Does anyone have any 4000 series GPU's running fah?

[dfonda]

EDog put up some 4070 Ti #'s in the front page review. WHawk will be getting his 4090 running soon, and I may have some #'s in a week or so...

Does anyone have #'s they can post up? There isn't much firsthand Team 32 info yet.

 

[JLK03F150]

E-Dawg also did F@H runs on the 4090 & 4080 he reviewed.
4090


4080

 

[freeagent]

Holy feck that card is an absolute beast. Not gonna lie, I was curious.

 

[Holdolin]

Well, I have 2 4090's running, but being 2 of several I'm not sure how to pull data on them specifically. That being said I do run Linux, and through observation they bounce between 15-29M PPD depending on WU. I have a pair of 4070Ti's on the way and will be able to give data on those once they come in and get to work. If there is a way to pull data from specific cards/computers please let me know as I'd be happy to give as much data as I can. I guess I really should update my sig...

Here is a current snapshot of my fleet:

 

[dfonda]

Well, I have 2 4090's running, but being 2 of several I'm not sure how to pull data on them specifically. That being said I do run Linux, and through observation they bounce between 15-29M PPD depending on WU. I have a pair of 4070Ti's on the way and will be able to give data on those once they come in and get to work. If there is a way to pull data from specific cards/computers please let me know as I'd be happy to give as much data as I can. I guess I really should update my sig...

Here is a current snapshot of my fleet:

View attachment 360473

That is the data needed. PPD per GPU.

Earthdog seems to be the only one to have checked fah #'s for the more affordable(and getable) 4070 Ti. I ordered one to speculate. If the #'s hold up that one GPU will do as much as a 3080 and 3070 Ti.

Cannot find anything else on line on the 4070 TI...sure that will change by next week...plus stock will be gone. lol

 

[Holdolin]

Good deal. Was also a great time to snap a pic as it shows how the same card varies based on WU. Additionally I can verify the cards use 300-350 watts (agian, depending on WU). I tested the prometheus' board using on onboard graphics to get a baseline on power draw before the card got put in. She's now at 450-500 watts.

 

[HayesK]

HFM has "Benchmark Viewer" and "WU History" under the "Tools" menu.

The Benchmark Viewer provides data for each project by slot.

The WU History has data for all the wu and has an "export" feature under the file menu.
WU History Data View defaults to "Select All", but new views of the history can be created.
Select "New", then enter a name for "New Query" and a project ID in field "Value".
The fields "Name" and "Operator" default to "Project ID" and "Equal", but other values can be selected.

 

[WhitehawkEQ]

Hayes, you took the words right out of my keyboard I was going to say the same thing (more or less).

 

[Holdolin]

Well, now I'm educated. Thanks!!

 

[EarthDog]

Does anyone have #'s they can post up? There isn't much firsthand Team 32 info yet.

Pretty sure our 4090 and 4080 review also had F@H numbers...

I just shutdown the 4070ti so you can see what it has done (under my id) when left alone for a few days.

 

[HayesK]

RTX4070Ti ppd, Windows, Earthdog

 

[dfonda]

RTX4070Ti ppd, Windows, Earthdog
View attachment 360482

Nice, I only looked at the screenshot posted with the review.

 

[dfonda]

Ubuntu , 2 4070 Ti's at 30 Million.

 

[Holdolin]

Ubuntu , 2 4070 Ti's at 30 Million.

View attachment 360754

Noice. What kind of mobo you got under those? To be more specific, is it running PICI4@x8/x8? In all the years I've been doing this thing we do, bandwidth has always been part of the equation, but to what extent I've never been sure. All the new boards run x8/x8 with 2 cards, and to get muliti-gpus at x16 requires stepping up to Workstation/Server boards which get expensive quick.

 

[dfonda]

Noice. What kind of mobo you got under those? To be more specific, is it running PICI4@x8/x8? In all the years I've been doing this thing we do, bandwidth has always been part of the equation, but to what extent I've never been sure. All the new boards run x8/x8 with 2 cards, and to get muliti-gpus at x16 requires stepping up to Workstation/Server boards which get expensive quick.

It is the Gig Z390 , The #1 in sig.

I think both slots are PCIe 4.0.

Each generation of PCIe is twice as fast as its predecessor. While PCIe 3.0 had a data transfer rate of 8 gigatransfers per second, PCIe 4.0 transfers data at 16 GT/s, and PCIe 5.0 at 32 GT/s. (The bit rate is measured in gigatransfers to show the theoretical max speed before encoding—realized speeds may be slower.)

 

[WhitehawkEQ]

Ubuntu , 2 4070 Ti's at 30 Million.

View attachment 360754

You do know you can hit Prt Scr (Print Screen) and paste in paint and save to HDD/SSD?

 

[dfonda]

You do know you can hit Prt Scr (Print Screen) and paste in paint and save to HDD/SSD?

I would need to install the equivalent of paint (Gimp) and actually use the keyboard. Much faster to take a pic which goes straight to Google photo's in the cloud and pull down the pic later in my recliner using my Pixelbook...The only thing I use to read your "Pearl's of Wisdom"

I like putting a strain on your old eyeballs as well....

 

[Farwalker2u]

I have been folding with a MSI Gaming Trio 4090 for a little over 3 days now. It seems to produce between 20 to 24 million PPD. This is on a Windows 11 system with 32 GB RAM, and a 12700K CPU (that is not folding).
Maximum GPU wattage has been 311.7W.