[BENCHMARKS] The OFFICIAL Thread

[Khenglish]

8 hours ago, johnksss said:

About what core clock did it start throttling at?

 

If I run a vbios that goes over 1.2v I get a performance reduction without throttle at any clock or voltage, with more reduction the higher the vbios can go.

[johnksss]

Then it's getting closer.

You will lose on stock, but you can gain when over clocked.in the 1.5 ghz+ range.

But there is a balance though...

[Khenglish]

1 hour ago, johnksss said:

Then it's getting closer.

You will lose on stock, but you can gain when over clocked.in the 1.5 ghz+ range.

But there is a balance though...

 

I've tried max overclocks at 1.2v, 1.25v, 1.275v, and 1.3v, with 1.2v beating all.

 

I only gained 20mhz with 1.25v, 30 with 1.275v, and 40 with 1.3v. The performance hit is too high for these small clock increases.

[johnksss]

On 12/23/2015 at 7:43 PM, johnksss said:

It's all to much voltage technically. And it has to have a balance. This trip, it might be base off of two mods. One for higher overclocking and one for lower over clocking, but with a higher performance rating on air. Just like what Vince did for the Kingpin Classified 980 ti.

 

For me, it's a work in progress..

That is why I made this post earlier....

[chap]

1.2v voltage offset? wow thats way to go, i have only the "normal" prema mod bios, id really like to get up to 1.2 but is it fix? becaue i would only need those voltages for benching

it would be perfext if it was possible to ov with nvinspector up to 1.2

 

is your gpu completely shut off in idle khenglish? and why cant i pm you?

[Guest]

5 minutes ago, chap said:

1.2v voltage offset? wow thats way to go, i have only the "normal" prema mod bios, id really like to get up to 1.2 but is it fix? becaue i would only need those voltages for benching

it would be perfext if it was possible to ov with nvinspector up to 1.2

 

is your gpu completely shut off in idle khenglish? and why cant i pm you?

 

Huh? All my GTX9xxM vBIOS allow up to 1.2v, just use NVIDIA Inspector to overvolt.

[chap]

so its 1.2 vgpu overall, not adaptive?

[Guest]

It's only applied to 3D Turbo clocks (P0).

[chap]

yeah i know, i only never checked how much total vgpu i got when i put the +137.5mv offset on, well then khenglish won the silicon lottery

unfortunately not me

[johnksss]

53 minutes ago, chap said:

yeah i know, i only never checked how much total vgpu i got when i put the +137.5mv offset on, well then khenglish won the silicon lottery

unfortunately not me

He actually didn't win the silicon lottery. That came from a prior skillset he already possessed. And with a premamod to help push it even further.

 

It's alot less luck here than there is skill...;)

[Thumper_23]

http://www.3dmark.com/3dm11/10759084

[Guest]

 

Very nice!

Come on you can do 25K total score...Try a newer driver!

 

You are getting closer to John's 32K GPU score record...

Can't wait to see what@Johnksss can do with@Khenglish vRAM Mod...

[Thumper_23]

4 hours ago, Prema said:

 

Very nice!

Come on you can do 25K total score...Try a newer driver!

You are getting closer to John's 32K GPU score record...

Can't wait to see what@Johnksss can do with@Khenglish vRAM Mod...

What is that vram mod all about? I've tried newer drivers but even though they give me a liitle more gpu score my cpu score goes way down.

[Guest]

28 minutes ago, usmc362 said:

What is that vram mod all about? I've tried newer drivers but even though they give me a liitle more gpu score my cpu score goes way down.

 

There are two vRAM Mods:

 

I. A voltage Mod (simple):

 

https://www.techinferno.com/index.php?/forums/topic/8820-hardware-mod-gtx980m-memory-overvolt-guide/

 

II. A total vRAM chip swap (the high road):

 

https://www.techinferno.com/index.php?/forums/topic/9021-hardware-mod-gtx980m-hynix-to-samsung-memory-swap/

[johnksss]

Yep, he is totally right about the cpu score. I just really noticed that just the other day.

 

Here is a 32.6k gpu score, with windows 8. And with windows 7 it was 32.3k

http://www.3dmark.com/3dm11/10686823

Edited January 4, 2016 by johnksss

[johnksss]

http://www.3dmark.com/3dm11/10783428

 

http://www.3dmark.com/fs/7149225

[Guest]

Thanks to @Khenglish for fixing my screwed up 980M. It works great. I added the extra MOSFETs to my other 980M to match the one he fixed for me.

 

Interesting observation... seems the Panther 5 draws almost 100W less benching on AC than it does on straight air. Must be an efficiency loss at higher temps that starts causing higher power draw. But, I was really surprised at how much change. These runs peaked just a hair over 700W.

 

Not as good as @johnksss and @usmc362 are getting, but a personal best in 3DMark 11. This is using a @Prema vBIOS mod. 

 

http://www.3dmark.com/3dm11/10812335

 

 

This one is not a personal best, but it completed the run successfully at 1470 | 1450 | 1.200V. Actually, three runs with that voltage, but this was the best of the three.

 

http://www.3dmark.com/fs/7221027

 

 

Sky X9 is actually kicking my Panther's butt in Fire Strike since CPU performance doesn't add much to the overall score.

Here is a photo of the MOSFET mod.

 

 

 

Edited January 15, 2016 by Guest

[johnksss]

I don't understand how your gpu score is so low for that overclock.....

 

And to add... No way a Sky9 should be beating your Panther.

 

 

[Guest]

13 minutes ago, johnksss said:

I don't understand how your gpu score is so low for that overclock.....

 

And to add... No way a Sky9 should be beating your Panther.

 

 

I know, right. And, if I knew why it wouldn't be so low because I would fix whatever is causing it unless it meant spending money on new GPUs. I'm not going down that road. I just hope the one the Khenglish fixed for me stays fixed and doesn't start doing stupid stuff again after a while.

 

I've never had GPU scores that come anywhere close to what you and usmc362 are getting from this pair of GPUs, but then again this particular pair of 980M cards have never been good ones. They've returned mediocre results since day one. Even in the testing you and I did a long time ago using Team Viewer they almost always returned results lower than expected in the R2 and the Panther. And, they've always had issues with the screen going black at high overclocks unless I keep them freezing cold.

Edited January 15, 2016 by Guest

[johnksss]

Exactly. And benching on room temp is a quick way to kill parts sooner than later though. It's possible to bench like that on room temp, but you run the risk of a lot of other things going wrong.  IMHO

Edited January 15, 2016 by johnksss

[Guest]

I wasn't today though. And, they're still too low. Only difference today was the screen didn't keep going black between 70-75°C like it does without AC. These GPUs are junk and so were the 980M cards in the Sky X9. And, one of each pair died. And, they both had black screen issues. And, @ssj92 has two 980M that run like crap (actually a lot worse than these two ever have, so it could be worse for me and I am thankful it is not) and both of his GPUs are already dead. They died in less than a month, about a day apart from each other. So, yeah... it could be a lot worse and maybe I should not complain too much.

 

That's why I said I'm not going down that road again. They're way too expensive and a huge crap shoot on getting a good pair of 980M with only a 90 day warranty. I will just wait and try my luck again on a getting a good pair of GPUs in my next new machine with at minimum a 1 year warranty.

Edited January 15, 2016 by Guest

[johnksss]

And even the ones that came with the sky9 are bad also?

[Guest]

12 minutes ago, johnksss said:

And even the ones that came with the sky9 are bad also?

Both of them had issues with the screen going black exactly like those in the Panther, and one of them is as dead as a doornail. Not even visible in the BIOS. I sent them both back and only using the 200W GTX 980 now. Now, that's one nice GPU. I wish there were two of them in the beast.

 

Those in my Panther actually can overclock higher without a black screen on AC than those in the Sky X9 were able to. They could barely handle 1425/1425 without the screen going black on AC and about 1325 on core without AC. Quality control on these Clevo 980M GPU seems like it's not very consistent. I can't say I just have bad luck, because mine seems better than Brother @ssj92 bad luck, LOL.

 

The only thing wrong with the GPU that @Khenglish fixed for me was bad MOSFETs, so maybe that's what the problem is with most of them that don't overclock well. I did not replace the three original MOSFETs on the one GPU. I just added 3 to the empty pads. The one he fixed for me has all 6 MOSFETs replaced.

 

Edit: One thing for sure, when you do get some good ones, they're really good. Yours, Harlons and Brother @GTVEVO are all excellent. I just wonder how many people have really crappy 980M cards and don't even know because most of them don't overclock.

Edited January 15, 2016 by Guest

[johnksss]

The main reason being that they used 3500 mhz memory with better timings. (Desktop 980)

 

Hummmm, And when these gpu heatsinks get put on, they are actually sitting firmly on those mosfets?

 

Still trying to see if my mosfet mod did anything different. I'm still able to do the exact same numbers as before the mod.

[Guest]

15 minutes ago, johnksss said:

The main reason being that they used 3500 mhz memory with better timings. (Desktop 980)

 

Hummmm, And when these gpu heatsinks get put on, they are actually sitting firmly on those mosfets?

 

Still trying to see if my mosfet mod did anything different. I'm still able to do the exact same numbers as before the mod.

It looks like it to me. The thermal pads have nice deep impressions in them from the MOSFETs. 

 

I don't think adding the MOSFETs actually helps overclock higher as much as make the GPUs more stable. Before each GPU had 6 I could not run 1.200V even stock without crashing. Now I can use higher voltage without losing stability and my black screen problem is quite a bit better now. It still does it, but takes a higher overclock or longer/hotter before the screen goes black.

 

It's really obvious without AC cooling. With only three MOSFETs on each GPU 1400/1400/1.150V would go black about the middle of 3DMark 11 Test #1 and almost at the end of Fire Strike Test #1 every single time. Once in a while it could get part way through Fire Strike Test #2, but never, ever, in 3DMark 11. But, just adding the MOSFETs the Panther can now run that OC and finish both benchmarks more than half the time without the screen going black. .

Edited January 15, 2016 by Guest