TPU's F@H Team

[Toothless]

Yeah but this one uses ddr4 ECC, and another Dell Mobo wouldn't help me as it would have the same issue. The only salvageable thing would be the CPU but a 2667 V3 is like 30 cad with shipping included from US. Case is proprietary and so is the PSU. So, not much you can reuse. Most x99 consumer boards on the market do not support ECC.

I used ECC in a Gigabyte x99 with my 2698v4 before going dual socket. Supermicro has fantastic ATX boards for that generation too.

Also has a dual socket ATX. Been my favorite server board so far.

Edit: Supermicro X10SRA looks like a good fit.

 

[dogwitch]

I used ECC in a Gigabyte x99 with my 2698v4 before going dual socket. Supermicro has fantastic ATX boards for that generation too.

Also has a dual socket ATX. Been my favorite server board so far.

Edit: Supermicro X10SRA looks like a good fit.

if your wondering. that whay i pick what i did.
i do some gaming.
but its more folding,boinc,vm,plex etc stuff on my pc.
i also notice how limited gaming mobo are with pci lanes.
they do out right lie to . wendell lvl1tech found that out this year i think on mobo he has.

 

[Marstg]

I saw an Asus Z87-WS combo with 4770k CPU for like 220 cad. Specs say quad x8, that would hold my 3 gpus plus one nvme onto an adapter, once I update the bios to support the nvme or I could do a raid5 on the onboard intel controller with 3 ssds. But most probably single nvme with dram cache will be my choice.
@dogwitch please link that article.

 

[Toothless]

if your wondering. that whay i pick what i did.
i do some gaming.
but its more folding,boinc,vm,plex etc stuff on my pc.
i also notice how limited gaming mobo are with pci lanes.
they do out right lie to . wendell lvl1tech found that out this year i think on mobo he has.

CPU has 40 lanes, more than enough for even a quad GPU config with everything running 8x. Board issue most likely. I had four GPUs on a 1680v2 and x79 Classified and everything ran fine.

I saw an Asus Z87-WS combo with 4770k CPU for like 220 cad. Specs say quad x8, that would hold my 3 gpus plus one nvme onto an adapter, once I update the bios to support the nvme or I could do a raid5 on the onboard intel controller with 3 ssds. But most probably single nvme with dram cache will be my choice.
@dogwitch please link that article.

Depends on the board. Haswell consumer chips are 16 lanes and I'd run 8x to GPU and 4x to NVME PCI card. Worked fine with the 1080ti (at the time) and whatever NVME I used then. SLI was a clear 8x/8x so yeah, differs between boards.

 

[Marstg]

This board must have a PLX pcie splitter/multiplexor, I guess it takes all 16 lanes from CPU creates 32 then splits in a quad x8 configuration, have you seen what a huge heatsink it has? It's my only explanation to it's capability. The NF200 chip is only a pcie 2.0 unit.

 

[phill]

This board must have a PLX pcie splitter/multiplexor, I guess it takes all 16 lanes from CPU creates 32 then splits in a quad x8 configuration, have you seen what a huge heatsink it has? It's my only explanation to it's capability. The NF200 chip is only a pcie 2.0 unit.

Are you trying to get a number of GPUs in a single system to fold with? What's the end goal here? Maybe I missed a previous post somewhere....

 

[thebluebumblebee]

Has F@H PPD gone up? This is just my 3GB GTX-1060 averaging what I thought was 1070 territory. Average close to 770,000 PPD.

 

[80-watt Hamster]

View attachment 326501

Has F@H PPD gone up? This is just my 3GB GTX-1060 averaging what I thought was 1070 territory. Average close to 770,000 PPD.

That's notably better than the 1060-6G I used to have, which averaged mid-600K. It's entirely likely you're getting assigned high-value WUs.

 

[phill]

Is there a way to get a report from FAH to see what each machine/GPU is averaging or doing so to speak? I'm just curious I've not seen anything so to speak, in your face with the information....

 

[thebluebumblebee]

Is there a way to get a report from FAH to see what each machine/GPU is averaging or doing so to speak? I'm just curious I've not seen anything so to speak, in your face with the information....

No.

 

[phill]

No.

Its a shame but not a problem Will just need to do more

 

[thebluebumblebee]

Wahoo! 100,000,000 in a month, with one day to spare..

 

[Athlonite]

Is there a way to get a report from FAH to see what each machine/GPU is averaging or doing so to speak? I'm just curious I've not seen anything so to speak, in your face with the information....

That would be a Yes and No the only way you could do it is to use a seperate setup and account with FAH but then that would defeat the purpose of using a bunch of machines to gain more points quicker on a single account

 

[thebluebumblebee]

Is there a way to get a report from FAH to see what each machine/GPU is averaging or doing so to speak? I'm just curious I've not seen anything so to speak, in your face with the information....

No.

Eh, now that I think about this a second time, F@H can't give you this, but HFM.net might.

GitHub - harlam357/hfm-net: Client Monitoring Application for Folding@Home

Client Monitoring Application for Folding@Home. Contribute to harlam357/hfm-net development by creating an account on GitHub.

github.com

This is from 2010:

HFM.NET - Client Monitoring Application for Folding@Home

Hello Tech Power Up! I'm posting today to share with you my Folding@Home Monitoring Application - HFM.NET. It has a wealth of features and is continually under development. Below you will find a link to the Google Code page where you can download HFM.NET - IT'S FREE!!! I'll endeavor to...

www.techpowerup.com

 

[thebluebumblebee]

FYI: GTX 1660 Super production (6/12-6/19):

 

[80-watt Hamster]

FYI: GTX 1660 Super production (6/12-6/19):
View attachment 352928

That echoes what I was seeing on my Titan XP before I shut it down for the summer.

 

[BUCK NASTY]

It's been a while guy's and it's "Hot as Hell" here in Florida. I've picked the worst time of the year to build a large scale PPD folding rig, but here it goes. I've got tons of hardware (mostly 10-15 yrs old) and I'm looking to put it to use. Luckily Mobo/CPU/PSU tech has not advanced as drastically as GPU's have. Looking to pick up 2x RTX 4080(AD103-300) and fold only on the GPU's. Also been looking at the RTX 4080 Super, but their efficiency and PPD is lower per 1m PPD.

Mobo - Asus P7P55 WS Supercomputer
CPU - Intel Core i7-870
PSU - EVGA 1000 Supernova Gold
OS - Ubuntu 24.04 LTS

Looking for some feedback on which GPU to choose, as the 4080 Super has alot less samples worth of Data(see links below).

GeForce RTX 4080 Folding@Home PPD Averages, Power Consumption & Research Projects

F@H GeForce RTX 4080 performance as of 12/8/2024. Averages across all projects PPD:16,772,016 - Work Units Per Day:24.56, performance history and more.

folding.lar.systems

GeForce RTX 4080 SUPER Folding@Home PPD Averages, Power Consumption & Research Projects

F@H GeForce RTX 4080 SUPER performance as of 12/8/2024. Averages across all projects PPD:18,277,921 - Work Units Per Day:31.83, performance history and more.

folding.lar.systems

 

[thebluebumblebee]

Howdy Buck!
That motherboard only supports PCIe 2.0. Not sure if that will have any effect.

 

[Super Firm Tofu]

Howdy Buck!
That motherboard only supports PCIe 2.0. Not sure if that will have any effect.

That was my concern as well. I don't know the other specs, but if you use two slots, you might be limited to PCIe 2.0 x8 per card.

Might consider a modern alternative that can give you two PCIe 4.0/5.0 x8 slots.

 

[phill]

I'm not sure that the PCIe 2 will be a massive problem, bit like when mining, but it'll be interesting to see the results!!

Welcome back @BUCK NASTY !!!! It's an honour!!

 

[BUCK NASTY]

@thebluebumblebee I don't remember F@H ever being limited by PCI-e revisions. I'll have to confirm this. I have a MSI MPG X570 GAMING EDGE WIFI with a 5600x as my daily driver. Could use that if needed.

 

[thebluebumblebee]

I don't remember F@H ever being limited by PCI-e revisions.

I'm not sure either, but I am concerned about that P55's age and whether it has UEFI or not. It might not work with UEFI GPU's.

As for 4080 vs 4080S, I would look at those number on https://folding.lar.systems (great effort and info) with a grain if salt. That site shows the 1060 3GB beating the 1060 6GB. The numbers shown there are really low for the 1660S in comparison to the numbers i posted above. W1zzard's testing of the the 4080/4080S showed they use the same power, but the 4080S has 512 MOAR (~5%+) shaders. To me, everything, including the price, points to the 4080S. I just can't see how https://folding.lar.systems shows the 4080 as 20%+ faster than the 4080S???

 

[80-watt Hamster]

I'm not sure either, but I am concerned about that P55's age and whether it has UEHI or not. It might not work with UEFI GPU's.

As for 4080 vs 4080S, I would look at those number on https://folding.lar.systems (great effort and info) with a grain if salt. That site shows the 1060 3GB beating the 1060 6GB. The numbers shown there are really low for the 1660S in comparison to the numbers i posted above. W1zzard's testing of the the 4080/4080S showed they use the same power, but the 4080S has 512 MOAR (~5%+) shaders. To me, everything, including the price, points to the 4080S. I just can't see how https://folding.lar.systems shows the 4080 as 20%+ faster than the 4080S???

Sampling limitations can skew LAR results. A high- or low-point run on just a few cards can push the numbers to an appreciable degree. The default display is also filtered for most recent, so sometimes switching to all-time can give a better picture. But since projects and point values shift over time, that's not foolproof either.

So yeah, keep that saltshaker handy.

 

[thebluebumblebee]

Hey @BUCK NASTY :

[FS][US] - MSI 4080 Super Gaming X Slim

I thought I was having issues with my 4090 that I have had for a couple of years. So, I ordered this GPU at the end of July. I ran it in my system for two weeks and figured out the issue I was having with my 4090. As the 4090 is good to go now, I no longer need this GPU. My loss is your...

www.techpowerup.com

 

[thebluebumblebee]

Isn't this lovely?

Study: AI and data centers could drive cost of energy up by 70% over 10 years

Prices jumped from $29 to $270 per megawatt-day “across the PJM region” and from $29 to $444 in parts of Virginia, home to more than half of the nation’s data centers, according to the study.

justthenews.com