Jump to content

Gorgon

Member
  • Posts

    3,436
  • Joined

  • Last visited

Everything posted by Gorgon

  1. I had one of my dual GPU motherboards fail during Folding month and so had to move the GPUs to my last two remaining spare AM4 motherboards which are Gigabyte Aorus b450m models I picked up for $80Cdn each when they were on Clearance last year. But these boards are micro ATA (mATX) and thus could only support one GPU each. With the 4080 Supers being better values now I picked up a couple of the Gigabyte Gaming OC models. Being 3.76 Slots wide these big boys fit with just a couple of mm to spare to not overhang the I/O connectors at the bottom of the motherboards. A case swap was also required for one system as the Fractal Define C was too short to accommodate the 342mm GPU length so the guts were transplanted to a spare Fractal Define S. As expected these GPUs are most efficient being at the top of the AD103 stack. One even draws just 3W at idle.
  2. Next up is the 4080 Super. I picked up a couple of Gigabyte RTX 4080 Super Gaming OC GPUs. As expected, these being at the top of the AD103 Stack they are most efficient at 2205MHz close to their 2295MHz Base Clock. Running at 2730MHz, close to their Default Power Limit setting they consume about 275W and yield about 21MPPD. Moving to the efficiency peak at 2205MHz they consume about 167W and produce 17.6MPPD a decrease in Yield of 16.5% but a 39% decrease in Power.
  3. Doing some research for something else I surveyed the x670e motherboards to see which ones have PCIe5 x8/x8 slots: About the only "reasonably" priced x670e boards out there are the Asus ProArt and Strix x670e-e and the ASRock TaiChi. The MSI Carbon WiFi would be a contender but rather than shifting all the PCIe slots down 1 slot like the rest they decided to shift everything down 2-slots meaning the lower card will, in most cases, collide with the PSU Shroud. Gigabyte appears to be avoiding having to buy re-drivers on their 670e motherboards in favor of even more m.2 slots and so are out of the running. Which makes me wonder why the Extreme and Master are so expensive.
  4. Jayz2cents likes a toohbrush and Crest 3Dwhite toothpaste for cleaning those
  5. Sigh. I'm not going to engage. You have offered your opinion and I mine.
  6. No, they are not perfect. Those are 20km rated 1310nm SFP modules and will require attenuators for such a small run or you WILL risk burning out the receivers at either end. The Multi-Mode model rated at 550m maximum would be better in this use case. See:
  7. Yes, you can use 100BaseTX (Copper) to 100BaseFX (Fibre) transceivers but these are generally a Pain in the Ass as they require external power and can be flakey (intermittent failures for no good reason). There also are 100BasFX SFP Modules but you'd have to figure out if the Switch and NIC support them. The fiber you linked is Single Mode and would be for more expensive "Long-Haul" applications. You would typically use Single Mode Fibre like this but that vendor only has lengths up to 100m but if you shop around you should be able to find longer lengths. The Transceiver you linked is for Single Mode Fibre but is designed to drive a signal 25km so WAAY more than you need so you would have to use attenuators to drop the signal level down or risk burning out the receivers at either end. It is also single-ended and only uses one fibre for both directions (each direction runs on a separate "color" (wavelength) of light) and uses a single SC connector rather than the more common (these days) LC ones. While you can also use Fiber with "SC" connectors they are much larger and harder to pull through smaller conduits (pipes). Yes, you likely can find 100BaseTX to FX Media Convertors it is probably easier these days just to use Gigabit (1000BaseT to FX) convertors or SFP modules as they are much more commonly available.
  8. You definitely want to use Fiber for this for Lightning protection etc as a long Copper Ethernet cable is like a giant antenna. You should just need 2 1000BaseSX Small Form-Factor Pluggable (SFP) and either 50/125 or 63/125 micron Multi-mode Fibre for Gigabit Ethernet. If you want 10 Gigabit then you'll need switches with SFP+ ports. If your burying your services (electrical, water) then be sure to install an additional separate conduit (plastic pipe) for networking and also install a pull cord along with whatever networking cabling you use. In the shop, if there's only one device (PC) you could just install a Fibre Network Interface Card (NIC or Adapter) and terminate the Fiber on it. Definitely purchase the Fiber cable of the appropriate length with the "LC" connectors pre-installed as terminating fiber is not for the inexperienced.
  9. Likely vmWare is "re-arranging" things. Long time vmWare user here - after this cash grab I'm totally dropping them and moving to KVM. Expect much less support and innovation from vmWare as they ditch all but the largest enterprises
  10. An oldie but a goodie - z370 Aorus Gaming 5 with a i9-9900k (remember when we used to think these ran hot?)
  11. Picked up a couple of 4070 Ti Supers to add to the Fold. They appear to be most efficient around 2400MHz. 450W 29MPPD Mini Space Heater:
  12. With the refresh to the 4000-series I picked up a couple of MSI 4070ti Super GPUs, one a tri-axial Gaming X Slim and the other a dual-axial fan Ventus 2x. Both of these cards are slightly under 2.5 slots in width and thus are suitable for running in a dual GPU rig with 3-slot spacing if one applies clock or power limits. Now these two "4070s," unlike the older 4070ti which sat at the top of the AD104 stack, sit at the bottom of the AD103 Stack (4070 ti Super; 4080, 4080 Super). The implication of this is, unlike the 4070 ti which would have been the best binned silicon in the AD104 stack, the 4070 ti Supers are likely the worst binned silicon in in the AD103 stack so we should expect them to be less efficient that the 4080 or 4080 Supers like we observed with the 4070 where it, and the lower models in AD104 stack showed less efficiency compared to the 4070ti. Here's the Efficiency vs. Clock Speed plots for both GPUs: Both these appear to have a peak efficiency around a clock speed of 2400MHz.
  13. what does nvtop show? About all I can think is the device is thermal throttling but I assUme you've looked for that.
  14. Mostly on Gigabyte x570 AM4 systems (Master & Pro) with a mix of 3000 and 5000-series R7 and R9 processors (see Signature)
  15. Hmm. Those Wattage and Utilization numbers look low for the 4070tis. Could it be a driver issue? Are you Power or Clock Limiting? (I doubt it as Persistence Mode is off.) Here's what two of my 4070tis look like (p18219 & p1812226), both at ~12.9MPPD Clock-limited to 2205MHz: +-----------------------------------------------------------------------------+ | NVIDIA-SMI 525.147.05 Driver Version: 525.147.05 CUDA Version: 12.0 | |-------------------------------+----------------------+----------------------+ | GPU Name Persistence-M| Bus-Id Disp.A | Volatile Uncorr. ECC | | Fan Temp Perf Pwr:Usage/Cap| Memory-Usage | GPU-Util Compute M. | | | | MIG M. | |===============================+======================+======================| | 0 NVIDIA GeForce ... On | 00000000:0A:00.0 Off | N/A | | 66% 76C P2 150W / 285W | 502MiB / 12282MiB | 89% Default | | | | N/A | +-------------------------------+----------------------+----------------------+ | 1 NVIDIA GeForce ... On | 00000000:0B:00.0 Off | N/A | | 32% 59C P2 138W / 160W | 496MiB / 12282MiB | 98% Default | | | | N/A | +-------------------------------+----------------------+----------------------+
  16. The 3070 has 5888 Shaders while the 4070 Super has 7168. So not exactly an apples to apples comparison. You'll want to compare it to the 4070 as that has the same number of shaders. So ~7.9MPPD versus 4.7MPPD. Yes, 4000-series GPUs are astounding for both their efficiency and price. Though the base clocks have increased from ~1600 to ~2000MHz the power consumption has also gone way down so not only do they perform well but they can run a lot cooler than Ampere.
  17. Here are your current "CPUs" in use (actually CPUs and GPUs) which show 2 CPUs and 3 GPUs currently folding, all of which are receiving the QRB. There is a AMD GPU which appears to be MIA since Feb 17th and a x86 GPU missing since the 18th, but that may just be a a slot being re-configured. I'd check at this point to see if you have a bunch of failed WUs. HfM is useful for this, otherwise you'll have to search the logs for "CORE_RESTART" "NaN" "exception occurred" The Log Viewer in the Advanced Client can be useful for this by checking "Warnings & Errors"
  18. Likely just a backlog of work units that need to be processed. Hopefully you'll see a big update shortly. After the 4 slots I had that were stuck trying to upload It's been smooth sailing here and I haven't noticed any major decrease in production being reported,
  19. Due to the cost of the DDR5-based systems I'm still running mostly x570 with Gen4 x8/x8 from the CPU. Unfortunately from what I recall only the b550 Vision D and Vision D-P were the only two low cost motherboards with x8/x8 from the CPU (i.e. with PCIe Gen4 switches) with 3-slot spacing. AFAIK all the b650 motherboards have only x16 from the CPU and all other slots are x4 from the Chipset and most of those are Gen3 as the PCIe Gen5 Re-Drivers and Switches are hellishly expensive and Motherboard manufacturers seem to be focusing more on lots of M.2 slots. Still if you can find a b650 motherboard with PCIe Gen4 from the Chip Set at x4 that would be equivalent to Gen3 x8 and so you shouldn't loose too much yield with "just" a 4070, But most b650 motherboards appear to have just PCIe3 x1 to the Chipset connected slots so your likely stuck with having to get a 670 or 670E which are waaay more expensive. The other option is to go off-label and get a PCIe Gen4 m.2 -> PCIe x16 adapter and use the CPU-connected m.2 slot for the second GPU and install the OS on one of the Chipset connected m.2s. But then your stuck with trying to figure out how to secure the second GPU.
  20. I'd just do a CPU and GPU Upgrade. Should get you pretty close to the performance of a 7800X3D at a fraction of the cost. If you have someone in mind to donate the system too (it's still perfectly fine for most 1080p AAA titles) then a 7800X3D, new motherboard, DDR5 6000 Memory, new case, new power supply & new GPU would work well. Intel has some pretty solid options too. That's a pretty solid motherboard. A 5800X3D or 5700X3D would work in it after a BIOS Update with the existing memory, and you could even add an additional 16GB RAM at 3200. For a GPU a 4070ti Super should give you lots of headroom at 1440p. AMD also has some better value options out there for pure fps but Nvidia is about a generation ahead in Ray Tracing and DLSS if that's important.
  21. This is a new one. I have 6 slots stuck Sending but the twist is it doesn't appear to be a network issue but with the server: core:0x22 unit:0x360300000c0200000000000025490000 15:18:19:WU01:FS00:Uploading 34.54MiB to 131.239.113.97 15:18:19:WU01:FS00:Connecting to 131.239.113.97:8080 15:18:25:WU01:FS00:Upload 19.90% 15:18:31:WU01:FS00:Upload 41.61% 15:18:37:WU01:FS00:Upload 63.50% 15:18:43:WU01:FS00:Upload 85.40% 15:18:47:WU01:FS00:Upload complete 15:18:47:WU01:FS00:Server responded PLEASE_WAIT (464) 15:18:47:WARNING:WU01:FS00:Failed to send results, will try again later My other slots, connected to other Servers, all uploaded successfully it just appears to be an issue with 131.239.113.97. The Server Stats show the unit has one error - Out of Disk Space. Whoops! There is already a post on the Support Site about the issue. I guess we'll just have to watch the QRB dwindle until they get it fixed.
  22. You may have lost the Silicon Lottery with your new CPU. Try Disabling XMP in BIOS. Have you checked if your BIOS is up-to-date and supports the newer CPU? It must be version F60 or later Upgrade you BIOS to the latest, be sure to Set it back to Default If that works try enabling XMP, If XMP doesn't work try manually setting RAM to 3200
  23. ... or a new mix of WUs have been released that perform slightly better ... I'd wait a week or so before passing judgement. I mean it's possible you might get one or 2 "bins" (15 or 30MHz) if folding at the default Power Target but I wouldn't expect to see that translate to a 2-5MPPD difference.
×