More CUDA cores = more PPD...yes?

[ihrsetrdr]

More CUDA cores has always been(in my mind) the key to more PPD. With current GPUs and WUs does this still hold true?

I'm considering either a GeForce RTX 3080 Ti 12GB or an RTX 4070 12GB. The older 3080ti has almost twice the CUDA cores as the 4070....

However, checking on folding.lar.systems shows the RTX 4070 producing 6,607,009 PPD avg, and the RTX3080ti pumping out 6,702,455 PPD....not a lot of difference, for the increase in # of cores and price.

Thoughts?

I'm mainly going to fold, but also do some video editing, so the more power the better.

 

[dfonda]

What is your budget Tim? I think PPD per $. The 4070 Ti is the one. I'll post a link to the chart.

GPU Folding@Home PPD Overall Ranks Database

folding.lar.systems

 

[HayesK]

The ppd/watt is also going to be much better on the 4000 series cards. I currently set the power limit at 230w on my 3070Ti (290w TDP), 280w on the 3080Ti (350w TDP) and 260w on my 4070Ti (285w TDP). (My 4070Ti power limit still needs to be optimized). Thinking the 4070 (200w TDP) would probably fold OK at 140-160w power limit.

 

[Cvsi3]

I have both a 3080Ti, and a 4070Ti.
Ball park my 3080Ti folding would run at about 300W whereas the 4070Ti was only about 150W. The 4070Ti is faster, it averaged like 12M ppd, Is the same price as the 3080TI,
and also runs about 15 - 20C cooler then the 3080Ti.
The downside is that it is a much larger card. Well, the card itself is smaller but the cooler is much larger.

Both of my cards are PNY btw.

 

[ihrsetrdr]

What is your budget Tim? I think PPD per $. The 4070 Ti is the one. I'll post a link to the chart.

GPU Folding@Home PPD Overall Ranks Database

folding.lar.systems

If I pay with Paypal the wife won't be tipped off...

 

[superducky]

What is your budget Tim? I think PPD per $. The 4070 Ti is the one. I'll post a link to the chart.

GPU Folding@Home PPD Overall Ranks Database

folding.lar.systems

Agreed. 4070ti is where I settled on the "I need a card to increase my PPD, now, that will never render a frame in anger ever in its life--its just for compute" spectrum.

https://imgur.com/RlWd3Hn

https://imgur.com/uujBgq9

And, Hayes is right, too--the 40s are SO MUCH cooler.

 

[Farwalker2u]

My MSI 4090 Gaming Trio does not go over 350 watts while folding (usually under 300 watts), even with a small overclock of 2910 MHz GPU clock, and 11352 MHz memory.

 

[Holdolin]

My MSI 4090 Gaming Trio does not go over 350 watts while folding (usually under 300 watts), even with a small overclock of 2910 MHz GPU clock, and 11352 MHz memory.

I'm seeing pretty much the same on my Gigabyte boards. I haven't touched the clocks. i just let them do their thing.

 

[ihrsetrdr]

What would make this purchase a lot sweeter is if I actually played video games. I really even wouldn't know how or what, never gamed online....last games I played were PC based off-line games like Ghost Recon and Quake 3.

 

[Holdolin]

What would make this purchase a lot sweeter is if I actually played video games. I really even wouldn't know how or what, never gamed online....last games I played were PC based off-line games like Ghost Recon and Quake 3.

In my case that's an amazingly true statement as well. Even though I do play games, my gaming rig has one of my old 2080Tis. All my 40 series cards are dedicated folders that I go to some lengths to slim down as much as possible. I want absolutely as few as possible compute cycles spend doing anything aside from folding. That ideology is what keeps me from using any of my new cards for gaming.

 

[ihrsetrdr]

In my case that's an amazingly true statement as well. Even though I do play games, my gaming rig has one of my old 2080Tis. All my 40 series cards are dedicated folders that I go to some lengths to slim down as much as possible. I want absolutely as few as possible compute cycles spend doing anything aside from folding. That ideology is what keeps me from using any of my new cards for gaming.

...back then I was using GeForce4 Ti 4200s, but my main "gaming" effort was GPU overclocking and Mad Onion benckmarking. I don't recall exactly when GPU folding started, maybe 2006?

 

[Holdolin]

That's a good qeustion. I don't know when exactly folding on GPUs became a thing. I started in 2009 and was only aware of CPU folding thus built Supermicro 4P servers to crunch -bigadv units. I swear it seems like just yesterday at times.

 

[WhitehawkEQ]

I started in 12/8/2010 2 days after I joined the forums , with a single core CPU, my GForce 6200 GPU would not fold. Then in 2011 I got my 880GA-UD3H MB, GTS 450 GPU and started those to folding, then in 2013 or 14 I built my 1st 4P folder and later added a 2nd 4P that I still have today running Win 10 Pro for Workstation. (I'm not folding with the 4P's)

 

[ihrsetrdr]

Before the Intel Core 2 processors came out I was building a bunch of AMD 2P machines, never did go with a 4P although that was on my wish list.

 

[unsafesteagle]

I had a Tyan Thunder K8WE
If I remember right

 

[Holdolin]

Ya, you remember right. Or at least that board exists. I remember looking at one

 

[ihrsetrdr]

My sigline from 2006, all folding, 2P machines in Bold:

Abit NF7 2600xp * MSI K7D #1 dual mp2000+* MSI K7D #3 dual mp2200+ * Soltek 75frn2-RL-XP2700 * MSI KM4MV 1666mhz XP 2000 * SoltekK8M800I-RL /A64 3000 * Gigabyte 8kvm800 AMD64 3000 * MSI-K7D #2 Dual mp2800 Water-cooled * A7M266-D #1 2x Athlon xp-m 2400's @2133mhz *A7M266-D #2 2x Athlon xp-m 2500's @2266mhz * Gigabyte GA-7DPXDW 2x Athlon xp-m 2500's @1800mhz * Dell Precision Workstation 530 Dual Xeon Processors @1.7Ghz *Asus K8NDL + dual Opteron processors.

 

[ihrsetrdr]

So I still don't know what it is, besides CUDA cores that makes for more PPD.
Example:
GeForce RTX 3080 Ti has 10240 cores but makes 5,777,977 PPD. Price: $1,251.44.

GeForce RTX 4070 Ti has 7680 cores and makes 10,498,914 PPD. Price: $849.99

The GeForce RTX 3080 Ti has...33% more cores that the 4070 Ti but is making half the PPD.

 

[dfonda]

So I still don't know what it is, besides CUDA cores that makes for more PPD.
Example:
GeForce RTX 3080 Ti has 10240 cores but makes 5,777,977 PPD. Price: $1,251.44.

GeForce RTX 4070 Ti has 7680 cores and makes 10,498,914 PPD. Price: $849.99

The GeForce RTX 3080 Ti has...33% more cores that the 4070 Ti but is making half the PPD.

Stop procrastinating ....see post #'s 2,3 4 and 6....and get a 4070 Ti.

 

[HayesK]

@ihrsetrdr

4070Ti vs 3080Ti ppd. Suspect most of the difference is related to the GPU core frequency.

When there are sufficient cores to load the entire wu, the extra cores are idle.

The 4070Ti boost frequency is more than 50% higher than the 3080Ti.

Most of the ppd comes from the Quick Return Bonus, which is not linear.


Just checked a few of my GPUs (values will vary depend on the project)


3070Ti (MSI Suprim) folding at 2025 MHz, 68C, 230w (power limit set 230w)

3080Ti (EVGA FTW3) folding at 1845 Mhz, 71C, 280w (power limit set 280w)

4070Ti (MSI Suprim) folding at 2955 MHz, 64C, 218w (powerlimit set 235w)