Posts by Paydirt

21) Message boards : Number crunching : Naive Quad Core to be released in August. (Message 47186) Posted 29 Sep 2007 by Paydirt Post: I think most things will benefit from multicore in the future. The newest games are utilizing multicore. Crysis, Bioshock, etc.
22) Message boards : Number crunching : OMFG!!!!!!! folding just broke the teraflop boundrie!!!!! (Message 47185) Posted 29 Sep 2007 by Paydirt Post: Cool. For folks willing to deal with the hassle and maintenance of keeping it running, the GPU folder (x1950pro or x1950xt) is twice as powerful per total dollar spent (including power costs).
23) Message boards : Number crunching : Naive Quad Core to be released in August. (Message 47107) Posted 27 Sep 2007 by Paydirt Post: Nice chart Tiago. Looks like the Q9450 with 12MB cache and 2.66GHz will be the way to go in January at $316. Not sure why anyone would buy a dual core unless they were restrained by motherboard...
24) Message boards : Number crunching : CPU Comparison question (Message 47089) Posted 27 Sep 2007 by Paydirt Post: This conversation has evolved... Anyways, some thoughts. Einstein awards more points for similar computers. In my mind, it has potential for astrophysics, but nothing as immediate as the medical potential of Rosetta. SETI has optimized apps, but they have more processor power than they need and actually need cash donations more than processing power (they need money for server and backup equipment) [they actually have a dude crunching on a Pentium 60MHz!]. I think it'd be cool if we detected ETs, but still protein research could have an immediate impact on the quality of life here on Earth. In the end: try to optimize your RAC for the project that inspires you the most. For me that's Rosetta & Folding (Rosetta being #1, and Folding allowing me to crunch on GPUs and PS3s).
25) Message boards : Number crunching : ATI R600 cards? (Message 46980) Posted 25 Sep 2007 by Paydirt Post: :) hehehe. Um, let's just say that the number is greater than zero and is significant and I'd probably have to figure out where to put all the computers.
26) Message boards : Number crunching : OMFG!!!!!!! folding just broke the teraflop boundrie!!!!! (Message 46969) Posted 25 Sep 2007 by Paydirt Post: Why not? Because Folding@Home is a legit project and why use 1/8th of a PS3 on a project you like a little more when you can use the full power for Folding@Home? Folding has already done a lot to improve efficiency of their crunching and I'm fairly certain they have shared these ideas/results. If I could donate 35 gigaflops to Folding or 5-7 gigaflops to Rosetta, I would donate to Folding, no question in my mind. Crunching on a PS3 in a suboptimal way would not be progress. Just look at Windows 3.1 through Windows XP... It'd be best for the Rosetta project to design/program something from the ground up.
27) Message boards : Number crunching : ATI R600 cards? (Message 46968) Posted 25 Sep 2007 by Paydirt Post: I guess what I'm trying to say is that Folding@Home is a worthy project. Though we may not be as excited about their way of doing the protein folding science, I still consider them #2 to Rosetta in their approach. I guess I worry that folks will spread themselves over many projects so that all projects do not have enough computing power for a first definitive BREAKTHROUGH. I wish folks would just stick to either Folding or Rosetta. Oh well. If Rosetta were to able to fully utilize PS3 or GPUs, then I would commit (on the spot) to donating 6% of my income to build and upgrade a crunching farm just for Rosetta.
28) Message boards : Number crunching : OMFG!!!!!!! folding just broke the teraflop boundrie!!!!! (Message 46909) Posted 24 Sep 2007 by Paydirt Post: F@H USES the SPEs. So the theoretical peak for the whole PS3 is around 200 gigaflops. Right now, F@H is getting 35 gigaflops (if you're crunching 24/7, it's 35, not the average of 25); they think they can realistically get 70 gigaflops out of the PS3. Please don't port stuff to only use the PPE. If you're gonna go PS3 or GPU, then use projects that utilize the whole system. F@H's GPU beta is getting 90 gigaflops per ATI x1950xt which is still superior than the PS3, but far more user-intensive. You can even do AGP (instead of PCI express) versions of the x1950xt...
29) Message boards : Number crunching : ATI R600 cards? (Message 46908) Posted 24 Sep 2007 by Paydirt Post: GPU vs CPU crunching... #1) You can't game and GPU crunch at the same time. You can surf the 'net and GPU crunch. You can watch videos and GPU crunch but it slows down the WU by 50% (even once you're done watching the videos). #2) If you're taking money out of your pocket to buy dedicated folders, it makes sense to go with some GPU or PS3 crunchers. Check out my first post on this page: http://forum.folding-community.org/ftopic18624-30.html Basically the total dollar outlay per gigaflop (from upfront purchase cost to ongoing power costs from 24/7 crunching), is much lower for GPU crunching; or stated differently, performance per dollar spent is much higher. My comparison was generous in estimates to the PS3 (to prove a point). GPU > PS3 > CPU. PS3) The DC programs that do not utilize the PS3's 7 or 8 extra chips (SPEs) are a waste. You can run Boinc on PS3 through linux, but you won't be using the full power, so you're better off with Cure@PS3. #3) ;) We still need CPU crunchers because there's a limit to the types of units/calculations that the GPU crunchers can do for now. So what I'm doing is putting current dollars to work as GPU crunchers and saving dollars for future CPU crunchers. :)
30) Message boards : Number crunching : ATI R600 cards? (Message 46730) Posted 21 Sep 2007 by Paydirt Post: OK... #1) You must dedicate a CPU core for each GPU you have crunching. Otherwise, performance stinks. You can use a low-end CPU core such as Pentium D. You can't crunch anything on this CPU core. If you have dual-core CPU, you could use your 2nd core to crunch. #2) No R600 client for several reasons. (a) remember the GPU client is still in beta. (b) lead GPU programmer was focusing on a PhD thesis. (c) Apparently, it's not simple to "port" from X1950 to R600. The shader processing units are different. However, Vijay Pande (the director of Folding@Home) just assigned two people to the GPU client. They will also be porting to the GPU client an algorithem they just implemented for the PS3 which will boost performance.
31) Message boards : Number crunching : Improvements to Rosetta@home based on user feedback (Message 44058) Posted 22 Jul 2007 by Paydirt Post: Hey Jim, I wanted to share some of my thoughts about BOINC using processor time. I've found that it does not cause any noticeable slowness in machines where I have it set to 100% CPU. The reason this is, is because BOINC is set to the "LOWEST" priority so whenever the computer needs the CPU for something else, BOINC will temporarily get out of the way. I don't even turn it off for gaming. I do turn it off for spyware and virus sweeping.
32) Message boards : Rosetta@home Science : FAQ on diseases being explored by RAH is out of date (Message 44057) Posted 22 Jul 2007 by Paydirt Post: I understand your point about the FAQ. I think the best source for information about what the scientists are doing is in Baker's Journal here in the Science forum. The science they are doing IS important and I think they have the potential to do a "better" job than FAH, but they don't have as much computing power. I presently find disease research more inspiring than astronomical or physics research.
33) Message boards : Number crunching : Naive Quad Core to be released in August. (Message 43044) Posted 2 Jul 2007 by Paydirt Post: Anyone who reads these forums knows that Who and Penguin throw zingers at each other. It's so easy to be offended by something, but it is really our choice to be offended by something, is it not? Who or Penguin do not choose our emotions, we do. Anyways, to the point of "planned obsolescence"... I disagree. I once sat down with an Intel engineer who said in 2000 that they had the next 10-15 years of designs already made and that they simply wait for the "right market" to release everything sequentially. A rumor... Yeah, something better will always be around the corner and a vast majority of informed PC buyers (and a decent number of the uninformed) realize this. Its important for these discussions to take place (especially reviews & side-by-sides), because most people plan their purchases. Speculation about the future is important, especially when people are plunking down hard-earned money for a dedicated crunching machine (essentially a donation to science). ...I have to doubt that AMD will have a solid cruncher until their CPUs have 2MB cache per core (which R@H makes good use of).
34) Message boards : Number crunching : Folding@home GPU app? (Message 43035) Posted 2 Jul 2007 by Paydirt Post: Not for F@H. SETI does not need the crunching power, they actually need cash more than crunching power. The R@H site and forums are now up. If you end up getting another card, do NOT go with the bare minimum or a 256MB card, you will be disappointed.
35) Message boards : Number crunching : Could GPU's mean the end of DC as we know it? (Message 42618) Posted 25 Jun 2007 by Paydirt Post: They could have "enough" computing power where those pesky humans at the science department at Washington aren't able to come up with enough new ideas to run computations on. I guess there are always proteins to figure out the 3D structure of and interaction simulations for designer proteins... Eventually though if they had a TON of computing power, they'd run out of stuff to do :)
36) Message boards : Number crunching : Could GPU's mean the end of DC as we know it? (Message 42604) Posted 25 Jun 2007 by Paydirt Post: I agree that it is not the end, only the beginning for DC science. Yeah, we might see a day where F@H & R@H have more computing power than they need (which would be GREAT). How much will ClimatePrediction need before it is "full"? Also, I think you will see DC solutions being used for a wider variety of issues/problems. AI, space travel, psychology, sociology, RNA medicine, etc. Maybe computing power will become very cheap? That would be cool. JUST NOW we have kids growing up with Google, the Internet, etc... They will have a different perspective on things and they will have new ideas on how we can harness it all for the betterment of mankind.
37) Message boards : Number crunching : Folding@home GPU app? (Message 42603) Posted 25 Jun 2007 by Paydirt Post: The F@H message boards are back up. My advice in general would be to consider the total system cost when you are pricing a cruncher/folder. So if a card is 25% better cruncher but costs 40% more than the lesser card, it probably is only raising the total system cost by 15% or whatnot, so it is a good purchase. You gotta take into account all the resources that are being tied up. Get a 2-slot card to keep your system heat down.
38) Message boards : Number crunching : Folding@home GPU app? (Message 42450) Posted 22 Jun 2007 by Paydirt Post: yeah you need at least a x1650, and only certain one(s) of that line work. Their forums have been down for what seems like a week :(
39) Message boards : Number crunching : GPU-Tech Releases GPU Computing API, Benchmarks (Message 42313) Posted 19 Jun 2007 by Paydirt Post: First, I'm not an expert, like you folks I just read a lot about this stuff. @Paydirt: What effect do you believe 2Q/3Q release of chipsets supporting PCIe 2.0 and the announcement that 1GB HD2900XTs arrive will have on this type of setup (quadcore + quad-gpu (crossfire/sli))? 8x and 16x PCIe lanes get about the same crunching performance. The performance hit is when they are dropping down to 4x. I don't think 32x would do much. I do think if more "lanes of data" are added and if quad cores can handle that and the various "buses" can handle that, then you may get better returns on running dual to quad GPUs. The crunching performance improvement for the 512MB 2900s is not expected to be linearly better than the x1950xtx (I was hoping that it would be...). The (good) problem is that the shaders are 5-dimensional vectors, whereas the x1950xtx uses 4-dimensional vectors. The Folding@Home code isn't designed to use that 5th "register" in each shader. The code will likely not utilize that for quite some time. So it's not going to be 6 to 7 times more powerful than x1950xtx (which is 3 times more powerful than PS3). It's estimated that it will be 70% to 3 times more powerful in the present state of the code. The code that will crunch on 2900 has not been released. I'm waiting for some numbers from after the release to make my next purchase decision. Can any current graphics cards saturate a PCI 8x link? Even if they did, would a crunching client require that much bandwidth? I'd think the GPU would be crunching on data held in memory for some of that time. I'd be interested to see how 2x 8x lanes cope ;) Crunching CAN require a ton of bandwidth. For instance, I remember reading that each work unit for SMP folding (where all the cores work in tandem instead of independently, I think) passes over a terabyte worth of data around in the system. For GPU crunching to work with the current DirectX, each GPU requires a CPU core that does "polling" to coordinate the work. (DX10 is supposed to get rid of this need, but it hasn't yet) I think what is happening is the CPU is asking "Is shader 15 done with the calculation yet? How about now? How about now?" That's why a CPU core is required for each GPU core. And though while the data is "dumb" a lot of data passes through. Dual GPU crunching on a Dual core CPU DOES take a performance hit. I'm guessing here that it is because total system bandwidth is being strained.
40) Message boards : Number crunching : GPU-Tech Releases GPU Computing API, Benchmarks (Message 42289) Posted 18 Jun 2007 by Paydirt Post: The problem with 2 or more GPUs is whether or not the PCIe lanes will run in 16x mode... Then you have overall system bandwidth and latency. For motherboards that can handle multiple 16x at once, the question becomes one of cost: Is the more expensive motherboard worth the increase in performance? When I made my decision with dual x1950xtx's, the then extra $100 for 16x motherboard was too much for the potential 7% performance gain over 8x. Turns out my mobo runs both cards in 4x, so it was a bad choice. Presently, there is no crunching client for the 2900, but it will be coming within two months or so (I think). FAH won't say when it comes because it isn't their policy to announce until release (beta or otherwise).

Previous 20 · Next 20