Search results

The new v7 client primarily benefits Kepler on the Nvidia side and 7000 series and newer cards on the AMD side. This is not really due to the improvements in the client (the client itself doesn't do any calculations) but because the new GPU core (core 17) that came out about a year ago only...

From the description you must be running a core 17 WU right now, as there's no way you could be getting 200k PPD if the GPU was processing a core 15 WU. The usage drops you're seeing are probably the core doing its checkpointing, and that's normal behavior. The one flag that would be...

Patriot, do the E7s work in the same socket as the SB E5s?

No, I don't remember that. I may have been folding since 2007, but I wasn't actively involved with the forums until mid-2008, and I probably wasn't paying much attention to GPU issues until mid-2009, when I finally bought a GPU that could actually fold a WU to completion. I didn't join a major...

What are these four episodes? In the discussions over the BA changes/eventual cancellation, several people have mentioned something to the effect of "they do this over and over," but I haven't actually seen these events enumerated. I've been with the project since 2007, and here's the...

There is no point bonus for folding beta WUs, be they CPU or GPU. Never has been, and I doubt there ever will be. Usually it's the opposite, where projects just entering beta have been poorly benchmarked and earn low PPD until enough beta testers say that it's low compared to what they're used...

One question about having mods be eligible for recall: what would be the quorum for the vote to count? If you set a 75% approval threshold, but only 12 people bothered to vote and 9 said yes, is that really enough to remove someone? There's around 9700 registered members on FF, but I don't...

BA may be time-sensitive and cutting-edge, but I have to think that a large part of their decision to end BA in a year is that it doesn't make up a large percentage of the total work they do, and the drama it caused was disproportionate to the usefulness of the results. There's only been...

I think they ended it because of the amount of drama it caused vs. the amount of their research it actually made up. It seems like 90% of the major griping over the past few years has related to BA when at any given time only a few hundred contributors were actually running it, and in total...

A major announcement:

You're not the first one to suggest this, and I like the concept. I'm not sure how closely it aligns with PG's plans, but it would eliminate the massive point drop that comes with being booted out of bigadv. I would guess that most people folding bigadv aren't particularly interested in the...

For me, at least, pausing and restarting the slots resets the retry timer. I didn't need to reboot the whole machine. I think the increase of time between requests is in large part to not overload work servers that have just come back online. At first, a lack of response could be a...

There's over 200,000 SMP WUs waiting to be assigned according to the server status page, and that's just the current gens. There's got to be tens of millions of WUs to get to the end of the all those runs. Even with a very coordinated effort, I don't think the folding community is going to run...

This is very true. On the couple of occasions that my Intel 4P has picked up a regular SMP WU, it's done around 200k PPD. You can now do better than that with a 780 Ti or 290X, and those are far cheaper than an Intel 4P.

They are almost equal. I have v7 running on my 4P 4650, and I get just about the same frame times as others with the same setup on v6. I have the Kraken and a ramdisk set up; I think the only optimization missing is you guys' kernel.

They're certainly a lot better in terms of price and power efficiency than any desktop CPU. Many bigadv rigs are more efficient, but most of them cost a lot more up front.

That sounds right.

The main caveat I see is that they explicitly mention testing one's CUDA application. Since core 17 is OpenCL, would it not run on this test setup? That would mean running core 15 WUs, which wouldn't be nearly as profitable points-wise.

Some thoughts on the Impact and Effectiveness slide: F@h may date back to 2000, but distributed computing in general goes back a few years earlier. Look at the Wikipedia page "List of distributed computing projects", sort by Launched, and you'll see that GIMPS goes back to the start of 1996...

Which drivers? I'm on 331.65 and I find them to be better than the last couple of pre-331 drivers (331.58 is virtually identical). 331 did drive down GK104 noticeably. I wouldn't be surprised by 210k. I have both my 780s running at 1215 MHz core. 8900 and 7811 nets about 173k, and 7810...

Is that stock clocks or OCed?

1. There were a few articles on gaming sites about the PS3 client shortly before it came out. I didn't realize at the time that there were PC clients as well, but I figured my PS3 was off most of the time and it sounded like I could do something useful with it. It helped that I was still in...

I'm guessing they'll still only fold well on Kepler and GCN. They run on Fermi and VLIW, just not particularly fast.

You only put one client-type flag. The better option PPD-wise is to use beta, as with advanced you'll probably end up getting a mix of core 17 and core 15 work, with the latter earning far fewer PPD. This isn't the officially endorsed route, as they only really want beta team members using the...

Something about the GCN architecture is much better suited for folding with core 17 than VLIW4/5; similarly, on the green side, Kepler is much better than Fermi. The PPD figures on a 6970 are rather sparse, and most of the google results are from the early days of core 17 before the OpenMM 5.1...

You got a link to these posts?

AMD cards can't fold in Linux at this time. Official word was once they put out some halfway decent Linux drivers it might become viable.

Double posting is super awesome.

I'm guessing you haven't added the slot option client-type:advanced to your GPU slot. The GPU core that you get right after a standard install is not very efficient and has low PPD. There's a new, much better GPU core that hasn't gotten a full release yet, which is what you'll get if you...

The CPU doesn't matter. Pretty much any vaguely modern CPU core can keep a GPU properly fed if it needs much CPU time. Currently AMD cards on core 17 do not need much CPU time, while Nvidia cards on core 17 need a dedicated CPU core. Core 15 for Nvidia is still out there and doesn't require...

You all think the SMs are loud? I've never done an actual test, but I swear my 4P is a) about the same loudness as the desktop that used to be my second box and b) quieter than my 3770k/dual OC'ed 780s box (I run one of the cards at 75% fan speed to keep temps under control, as that one doesn't...

On that note, one of the handful of narrow HSFs is the SuperMicro SNK-P0050AP4 (it comes with both square and narrow brackets). I find that they keep my processors sufficiently cool, at a max of around 60C when the processors are set to always turbo-boost to 3.1 GHz.

I don't think it makes sense either. I'm actually sort of waiting to see how Ivy-EX shakes out. I get the feeling 15 cores will be even more ludicrously expensive than 10...

I'm running BIOS 1.0d, which is the latest one for the X9QRi-F+, and it recognizes my C0 chips. I assume it recognizes C1s and C2s as well. On a related note, while I've seen the handful of 4650 v2 listings on ebay, I've been assuming that I couldn't run them right now since SM hasn't released...

It hasn't been whitelisted yet, so you won't be able to fold with it right now.

I don't think your slot 1 card's PPD woes are due to it driving the display. I recently added a second 780 to my rig running on a Sabertooth Z77; this rig has had the iGPU drive the display for a long time. After the addition, the card in slot 1 is consistently slower than the one in slot 2...

Desktop CPUs have really fallen out of favor among the larger contributors since their PPD/W and PPD/$ is not very good. GPUs became much more attractive in both regards with the introduction of a new GPU core earlier this year. It runs equally well on AMD and Nvidia cards. The AMD...

For $300 you'd probably be best off buying a 7970 and adding it to the machine in your sig. That would get you in the neighborhood of 90k-110k Points Per Day depending on overclock, which is a pretty nice amount, and it should be very simple to get it up and folding.

I don't know how they do on anything other than a 780, but before these came out I was running 327.24, which were developer beta drivers. They were stable and faster than 327.23 and 326.80, albeit slower than 320.49. As a point of comparison, on my 780 project 8900 had a TPF of 2:51 on 320.49...

Based on reports from EVGA and OCN, it appears that the latest beta drivers, 331.40, really tank people's PPD unless they have a 780, e.g. going from 96k to 35k PPD on a 770. For me, at least, they improved my 780's PPD over 327.23, which had gotten rid of the 36-hour crash but which was...