X1800XT up to 7X faster than 7800GTX

UnG > I generally agree, and it was not my intention to display severe symptoms of either fanatism or nvidiotism... However, one has to remember that Epic really is a TWIMTBP company, so I wouldn't be too sure. Moreover, I'm not sure what other threads you're referring to, because it's still very much unclear in those in the News section (smells big coverup from ATI, although it isn't an absolute requirement for SM3.0 - I give you that one). However, my reply was an answer to Cooper's intriguing suggestions in his posts...

Quote:

---

Originally Posted by Mikesta

Athlon 64's stack up badly against P4's in F@H. This is because F@H uses Intel compiler that does not allow SSE, SSE2 & 3 to be used with non-intel processors.

---

Just force SSE on at the command line. That's close enough for me. Darned if I will go out and buy Intel chips to fold with specifically, I'll use whatever is lying around. Be nice to have the video cards (even lesser cards.... 6800's, 9800's etc) be able to fold when not in use.

No worries Slider, it's just that in this particular case of folding with the chip I don't really care which card is the better gamer :) . The two companies use their pawns to hype up issues on forums like ours and get the guys fighting amongst themselves and it depresses me a little.

To no one in particular: If you are being funded by ATI or Nvidia directly or indirectly, you should step up and say so everytime you post...ya hosers. Get some ethics.

I'm going to get some time on the stanford DC forums in and learn a little more about the folding on video card way of the world. It seems the most promising..

GO TEAM 36362 :toast:

all this means to me, is I just need to put that card in my folding rigs but I can keep my Game box pure...

Cool stuff.

Folding Power!

Shadow, pls stay on topic, this thread is about X1800XT's potential folding capabilities.

Perkam

So, what makes for these excellent folding capabilities of the R520? Must be some extreme efficiency, considering less pipes and just 50% higher clocks. Possibly also the superior mem controller with 512-bit internal capabilities as well as the extra memory? When will GFX cards start help out with crunching? (Was never that into folding anyway.)

UnG > Yes, it's probably not too far-fetched to assume both nvidia and ati reps have a finger or two in all of this mess. However, they wouldn't probably know any more about the who is who than the next guy is, so I say, just let them fight it out while everyone else are enjoying their brand new XGI card. :D
And that other thread seems quite closed btw...

Fairly OT, but...
Shadowmage > Dil maintains that the feature's used for every texture in every frame in U3, and (assuming he hasn't been hallucinating too much lately) I wouldn't underestimate its impact on IQ and performance. But, as for now, we can only speculate, so I'm not sure this is a cause worth spilling too much blood for.

edit: well... seems like the black panther's been out hunting again. guess they don't want us discussing this issue, shadow. (*cough*coverup*cough*)

You mean X1800XT, right? ;)

Quote:

---

Originally Posted by perkam

Yea, but Nvidia's GPU was never designed for this type of folding and I'm sure there are other types of calculations it can do faster.

Exhibit A: 7800GTX doesnt lag behind as much in every test

http://img192.imageshack.us/img192/4...0sgemv14lx.jpg


Exhibit B: This has more to do with ATI's Architecture and less to do with whether the X1800 is faster than the 7800 (even the x800xtpe ends up faster in these calculations)

http://img192.imageshack.us/img192/2...0slides9su.jpg


Exhibit C: In ATI's defense, the X1800 could show even more performance seeing as how it isnt optimized for the X1800's new "scatter" feature

http://img192.imageshack.us/img192/3...ptimize3py.jpg

I just wish this defeats the need for a ppu in the future if companies continue to keep making hardware with 3 to 4 times the processing power of today's cpus.

Perkam

---

Yes that was a typo rcofell, I meant X1800. But If you look at my post above rcofell, even the x800xt wins over the 7800GTX sometimes...so its a matter of architecture.

Perkam

A little birdy told me those numbers had been understated, the actual optimized results will significantly higher....

Fold baby fold..... daddy needs soem new PC's :)

Quote:

---

Originally Posted by Ugly n Grey

Just force SSE on at the command line. That's close enough for me. Darned if I will go out and buy Intel chips to fold with specifically, I'll use whatever is lying around. Be nice to have the video cards (even lesser cards.... 6800's, 9800's etc) be able to fold when not in use.

---

You still can't get a64's to run with SSE2 on cores like QMD which are the most efficient cores to get points/time wise. Even on 364 pointers (The best PPW cores for a64's) a64's get outgunned by Intel CPU's. It sorta makes me sick. A64's would kick some major ass if they would use SSE2.

Anyway back to the topic at hand....ATI will have my money when I decide to upgrade my 6600GT because of the fact that it Folds so much better. The question I guess is...why is it so much better then Nvidia products?

Oh and on the clearspeed topic. I will be getting one for sure. What is great about it is that it's 50Gflops of processing power @ 25W. A big thing with hobby folders is the electricity bill. The Clearspeed math coprocessor has around 4 times +, the power of a good P4's which would consumer over 400W+ of power.

I just hope that the Clearspeed math co-processor will be supported or be able to be used with programs for other uses like encoding.

But back to the R520, around 25Gflops of power, unoptimised in Folding Gromac's is just great. I am really looking forward to the r580 hoping it will have a 512bit bus. 50+Gflops of folding power would be the equal of the clearspeed at around the same cost but have other uses (Crossfire!). Unfortunately they will probably consume 80-100watts of power.

Yeah, I was thinking the same thing. I don't play games but I'd buy one to fold on. As far as the SSE2 fiasco and the QMD packets, I have some of both camps products (A64/P4), so like I said, that good enough for me (at least for now). It won't be long before there's a 64 bit edition as well which should help both camps.

Well the next gen cards alone should match the Clearspeed dual chip cards if they can get the code optimized. I think the clearspeed is an easier product to code for, but I won't buy one. It's a limited use product to me. At least on a video card I have a display function :D which I DO need.

Heres two threads on the hacked intel compiler(s).

Cant find the 3rd one, think it was started by Lithan but I could be wrong.. let us know if you figure out how to use this with F@H and how it impacts perf. ;)

http://www.xtremesystems.org/forums/...hlight=compile
http://www.xtremesystems.org/forums/...hlight=compile

hmmmmm im not too likely to try it.

Like Ugly n Grey , I am waiting for the 64bit patch. Apparantely it will play nice with A64's.

Anyway I hope Stanford gets off it ass and releases GPU client!

Even more so I hope that ATI gets off it's arse and gets its r520 or even r580 out as soon as they can!!!

If some1 wants to try patching folding@home for amd 64 here is the iccout program to do it http://www.xtremesystems.org/forums/...chmentid=29223
originally posted by Walrusbonzo

If I had an athlon 64 here I would try it

Also what you need to do is recompile the program using the -xN flag first, then use this program to patch it, otherwise it probably wont work.

Quote:

---

Originally Posted by perkam

Yea, but Nvidia's GPU was never designed for this type of folding

---

but atis vpu was?

Quote:

---

Originally Posted by saaya

but atis vpu was?

---

Well one of the architects talked about using their vpus as physics proccessors as well, so i guess they had it in mind.

Quote:

---

Originally Posted by Ugly n Grey

Distributed computing programs are nothing new and neither are discussions about using the VPU of your video card to fold. A recently leaked document from the Stanford University Graphics Lab has the X1800XT leading it's rival the GTX by over 7x in certain types of calculations with performance equivelant to a 10.5 GHz P4 CPU.

---

I do a fair amount of work in similar areas to this group at Stanford. I can tell you with near certainty that I can find a computation kernel that has a 7800GTX beating a Radeon X850 by 7x. The programming model, performance bottlenecks, and data-access and flow restrictions are so bizarre, and variable from GPU to GPU that disparities like this are really not particularly unusual.

As far as folding goes, I think that GPUs are pretty well suited to folding. In fact I know a guy building a custom GPU-based cluster to do private folding. Kind of neat. Anyway. I dont think the stanford folding guys open source the folding algorithms, so someone with access will have to rewrite the computational kernels to take advantage of GPUs. Fortunately, Stanford also has many of the pioneers of GPGPU.

What if you get 2 and crossfire them, and then run a CPU based folding client as well?

Would that be pure folding ownage?

Quote:

---

Originally Posted by Le_Petit_Lapin

What if you get 2 and crossfire them, and then run a CPU based folding client as well?

Would that be pure folding ownage?

---

ABSOLUTELY!

But watch the power bill....

Then again a system with 2x 512mb X1800 will do the work of many systems (folding boxes) and will be much more power efficient points wise.

agreed on the power use, multi core, multi card systems are more efficient that way. I think this is going to be fun, folding on cards especially at a place like xtremesystems where everyone and their uncle owns something fast!

Now think of the XBOX 360 and folding with an even faster core than R520 and what is it 3 or 4 3.2 ghz g5's?

Dragon

Since the X1800XT isn't available yet that wouldn't make it faster at anything....

Quote:

---

Originally Posted by freecableguy

Since the X1800XT isn't available yet that wouldn't make it faster at anything....

---

:rolleyes: :sick:

now I wonder if you could set up 4 x1800xts in a Quad Sli board with a Dual core proc for insane folding...

If you were JUST buying cards to fold with, The clearspeed option looks easier to manage technologically really. Mind you, I'm sure you could find another use for a quad card system occasionaly :slobber:

Buy a video card... fold for XS :)