QuintLeo

If you like small work units (but configurable) you should check out distributed.net The "default" work unit size for RC5-72 runs so fast on my higher end GPUs that it takes them longer to load and unload than to run the units. Can also check out the BOINC MooWrapper project, which is basically distributed.net RC5-72 work with a "make it work with BOINC stats reporting" wrapper.

Curecoin no longer uses SHA256 ASIC - they used to for "20%" of their output to secure their blockchain, but Curecoin 2.0 moved to a different option for that some months back. I've sometimes thought it would be interesting to get a chip designer to come up with a Folding-specific ASIC, but I doubt it would be worth the cost as Folding is a LOT more complicated process than cryptocoin processing is.

ATI/AMD cards are compatable with Folding - but they tend to offer lower performance by quite a bit per watt. Some of the newest ones are getting closer - but even the current top-of-the-line AMD Vega VII is a thin hair faster to the same on PPD and similar power draw vs my generation-old Nvidia GTX 1080ti cards, while the current NVidia 20xx high end cards are almost twice the PPD/watt. Folding for some reason is more efficient on NVidia cards and has been for several generations now, I'm not sure why. There are projects where AMD cards have been higher to the same performance though - I'm not sure what the current numbers look like, but the MooWrapper BOINC project (Dnet RC-72 underneath) showed a slight advantage to the Vega64 vs the GTX 1080ti on stock settings that grew with tweeking, and even my "reference" model Vega56 is very close when tweeked to my top-end Aorus GTX 1080ti cards when tweeked - both are capable of close to 10 GigaKeys/sec (Vega a bit below, 1080ti a little above).

They've had ongoing intermittent issues with some of their work servers getting full disks. Just have to wait 'till the work server get some workspace cleared off it. Try HFM.NET as a monitor program - it's been a bit flaky lately with all the issues the clients have due to the out of work problems, but generally does a good job and is highly configurable. I've NEVER gotten the actual FAHClient to work worth beans at monitoring any client not on the same machine.

Is that an original IBM with the moving plate cap pickups, or the later IBM by Lexmark dome pickup junk? I've got a couple of both, the original is a LOT better overall.

F@H stat server might be getting overloaded. Wouldn't be the first time stats lagged, but this time they have more reason for it. The Overclocking stats site relies on the actual F@H stats servers for IT'S data, on an every 3'd hour stats file pull. Problem MIGHT have cleared up, there was a big "bump" on everyone early this morning.

So they likely are still working on that million USER count, I'm sure AWS has a lot more devices running than CoreWeave's "over 6000".

Do you have your passkey set up?

Remind me to post a few shots of my "built on a 2x2 foot sheet of particleboard" builds. There are pics of the "rack/shelves" they're mounted in posted somewhere on the bitcointalk forum but I don't remember if I ever posted any of the rigs themselves - probably in one of the PhillpMa1957 threads.

No shock - there were a TON of them for a while working in the Distributed.Net project. MOSIX or Beowulf were widely popular around that timeframe.

Drop the Ryzen down to a 4 core if you don't already have it - one core (NOT Hyperthreaded core, real one) per GPU is plenty and that lets you upgrade the machine to 3 or 4 GPUs eventually. Most of my folding systems run either an AMD A10 (commonly the 7860K) or an overkill FX-6xxxx/FX-8xxx series CPU (I used to mine Monero on those). 2080 cards seem to be the current sweet spot on PPD/$ and PPD/watt - and leaves you more space for future expansion. I like the EVGA B2/G2/T2 series but do NOT like the longevity on the fan on their non-ball-bearing models (which I think includes the BQ/GQ series) while the newer BA/GA series seem to be having some teething pains right now. Seasonic X or SS series are also good choices, but the X series is no longer made and I think they also discontinued the SS series. Their Prime and Focus series are non-BB stuff. I used to list EVGA as my go-to card maker on the Nvidia side, but their move in their 2xxx generation to "fancy name sleeve bearing fan" usage has dropped them off my list entirely. Gigabyte Aorus series is currently top of my list on the NVidia side, though I have OFTEN wished Sapphire would start making Nvidia based cards not just AMD based ones. RUN LINUX. It is a little more efficient even on single-GPU machines, and if you have to start using 1X slots for expansion it takes a VERY SMALL performance hit compared to running Windows (ANY version). All of my folding machines have been running Ubuntu for years, but Debian or Mint work just as well by report and any Linux distribution you can get the needed libraries and NVidia drives working on should also work. I recommend against Slackware, getting it working is a NIGHTMARE at best due to the lack of a good package management tool like apt / dpkg. I have no case recommendation, as none of my current rigs use a case - on the other hand, they are NOT child or cat safe so I don't recommend them for everyone.

OK, some of you been around a little while. Any "Quintleo" entry however is me - though I was working with Distributed.Net long before F@H existed.

When the PS3 first showed up, there was NO "GPU compute" at all. The CPU in the PS3 was one of the most powerful in the world at the time - in a FAIRLY CHEAP game machine. Today, a single low-end GPU would blow away ANY console - for reference, the graphic GPUs in the current PS or XBox models are very close in specs to old AMD A10 7860K/7890K APU except that one of them uses GDDR5 instead. That is about equal to the HD7750 in capability (same core count, same or slower RAM, a little higher clock rate). By CURRENT GPU standards, that's not even an ENTRY level card - the GTX 1030 was comparably to a bit faster (Folding works somewhat better on NVidia as they handle complex calculations better due to how they are designed, presuming same core count and clock speed and ram speed) and anything current in a new GPU is quite a bit faster. It's not comparable at all to the early PS3 days when they offered an insane amount of compute power for cheap. BTW - the PS3 was NOT initially intended to run LINUX, folks had to hack into it to make that possible - then eventually Sony gave up and allowed it semi-officially.

The issue was more about there not having been enough WORK UNITS to put on the servers, till the last couple days, more than a shortage of servers to handle the load. When the user count multiplied by ballpark 50-100 TIMES over less than a week (number of GPUs didn't climb as much due to the many of us long-timers with multiple machines and multiple GPUs), the available work just poofed in no time flat. The folks at WU running Folding@Home seem to be getting a handle on that issue, particularly today units have been nearly always available when my machines have asked for them, after almost 2 weeks of shortage.

2 machines (I just got done consolidating and dropping some lower-end cards). When they all have work, 6-7 million PPD. That's 3 GTX 1080 ti in each machine though, but TDP limits set DOWN a little for cooling reliability reasons. At one point back in the day, I had more like 30 GPUs working on Folding - the 1080 tis, 7 or 8 x 1080, 6 1070ti, and the rest 1070s - but I can afford to run all of those on Folding any more.