My ATI vs Nvidia Folding Battle

[Psywar]

So I do not do much on my PC other than moderating the forums and catching up on the latest news on the internet.

 

So I try and always keep folding@home running with 100% uptime.

 

I built this computer last summer and ever since I got it up and running I have had folding installed on the machine.

My ATI Card: SAPPHIRE Radeon HD 4870 512MB

Did not seem like it was cutting it compared to some of the other guys PPD's (Packets per Day) they were pushing out.

It seemed as if my ATI Card was only giving me 2000 to 2700 PPD's.

 

I thought this was a bit low compared to others so I was thinking I wonder if the Nvidia GPU would give me a better PPD then the ATI Card.

So this last Wed. I got my hands on a Nvidia GTX260 and slapped it in to my machine.

It has been running 24hours so far and I have already completed 14 WU's with it.

 

Now these two cards are not exact matches, but they are closely matched up in specs and performance. Actually to be perfectly honest my ATI Card usually whips the GTX260 in most of its Benchmarks.

So Honestly in theory I think the ATI Card is superior and if I was a gamer I would chose it over the GTX260.

 

But when it comes to Folding the GTX260 is almost/if not doing three times as many PPD's as my ATI card was doing.

The GTX260 is giving me between 6000 and 6500 PPD's - The Screen shot I am going to post has me at 7000 PPD but I think this is a small Protein because I actually just received this WU only an hour ago or so.

 

But anyways I took two screen shots of HFM with my ATI Card and one with my GTX260.

 

Here is the ATI Card:

 

 

 

Here is the Nvidia Card:

 

 

 

I just thought I would share with you guys anyways, I found it super interesting.

 

The other Folding Client HFM is showing is my computer. I only have a Dual Core CPU so I only run that Single Core Client.

I tried using the SMP on this machine with my GPU and I ended up getting less PPD's off my CPU then I did running the Single Client.

[Locutus]

It is a well known fact that nVidia cards fold a lot better than ATI cards. This is because the tools at stanford who are in charge of programming don't really care about the huge gap in scores.

[ivangela]

Hey.

I know this isn't quite on topic but I noticed that in one screenshot in the "MHz" column there is a frequency plugged in and in the other screenshot there is no frequency, actually it says 1.

When setting up the client, the only 2 things I set up was the pathway for my Folding client and I was obligated to give the client a name.

Should I put in the frequency of my CPU? It's a quad core PhII (see sig).

Thanks for the help, I just want true PPD numbers.

[Psywar]

Hey.

I know this isn't quite on topic but I noticed that in one screenshot in the "MHz" column there is a frequency plugged in and in the other screenshot there is no frequency, actually it says 1.

When setting up the client, the only 2 things I set up was the pathway for my Folding client and I was obligated to give the client a name.

Should I put in the frequency of my CPU? It's a quad core PhII (see sig).

Thanks for the help, I just want true PPD numbers.

 

Don't quote me on this, but it does not matter what you set the MHz at. It will still run the same.

In the one I had it at 2900mhz I think and when I reset the clients up with the other GPU I just set it at 1Mhz cuz it needed a number.

 

But my PPD's always turn out the same..

 

The PPD's come from the FAHLog.txt so the Client name, MHz n such have nothing to do with it.

 

(Or at least I think as far as I know.)

[IVIYTH0S]

It is a well known fact that nVidia cards fold a lot better than ATI cards. This is because the tools at stanford who are in charge of programming don't really care about the huge gap in scores.

So true, and yet so sad . They were supposed to optimize it for ATI again but never really followed through in that.

[ivangela]

Don't quote me on this, but it does not matter what you set the MHz at. It will still run the same.

In the one I had it at 2900mhz I think and when I reset the clients up with the other GPU I just set it at 1Mhz cuz it needed a number.

 

But my PPD's always turn out the same..

 

The PPD's come from the FAHLog.txt so the Client name, MHz n such have nothing to do with it.

 

(Or at least I think as far as I know.)

 

So it seems. The only things that changed when I put in my appropriate chip speed were the MHz & PPD/MHz columns. The PPD column didn't change...7071.9 currently

[beeiilll]

My 260 gives me 7200 to 7700 PPD pretty consistently at stock. One of these days I have to see how far it can be pushed and what that will do to the PPD output.

 

Boy am I glad I didn't get suckered into that goofy ATI nonsense.

[OwinC]

My GTX 470 gets around 14,000 to 16,000 PPD GPU3 on the i7 machine. Hitting 90c 100% fan.

 

The GTX 260 gets around 6000 to 7000 PPD GPU2 on the Pentium 4 machine. Hitting 82c 100% fan.

[engstudent]

 

Looks like my folding won't contribute much for a while had no idea I'd find something where ATI sucked this bad vs the green team.

[paulktreg]

 

Looks like my folding won't contribute much for a while had no idea I'd find something where ATI sucked this bad vs the green team.

 

Your GTS 250 would be the better folder? Probably 5500~6000 PPD. I wonder if there's anyway to force that one to do the folding in your rig, kinda complicates things with the 5850 installed?

[IVIYTH0S]

Your GTS 250 would be the better folder? Probably 5500~6000 PPD. I wonder if there's anyway to force that one to do the folding in your rig, kinda complicates things with the 5850 installed?

No I seem to remember folding with both my 4870x2 and my physx 8800GT (All three GPU's) and it worked.

Btw to ati folders, it helps a little to leave a processor core free for each ati gpu folding, still no where near the nvidia output but the gpu will fold better.

[Hyper Threaded]

I'm quite disappointed really, with the folding performance of the ATI client. On paper, the HD4870 is better than the GTX260, but the client lets it down. For shame, Stanford. When will they ever catch up with the clients?