do not ever let your card surpass 90*C, 95*C at most, during Furmark. If the temperature continues to rise, terminate the stress test.

 

Try resitting your GPU? Maybe there's no good contact.

 

room temperature is 25 degrees. i bought it at xoticpc. the cpu temps are fine but the gpu is very high. i think that there is no good contact i'm goint to wait to finish a download and then open the laptop and see whats going on.

by the way, no cooler i'm on a flat surface

 

make sure you are not on a table cloth

 

I'm sorry that I haven't bothered to read through all 64 pages of this thread... But I must ask a question:

Does the EVGA Precision OC utility work for mobile gfx cards, including the 280m?

http://www.evga.com/PRecision/
(It works on Non-EVGA cards, best tool I've used so far.)

 

jupapri,

open the notebook up and re-secure / re-seat the videocard.

also, you might want to check if the videocard's thermal compound is sufficient.

 

make sure the fan is running on high...

 

finallyyyyy my top temps are now 84 degrees, i remove the heatsink and there was a loooot of as5 spread, so i clean the die and re applied a thin layer of as5. the temps went down 10 degrees running ati tools scanning for artifacts. now is below of 90 thats very important for me

 

glad its back to under 90C

 

gophn. i dont know if the scanning for artifacts in the ati tools stress the gpu 100%. i want to run a test that stresses my cpu 100% and know the top temps

 

i would forget ati tool for nvidia cards...use furmark..if you want to sress your system and find out if it will make the grade...then use that and run it at your max screen resolution and max AA. that's all you need...and watch your system start to cook...

 

where i can download that john. what is the name of the program. sorry for the noob question

forget about that. i read futuremark haha no furmark. now i'm downloading

 

http://www.ozone3d.net/benchmarks/fur/

 

woooowwwww. 100 degrees in 1100 seg. this is really an stress test. is my gpu wrong? or is normal this result of furmark stress test

 

the furmark does stress your GPU to the limit.

so do not worry about it too much.... no game will do it like this.

you can toggle the fans to max speed... but that will only help it so much.

your system is fine, do not push it.... unless if you are willing to.

some have undervolted their videocards as well... keeps it a bit cooler without losing performance.

 

the fan in the bios are set to continous. tomorrow i'm going to play crysis and see if the temps reach the same temps before all this operation.