Folding Community Board

[LazyDev]

30 minutes ago, tep said:

 

I did feel that Turing is still more efficient than Ampere in terms of PPD/W when I gave the 3080 it's first WU. I was seeing a power draw of about 370W. Have since undervolt it and I have getting an average of 6.2M OOD on an average power draw of 300W so that translate to about 20k PPD/W. This is an EVGA 3080 FTW3 Ultra.

 

The Strix 2080ti rev.A is getting 5.5M PPD on an average power draw of 250w.

What is the clock speed of your 3080? Mine has never hit above 5.5M PPD

[Juggy]

Finally got my grubby paws on an RTX 3080 so now have the following Folding...

 

EVGA RTX 3070 XC3 Black

EVGA RTX 3070 XC3 Ultra

EVGA RTX 3080 XC3 Ultra

 

Looks good for around 11-12 million PPD.

[tep]

8 hours ago, LazyDev said:

What is the clock speed of your 3080? Mine has never hit above 5.5M PPD

2000-2100mhz depending on the project due to power limit. The 14911 beta WU is giving me 12M PPD though on the 3080. 

 

 

 

[Bitter]

Man, I wish my PSU had data like that...Stupid RM1000X

[LazyDev]

1 hour ago, Bitter said:

Man, I wish my PSU had data like that...Stupid RM1000X

I wish my AX1600i would get detected in HWInfo or HWMonitor. 

[Kilrah]

1 hour ago, Bitter said:

Man, I wish my PSU had data like that...Stupid RM1000X

This is GPU, not PSU that he showed

[Bitter]

D'OH!

 

Well I wish my GPU showed cool power data like that. Stupid Vega 56, just kidding I love it.

[LazyDev]

After some faffing around, finally got my power supply to load data into HWinfo64. Been over a month without data, which is fuel for a geek

 

 

GPU #5 is my 3080.

[Gorgon]

22 minutes ago, LazyDev said:

After some faffing around, finally got my power supply to load data into HWinfo64. Been over a month without data, which is fuel for a geek

Do you have to load iCue to get the i2c data from the 1600i or does HWinfo64 just find it?

[LazyDev]

47 minutes ago, Gorgon said:

Do you have to load iCue to get the i2c data from the 1600i or does HWinfo64 just find it?

I uninstalled iCue, rebooted. Checked HWinfo to find nothing. I then installed iCue (As that's where the drivers are in for the Corsair Link), making sure that iCue does not start with the system, then checked after a second reboot (Windows things...). The AX1600i showed up, but I had to leave it for a few minutes as it was on a refreshing frenzy. All stable and does show correct data.

 

I've not been able to get it to work for over a month, possibly due to a conflict between iCue and HWinfo.

 

 

[Favebook]

1 minute ago, LazyDev said:

I've not been able to get it to work for over a month, possibly due to a conflict between iCue and HWinfo.

This is a big problem for me too. If I want to run HWinfo, I have to reboot my PC after using it if I want to use my iCue again. I lose almost all functionality from iCue when I run HWinfo. 

[LazyDev]

2 minutes ago, Favebook said:

This is a big problem for me too. If I want to run HWinfo, I have to reboot my PC after using it if I want to use my iCue again. I lose almost all functionality from iCue when I run HWinfo. 

Interestingly, I never had an issue with my HX1000i. It was always plug and play, without issue. The only change I did was the power supply, and I get all sorts of issues now. I wonder if the data from the AX1600i is different from the HX1000i, in terms of what data is on the link. The only differences between the two supplies, is increased power rating, and improved voltage ripple. They must have tweaked the link.

[LazyDev]

Project 13440 a new sprint? Nothing showing up on Lar_System's database it seems.

 

[jctappel67]

Made 500 Million! Woot!

 

Would've made it faster if @cbigfoot hadn't stolen my 2060 and left a worthless stack of cash in its place.

 

Confirmed footage of the heinous perpetrator:

[maartendc]

Ive just moved my home office to my frigid attic space.

 

I figured I will be folding on my gaming rig to keep me warm up here.

 

Do something useful with the power rather than just getting a space heater ....

[Juggy]

On 12/13/2020 at 3:11 PM, Juggy said:

Finally got my grubby paws on an RTX 3080 so now have the following Folding...

 

EVGA RTX 3070 XC3 Black

EVGA RTX 3070 XC3 Ultra

EVGA RTX 3080 XC3 Ultra

 

Looks good for around 11-12 million PPD.

Everything is up and running....

 

 

 

The 3080 is in my main rig

[TVwazhere]

I think 2.73M PPD is just now my new normal

I'm definitely not complaining, two 1070ti's at 60% power is awesome and 1M ppd over what I was doing before the CUDA and moonshot updates. Though, I'm feeling an itch to get some RTX cards just to keep up wit the new wave of folders, as a single card now is out performing me while consuming less power. 

[Bitter]

The supposed 3050 might be a great folding value, I need more specs on it!

[Collin_S]

@LAR_SystemsI like how you have separated the OS's on the GPU PPD DB. I now can confirm that I was NOT the person to do the 3.1 mil PPD record on the RTX 2070 Rev A, as that record was set on windows... Though I may have set that Linux record! (and could be the only person to have contributed so far). 

Also, the addition of GPU folding suggestions for the CPU PPD:PLP DB is kinda cool. Keep up the good work!

[Gorgon]

12 hours ago, TVwazhere said:

I think 2.73M PPD is just now my new normal

...

I'm definitely not complaining, two 1070ti's at 60% power is awesome and 1M ppd over what I was doing before the CUDA and moonshot updates. Though, I'm feeling an itch to get some RTX cards just to keep up wit the new wave of folders, as a single card now is out performing me while consuming less power. 

Nice. But don't you mean a -50% Power Limit? The Founders Edition and most EVGA GPUs list a TDP of 180W and -50 to +21% or +31% for the FTW2 so 90 to 202W or 236W depending on the model.

 

So 2,730,000 PPD / (2 x 108W) or 12.6 kPPD/W being the efficiency at the lower Power Level.

 

This is consistent with what I'm seeing with my EVGA 1070 ti at a 90W PL:

Which in my experience is about the best you can do with Pascal. Turing will get you about 22.4 kPPD/W for a RTX 2070s so a 47% uplift in efficiency:

and the jury is still out on Ampere as I haven't seen any long term averages at the minimum PL for 3070s yet.

 

In comparison my RTX 2060 and GTX 1660 ti both run about 14.8 kPPD/W at minimum Power Limits whereas the 2060 Super yields 19.8 and RTX 2070a 20.1 kPPD/W so while the 2060 may have been the value leader (PPD/$) the mid-range 2070 are more efficient.

 

Your results also confirm show that the latest CUDA cores might have eliminated the "Windows Tax" and Windows based systems are at are closer to parity with Linux. Now if they were as easy to just "set and forget" ...

[TVwazhere]

1 hour ago, Gorgon said:

Nice. But don't you mean a -50% Power Limit?

I think it's actually 60%, I dont believe afterburner allows me to go below 60%. I'll double check ITNOS when I get home though. 

 

*Edit* it does go down to 50%, but I have it set to 60%

 

I have the FTW2 model and a Zotac Mini (cute little card, does just as good as the big one above it)

[Kilrah]

Nice and steady, bad driver mess, then good driver that's apparently a little better since the nice and steady is a tad higher... or it's me not remembering whether I had set 80% or 85% power limit before the mess, 85% now... 

 

 

But that's aaaaaalllll getting messed up again tomorrow since I'm finally picking up my 3080 in the morning

[LazyDev]

My system is stuck with 13M PPD. Might go over 14M tonight, will have to see. Sprint #6 might be in effect, I do suspect. Getting good moonshot work units lately.

[Gorgon]

2 hours ago, LazyDev said:

My system is stuck with 13M PPD. Might go over 14M tonight, will have to see. Sprint #6 might be in effect, I do suspect. Getting good moonshot work units lately.

No, Sprint 6 has not been launched yet We're likely just getting a nice consistent batch of new COVID WUs

 

5 hours ago, TVwazhere said:

I think it's actually 60%, I dont believe afterburner allows me to go below 60%. I'll double check ITNOS when I get home though. 

 

*Edit* it does go down to 50%, but I have it set to 60%

Try 50% rather than 60 - your production will go down also but likely by much less than 10%

 

5 hours ago, TVwazhere said:

I have the FTW2 model and a Zotac Mini (cute little card, does just as good as the big one above it)

Yes, my conclusion is to buy the least expensive card for folding if running on low power limits as the primary difference with the more expensive cards is a better cooler design to allow it to boost higher at maximum power limits.

 

By limiting the power, however, you don't need the XC Ultra or FTW and can get by just fine with a "Black" speaking in EVGA terms. You just might have to add a tad more overclock to get the base clock a little

[TVwazhere]

28 minutes ago, Gorgon said:

Try 50% rather than 60 - your production will go down also but likely by much less than 10%

 

Yes, my conclusion is to buy the least expensive card for folding if running on low power limits as the primary difference with the more expensive cards is a better cooler design to  allow it to boost higher at maximum power limits.

 

By limiting the power, however, you don't need the XC Ultra or FTW and can get by just fine with a "Black" speaking in EVGA terms. You just might have to add a tad more overclock to get the base clock a little

Both are OC'd +150mhz on the core, bringing them somewhere in the 1900's for core clocks. Cant remember what the memory offset is either but it's at least double that.