Jump to content
Search In
  • More options...
Find results that contain...
Find results in...

Pooky

Member
  • Content Count

    31
  • Joined

  • Last visited

Awards


This user doesn't have any awards

About Pooky

  • Title
    Member

System

  • CPU
    Xeon E5-1650 v3
  • Motherboard
    HP 761514-001 C612
  • RAM
    32GB DDR4-2133R CL15
  • GPU
    Gigabyte GTX 1070 Windforce OC gen. 2
  • Case
    HP Z440
  • PSU
    700W 80+ HP OEM
  • Keyboard
    Redragon K580
  • Mouse
    Logitech G502

Recent Profile Visitors

50 profile views
  1. You're more likely to find better performance with a new CPU platform for the same net cost. X99 is great, but going back from Broadwell to Haswell to seek better performance is more indicative that you need a platform upgrade than slightly more overclocking potential.
  2. You keep saying that you found a 9700k for $292. You did not. You found a 9700, which is not unlocked.
  3. When you are looking at these cheaper listings, make sure you are looking at QS (quality sample) chips and not ES (engineering sample) chips. The former are very very close and often identical to the final tray stepping, whereas the latter are often less stable and do not clock as high. You can use resources like CPU-World to verify the steppings of the chips. For example, one of the first listings for ~$250 is a E5-2690 v4 QHV4, but CPU-World says that the closer-to-retail stepping is QKE3. https://www.cpu-world.com/CPUs/Xeon/Intel-Xeon E5-2695 v4.html In general, for v3 chips, you want something later than QExx, and QKxx for v4. The steppings are in alphabetical order, so QGN3 is a later and better stepping than QGEF for example. If you want more info about how to read core steppings, feel free to inquire. Haswell-EP is the v3 mid-grade server/workstation class of CPUs based on the 22nm Haswell architecture (similar to 4th gen Core processors), and Broadwell-EP is the next generation up from that, v4, based on the 14nm Broadwell architecture (similar to 5th gen Core processors). Xeon E5 v3 and v4 both fit into the LGA2011-v3 socket and use the X99/C612 chipset. Although the E5 v4 CPUs came out about a year and a half hafter v3, they are not that much faster in terms of clock speeds or IPC, but because they are the newest available CPUs for the socket, they are notably more expensive than the v3 with very little performance gains. The biggest difference between Haswell-EP and Broadwell-EP is that the latter has a higher max core count at 22c/44t on the E5-2699 v4, vs 'only' 18c/36t on E5-2699 v3. Because of this, the E5 v3 CPUs tend to be a much better value second-hand even if they run slightly warmer. The E5-26xx CPUs do not scale up to 4 sockets. For that you will need the E5-46xx CPUs. Hint: the first number in the names represents the number of CPUs it scales with. So the E5-16xx is single socket, E5-26xx supports dual, and E5-46xx supports quad. The more scalability a CPU supports, the lower its clock speeds will be, and quad-socket motherboards are prohibitively expensive so I wouldn't recommend getting one anyway. At that point I would just get a single Threadripper 3970X. It would be better in the long run with more upgradeability, newer platform features, and much better single-core performance--likely double. Dual-socket workstations though? Yes, I think they are still a very good value for people that needs the cores. I think they're really interesting and can be an extremely good buy for someone that needs more cores than Zen offers for the same money. A dual E5-2697 v3 setup would be very competitive in rendering, but not as much with modelling and animation. The lower IPC and clock speeds might be noticeable on zBrush, but since I do not use that software I could not tell you. Keep in mind that a dual E5-2697 v3 workstation only nets ~43% more performance in C15 than a 3900X, despite having over twice as many cores. This is because Zen excels in tile-based rendering where data is not bottlenecked by the Infinity Fabric. Since you already have a 3900X and you are doing tile-based rendering and real-time modelling, I would recommend to you to skip the headache of learning another platform and sticking with your current CPU. It is very good, but if you feel like you need more power, then upgrade to a 3950X or wait for Zen3. A 28c/44t Xeon system could be a lot faster in multicore workloads that share data frequently between cores, or just needs lots of cores for many virtual machines, but tile-based rendering is not one of those tasks. If you see the value of 43% faster rendering time over slightly laggier modelling, it could be a good option, but I would reiterate that since you are already on a good platform it makes more logistical sense to stick with it. ------------------------------------------------------------------------------------------------------------ PS: If you are comfortable doing a BIOS microcode patch on an X99 motherboard, you can set the all-core turbo to be equal to the single-core turbo of the CPU and increase performance significantly. Normally, E5-2697 v3 will run at 3.1GHz all-core, but you can increase this to 3.6GHz for even faster multicore performance. 3900X vs 2x E5-2697 v3: https://browser.geekbench.com/v5/cpu/compare/3963170?baseline=3912842 3900X vs 2x E5-2697 v3 OC: https://browser.geekbench.com/v5/cpu/compare/3963170?baseline=3701802 You can see that the OC improves multicore performance by about 21% and goes from 47% faster than the 3900X to 79% faster. If you feel comfortable doing such a modification (check out a guide here or here) then this may help to sway your decision. If it seems over your head, which is perfectly okay, then you can ignore this last bit and only focus on stock performance numbers. Cheers
  4. I would not recommened the E5-2695 v4 at its current eBay price of $400+. I would recommend instead either dual E5-2697 v3 QS (28c/56t) for $400 or dual E5-2697 v4 QS (36c/72t) for $500. The v3 Haswell-EP chips are generally much cheaper than the equivalent Broadwell-EP chips on eBay, at least at the moment, but it is possible to find some good deals. For example.... If you pair two of these together, they will absolutely spank the 3900X in raw multithreading performance. If your goal is absolute multithreading capability and you aren't as concerned with single-thread performance, going with a dual-Xeon setup will be a much better value. However, their single-threaded performance lacks by significantly by comparison. For my purposes, buying a E5-2697 v3 at half of the cost of a 3900X was abolutely the better buy. You may be able to find deals on Threadripper Whitehaven (1st Gen) or Colfax (2nd Gen), but from my experience, the platform and CPU costs are still outrageously high on eBay ($800! for a 2920X, and $320 for a 1920X at the time of writing). Cheers
  5. Just snagged an E5-2697 v3 QS for $180 on eBay. This is will make a nice upgrade to my E5-1650 v3 and should allow for much greater virtualisation expansion. What a bargain! Not bad for 3.6GHz dual-core and 3.1GHz all-core. This should satiate me until Skylake-W or Threadripper Colfax come down in price. There are good deals to be had with the Skylake-W Apple QS chips on eBay but unfortunately the cost of entry into Basin Falls C422 is still too high at the moment. What are some good deals you have found?
  6. Great! Out of curiosity, could you share the turbo steppings per core? This information is not readily available online. All I currently know is that it does 4.2GHz on the first two cores and 3.9GHz with 7-8 cores. Information about turbo for 3-4 and 5-6 is unknown. The information should be easy to find with a tool like HWiNFO: Cheers
  7. I suppose I might have won the eBay silicon lottery, then! Not bad for $50. It was a good chip for sure, but I traded up for a E5-1650 v3 and haven't looked back since. Now that the E5-1650 is cheaper than a W3680 on eBay, and the E5-1650 v2 is only slightly more, it's hard to justify buying or upgrading to Westmere. But I do still use it in my home server! Cheers
  8. The Westmere (Gulftown) chips are the better choice, in my opinion (unless, of course, you meant Gulftown as well since the W3590 does not exist). I was able to attain an easy 4.0GHz with a W3680 in a Dell T3500, just from increasing the multiplier. I never messed with voltage, though.
  9. This was my first thought as well, but OP seemed to imply that this is a relatively new PC build. It's worth checking for malware, but if they are starting from a new OS install, malware does seem rather odd. @LeeCope3 If you open up Resource Monitor, does the PC exhibit the same behaviour or can you find the resource hog?
  10. Okay, I've just had a look at some benchmarks with hardware similar to yours, and I think that v-sync is limiting your FPS to your screen's refresh rate. Check the game's settings to see if v-sync is enabled in there and if so, try disabling it. If v-sync is disabled in the game's settings, try Nvidia Control Panel instead. Follow this guide but choose "Off" rather than "Forced on" and save your settings. Let me know if that uncaps your framerate. Cheers.
  11. unclewebb, the author of ThrottleStop, explains it like this: So basically, it is the normal limit the system puts on the CPU to prevent it from overheating. This does not appear to be anything out of the ordinary. Apologies if I am forgetting, but did you ever tell us what resolution and FPS you were getting on Valorant? If your settings are low enough and you are capped at your screen's refresh rate, it's possible that the low GPU usage is cause by v-sync is being forced on. Unfortunately I need to sleep now but hopefully some other, more helpful fellow can pick up the slack while I am off. Cheers.
  12. It wouldn't hurt to try reinstalling drivers, but since you claim that it did not work when you tried, it might be better to check for power limits first. All you need to do is run the software I linked and press the "Limits" button under "PKG Power". That window will show you anything that is limiting power to your system, both CPU and GPU.
  13. Check my edit. I have a few more ideas about power throttling but it would require downloading a software called ThrottleStop to check for limiters. Ex:
  14. I do not see any issues on the CPU side of things. From the way it looks, there might be an issue with your GPU drivers. I noticed in your post history that you had an issue with Valorant before following a GPU driver update. Have you tried uninstalling your current driver with DDU and then reinstalling the driver from GeForce Experience? I can't quite make out what model of GPU your laptop has but since your CPU is very capable, Valorant is well-optimised, and your GPU isn't pegged, I suspect that there may be a driver issue at play. It might not even be a GPU driver issue, but that would be the first thing to check. EDIT: I've also just thought, since you're on a laptop, it's also possible that there is power and/or thermal throttling limiting your GPU's performance. Could you check which power plan you are using in Windows? If you are not plugged in or if you have the wrong power plan selected, the system may be throttling power delivery. Cheers.
×