Jump to content

[BENCHMARKS] The OFFICIAL Thread


Brian

Recommended Posts

Adding the MOSFETs might even lower memory OC potential without a voltage mod to go along with it. That was the case for @Meaker and it kind of seems like it on these also. This pair has never been stable with a high memory overclock so it's hard to tell. They may not be any different in terms of memory overclocking, but I for certain know the memory overclock is not improved. 1425 to 1450 is the limit. 1475 almost always starts having severe artifacts and I have to stop the benchmark immediately to avoid damaging them. I tried 1465 today with the new MOSFETs using AC and it's not stable with the memory that high. 1450 or less is stable.

Edited by Guest
Link to comment
Share on other sites

Just an FYI, the screen going black is the GPU doing an emergency shutdown. I got them too on my MSI card when I only had 3 FETs. Since I'm on a P150EM with optimus I still have an image after the GPU shuts off. The GPU completely disappears from the system and the Nvidia driver says the GPU was unexpectedly removed. I'm not sure why having 6 FETs prevents this.

 

Also, whoever designed the inductor voltage loss compensation circuit on the 980m did it completely wrong. It looks like the circuit was copy/pasted from a much weaker, lower power GPU, with no adjustments made for the big inductors and 3rd power phase on the 980m. The 3rd phase and big inductors mean there is much less voltage loss at a given current, so the compensation circuit is boosting the voltage much too high. At full load with 1.062V VID my real steady state voltage is 1.11V. With 1.2V VID my real steady state voltage is 1.265V. I'm pretty sure this extra voltage is worthless for overclocks since it varies with load. Voltage will be at VID at light load, so you never get the full elevated voltage at all times for higher stability.

 

I think I figured out how to fix the overvoltage. I will be attempting my idea either later tonight, or tomorrow.

 

17 minutes ago, Mr. Fox said:

Adding the MOSFETs might even lower memory OC potential without a voltage mod to go along with it. That was the case for @Meaker and it kind of seems like it on these also. This pair has never been stable with a high memory overclock so it's hard to tell. They may not be any different in terms of memory overclocking, but I for certain know the memory overclock is not improved. 1425 to 1450 is the limit. 1475 almost always starts having severe artifacts and I have to stop the benchmark immediately to avoid damaging them. I tried 1465 today with the new MOSFETs using AC and it's not stable with the memory that high. 1450 or less is stable.

 

Those FETs are only for the core. Higher power draw for the core allowed by the higher voltage will put more noise in the 19V supply voltage to the card. This noise may add significant ripple to the memory voltage. I added a lot of supply voltage caps to my card and I had no change in memory overclock when adding the FETs, and no change with raising the core voltage and clocks.

Edited by Khenglish
  • Thumbs Up 2
Link to comment
Share on other sites

1 hour ago, johnksss said:

@khenglish

 

When you repaired his gpu, did it over clock the same on your system?

 

I only tested as high as 1450/1400. When that passed 3dm11 I boxed the card up and shipped it back.

 

The supply caps I added were directly on the card, so motherboard won't matter.

Link to comment
Share on other sites

Overvoltage issue fixed. 1.062V vid is now 1.065V instead of 1.111V. I'm trying to see if I have a clock reduction. If I do, it's less than 10MHz, but so far it seems like none. Over 10% overall power reduction for the card.

 

Would anyone like a guide on the fix? It's just 1 resistor that needs to be modified.

  • Thumbs Up 1
Link to comment
Share on other sites

7 minutes ago, Khenglish said:

Overvoltage issue fixed. 1.062V vid is now 1.065V instead of 1.111V. I'm trying to see if I have a clock reduction. If I do, it's less than 10MHz, but so far it seems like none. Over 10% overall power reduction for the card.

 

Would anyone like a guide on the fix? It's just 1 resistor that needs to be modified.

Now don't take this wrong, but ummm.....quit asking silly questions! LOL

Of course we do! :)

 

 

Question though. How are you measuring the voltage for the card?

Edited by johnksss
Link to comment
Share on other sites

4 minutes ago, johnksss said:

Now don't take this wrong, but ummm.....quit asking silly questions! LOL

Of course we do! :)

 

 

Question though. How are you measuring the voltage for the card?

 

I'm measuring the voltage of the 6 big electrolytic core caps just above the memory with a multimeter. I can do this when the card is running using the older 2 piece heatsink with the memory/FET/inductor half removed, and just the core half in place.

 

Now through 7 Unigen Heaven runs at 1312/6930 (my 100% stable max overclock prior to this mod) and no crashes or artifacts.

  • Thumbs Up 1
Link to comment
Share on other sites

8 minutes ago, Khenglish said:

 

I'm measuring the voltage of the 6 big electrolytic core caps just above the memory with a multimeter. I can do this when the card is running using the older 2 piece heatsink with the memory/FET/inductor half removed, and just the core half in place.

 

Now through 7 Unigen Heaven runs at 1312/6930 (my 100% stable max overclock prior to this mod) and no crashes or artifacts.

Ah, got it. I Was trying to see how i would have to do it without soldering a temp wire in place.

 

So the current clocks at 1.062V? And is that the stock voltage for the card when running a stock test?

Link to comment
Share on other sites

14 hours ago, Mr. Fox said:

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.

 

 

Nice, but you used the wrong vBIOS...higher voltage range versions result in lower per clock scores even if you "only" use 1.2v.

Just redo the benches with regular PM v2 1.2v base... 

 

14 hours 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.

 

 

 

Right, see above... :thumbsup:

Link to comment
Share on other sites

5 hours ago, Prema said:

 

Nice, but you used the wrong vBIOS...higher voltage range versions result in lower per clock scores even if you "only" use 1.2v.

Just redo the benches with regular PM v2 1.2v base... 

 

 

Right, see above... :thumbsup:

Maybe I will try the new base later today and see if it improves the score. I haven't ever had as much passion for benching the Panther like I used to because my lackluster 980M SLI performance has always been so discouraging. It would be awesome if they finally run as well as they should. I was so excited to get the Panther after having them turn out to be such a horrible investment on the M18xR2. After spending $1400 on GPUs for the R2 upgrade and another $1200 for a base system that could actually use them without throttling and 8-beep UEFI crap, it was very deflating to find their performance was mediocre in the Panther, too.

Edited by Guest
Link to comment
Share on other sites

8 minutes ago, Mr. Fox said:

Maybe I will try the new base later today and see if it improves the score. 

 

Latest vBIOS base no longer works with the older driver...use the old base but the 1.2v version.

Link to comment
Share on other sites

4 minutes ago, Prema said:

 

Latest vBIOS base no longer works with the older driver...use the old base but the 1.2v version.

That's what is on it now... old base 1.2v version. I downloaded it from the latest link you sent me. That's what that Fire Strike score is run with, using 353.06 drivers. I need to use a newer driver than 353.06 with that vBIOS?

Edited by Guest
Link to comment
Share on other sites

16 minutes ago, Mr. Fox said:

That's what is on it now... old base 1.2v version. I downloaded it from the latest link you sent me. That's what that Fire Strike score is run with, using 353.06 drivers. I need to use a newer driver than 353.06 with that vBIOS?

 

It is a 1.25v vBIOS, not a 1.2v vBIOS that you used there for benching, hence the score-loss. 

I never made a 1.2v "84.04.22.00.C1" version until the one I sent you a couple of hours ago after seeing your bench... ;)

 

Use that one with a driver that you can compare to the ones used by Kpaxx and John (347.88 or 344.75 for Firestrike), those driver no longer work with the latest vBIOS v88+ base. They also used W7...

Only then you can compare your results...I bet your cards are just fine.

Link to comment
Share on other sites

5 hours ago, Prema said:

 

It is a 1.25v vBIOS, not a 1.2v vBIOS that you used there for benching, hence the score-loss. 

I never made a 1.2v "84.04.22.00.C1" version until the one I sent you a couple of hours ago after seeing your bench... ;)

 

Use that one with a driver that you can compare to the ones used by Kpaxx and John (347.88 or 344.75 for Firestrike), those driver no longer work with the latest vBIOS v88+ base. They also used W7...

Only then you can compare your results...I bet your cards are just fine.

@Prema - that vBIOS is much better, bro. If I can get that combined score looking a little better should be awesome. That's the only thing allowing the Sky X9 to beat my Panther now. Combined score is way low, but graphics and physics are looking good. Finally broke my personal best Fire Strike score from 6 or 7 months ago. 

 

http://www.3dmark.com/fs/7234823

 

vdqZmFG.jpg

Edited by Guest
Link to comment
Share on other sites

http://www.3dmark.com/3dm11/10841880

http://www.3dmark.com/3dm/10413482

 

Power consumption (180W PSU):

 

Spoiler

9Cigkx4.jpg

 

@Prema does the 200W PSU that comes with the P65xRx work with the P65xSx? Dell/Alienware 240W/330W ones with the converter cable do not, they shut off under load.

Edited by octiceps
Link to comment
Share on other sites

  • 2 weeks later...

Join the conversation

You can post now and register later. If you have an account, sign in now to post with your account.

Guest
Reply to this topic...

×   Pasted as rich text.   Paste as plain text instead

  Only 75 emoji are allowed.

×   Your link has been automatically embedded.   Display as a link instead

×   Your previous content has been restored.   Clear editor

×   You cannot paste images directly. Upload or insert images from URL.

×
×
  • Create New...

Important Information

By using this site, you agree to our Terms of Use. We have placed cookies on your device to help make this website better. You can adjust your cookie settings, otherwise we'll assume you're okay to continue.