LTT Official Folding Month 2020!!!

[cbigfoot]

6 minutes ago, marknd59 said:

My output is maxed out so I expect to slowly slid down.

izac is even joining in in a small mannor got his issue fixed turns out SERVER GRADE MBs dont like Gaming Memory...go figure

 

threw in 4x8GB 3200MHz crucial CL22 and been rock solid for 2 weeks and actually can reboot with out crashing

[Guest]

22 minutes ago, marknd59 said:

My output is maxed out so I expect to slowly slid down.

[marknd59]

5 minutes ago, Den-Fi said:

 

[Guest]

3 minutes ago, marknd59 said:

 

[Collin_S]

36 minutes ago, marknd59 said:

My output is maxed out so I expect to slowly slid down.

I am too, but I am tuning my system still. One thing that is annoying is I can't just skip online classes, and even though I keep F@H running, they really cut down on my output. Though I know If I didn't need to be on my system all day, I could probably get a few hundred thousand more points. I am working on getting some of my friends into it and seeing if I can get them to use my passkey for this month.

[marknd59]

The problem is I'd need about 17 P100 to get over your output for a day. I'll stick to my 2 x 2060 and 2 x 2060 super and end up where I end up

[Collin_S]

Is there a way to track the PPD of a specific client?

[cbigfoot]

8 minutes ago, Collin_S said:

Is there a way to track the PPD of a specific client?

i use https://github.com/harlam357/hfm-net to monitor them all.  and can look up what they done etc

but there are ALOT more details you can look up some where there is a howto on how to set it up

also (For those that stalk my signature Scooty Puff JR has moved on to greener pastures...) 

[Shlouski]

28 minutes ago, marknd59 said:

The problem is I'd need about 17 P100 to get over your output for a day. I'll stick to my 2 x 2060 and 2 x 2060 super and end up where I end up

Wow, you love your 2060's

 

An rx 470 and 570 are closest I have to having the same card. 

 

How many ppd on those cards?

 

I have my first 2060 coming, it's the evga ko, I imagine it's a little better than a reference 2060, but slower than a super.

[Collin_S]

1 minute ago, Shlouski said:

Wow, you love your 2060's

 

An rx 470 and 570 are closest I have to having the same card. 

 

How many ppd on those cards?

 

I have my first 2060 coming, it's the evga ko, I imagine it's a little better than a reference 2060, but slower than a super.

Yeah 2060 Supers are a ton more like a 2070 (what I have) than a 2060. The 2060KO is the same amount of CUDA cores as a normal 2060, but has a cut down version of the dye of the 2080. It has the same performance a normal 2060 for gaming, BUT it has up to 47% in some workstation performance (different per application). I wouldn't be surprised if Folding At Home ends up being one of those applications that does get a bit of an uplift from it.

 

 

[Shlouski]

34 minutes ago, Collin_S said:

I wouldn't be surprised if Folding At Home ends up being one of those applications that does get a bit of an uplift from it.

 

I did know about it using a different chip and being quite a bit better in some tasks, also it's the cheapest 2060 I could buy.

 

Unfortunately I don't think fah will get a boost, fah likes cuda cores and the 2060 ko has the same number of them as a reference 2060, so I guess it will perform like a reference 2060, if it running faster memory speeds then it might be a little faster.

 

Edit: it's running similar clock speeds.

[Collin_S]

Just now, Shlouski said:

 

I did know about it using a different chip and being quite a bit better in some tasks, also it's the cheapest 2060 I could buy.

 

Unfortunately I don't think fah will get a boost, fah likes cuda cores and the 2060 ko has the same number of them as a reference 2060, so I guess it will perform like a reference 2060, if it running faster memory speeds then it might be a little faster.

It has the same number of everything on the 2060 KO. Applications using CUDA (as in the compute method, not just using the cores for rendering games) (Including Blender) are the ones that get the boost. Blender gets a boost of 26% for CUDA, but 19% for Optix (which uses a combination of CUDA and some of the dedicated ray tracing hardware). To my understanding, it is something with the engines of the card that makes the difference, not so much the actual cores.

[Collin_S]

2 hours ago, GOTSpectrum said:

It seems to be something that has come up since the CUDA update, as before you could just leave a core per GPU and it would be fine 

I seem to still be doing better with CPU folding on, but I have a 3900x, and I am leaving at least 2 threads open for the GPU. And my CPU currently has a TPF of 1min 6 seconds, with my GPU at 59 Seconds TPF, so I am a bit of an outlier (no CPU OC though).

[Shlouski]

9 minutes ago, Collin_S said:

Applications using CUDA (as in the compute method, not just using the cores for rendering games) (Including Blender) are the ones that get the boost. Blender gets a boost of 26% for CUDA, but 19% for Optix (which uses a combination of CUDA and some of the dedicated ray tracing hardware)

 

It uses the same number of cuda, tensor and rt cores, so I dunno why it's better, I guess it's possible it's using newer cuda cores. 

[Collin_S]

Just now, Shlouski said:

 

It uses the same number of cuda, tensor and rt cores, so I dunno why it's better, I guess it's possible it's using newer cuda cores. 

It shouldn't be that, especially since the 2080 actually came out before the 2060. EVGA doesn't even know the cause of it, and didn't even know it happened until after Gamers Nexus asked them about it... hahaha

[Shlouski]

 

13 minutes ago, Collin_S said:

It shouldn't be that, especially since the 2080 actually came out before the 2060. EVGA doesn't even know the cause of it, and didn't even know it happened until after Gamers Nexus asked them about it... hahaha

Yeah that's true, dunno what I was thinking...........it is really late though

 

Seems the 2060 and 2060 ko have a different configuration of streaming multiprocessors, and the larger chip draws a little more power, these are the only differences I can find.

[Collin_S]

1 minute ago, Shlouski said:

the larger chip draws a little more power

Well, yes and no. It actually doesn't have the ability to raise the power limit on it like the normal 2060's or the partner's OC'D versions especially.

[Collin_S]

1 hour ago, cbigfoot said:

i use https://github.com/harlam357/hfm-net to monitor them all.  and can look up what they done etc

but there are ALOT more details you can look up some where there is a howto on how to set it up

also (For those that stalk my signature Scooty Puff JR has moved on to greener pastures...) 

I got it working for my laptop, and the slots show up for my desktop. but it says both are at 0% and shows no other details. Though what I was looking for was the ability to track the overall points made by an individual client, like in a history.

[Collin_S]

For any of you folders out there with over a 16 core CPU and on Linux, Big Advanced WU's are definitely worth considering... https://foldingathome.org/support/faq/installation-guides/configuration-guide/

[Intrinsic]

How often are these LTT folding events held?

[Gegger]

2 minutes ago, Intrinsic said:

How often are these LTT folding events held?

AFAIK, there is folding month every year usually in october, and a summer(for the northern hemisphere) week-long folding sprint earlier in the year

 

earlier this year there was a COVID-19 related folding event, with many many people joining

[DailyProcrastinator]

For an all out folding rig, what specs should I look at?

 

For the CPU do high clocks speeds benefit greatly? Or would going for higher core count provide better performance?

I may have a line on a used TR 2970WX system. Otherwise I would consider a 3950x.

 

For the GPU does a high cuda-core count provide a big benefit? Looking at a used GTX 1080ti or RX 5700xt. 

 

Thanks in advanced!

[DirectingGnu2]

2 minutes ago, TheDailyProcrastinator said:

For an all out folding rig, what specs should I look at?

 

For the CPU do high clocks speeds benefit greatly? Or would going for higher core count provide better performance?

I may have a line on a used TR 2970WX system. Otherwise I would consider a 3950x.

 

For the GPU does a high cuda-core count provide a big benefit? Looking at a used GTX 1080ti or RX 5700xt. 

 

Thanks in advanced!

Uh don't bother with high end cpu, its all in the gpus. Nowadays cuda cores matter now, I personally wouldn't bother with AMD until Big navi comes out.

[Spotty]

39 minutes ago, TheDailyProcrastinator said:

For an all out folding rig, what specs should I look at?

 

For the CPU do high clocks speeds benefit greatly? Or would going for higher core count provide better performance?

I may have a line on a used TR 2970WX system. Otherwise I would consider a 3950x.

 

For the GPU does a high cuda-core count provide a big benefit? Looking at a used GTX 1080ti or RX 5700xt. 

 

Thanks in advanced!

CPU isn't worth it for folding, though it will help if you're looking at doing BOINC as well. 

The 1080ti outperforms the 5700xt. 

 

https://folding.lar.systems/folding_data/gpu_ppd_overall

Don't just look at raw performance though also check PPD/W of the cards if you're planning on running full time.

[marknd59]

5 hours ago, Shlouski said:

Wow, you love your 2060's

 

An rx 470 and 570 are closest I have to having the same card. 

 

How many ppd on those cards?

 

I have my first 2060 coming, it's the evga ko, I imagine it's a little better than a reference 2060, but slower than a super.

The first folding system I put togeather had a budget of around £800 for GPUs and at the time it worked out that 2x 2060 was the best way to go. Next when it came time to update my daily driver that had a RX 580 and a RX 590 in it the best PPD/KwH was the 2060 super and it may still be we don't have enough data on the 3xxx and the CUDA processing yet to run the numbers.
 

https://docs.google.com/spreadsheets/d/1iauimDyaUS0jnpeYqRNbM2uub-ASvS1OEz2yEUNJgls/edit#gid=0