Jump to content

David Petrofsky

Member
  • Posts

    35
  • Joined

  • Last visited

Awards

This user doesn't have any awards

Recent Profile Visitors

The recent visitors block is disabled and is not being shown to other users.

David Petrofsky's Achievements

  1. Apparently the "other Linus" liked this cpu/cooler combo as well: https://www.zdnet.com/article/look-whats-inside-linus-torvalds-latest-linux-development-pc/ Before I ordered the parts I found a site that actually benchmarked 3970x on this cooler vs. several air and water cooling setups and it came out very favorably, beating a lot of water coolers in temperature and beating everything in noise. I have tried to find that article again but I can't find the right search terms this time.
  2. There doesn't seem to be anything with better timings in a size that will max out my RAM, at least on newegg. Is there one you had in mind? What kind of numbers should I look for in memtest benchmarks, etc. to know whether it is actually a problem or not?
  3. Fair enough - it's worth pointing out though that if someone doesn't know they're supposed to connect more cables (in my case the manual actually said not to), that instruction doesn't really tell you to go against the manual. I would suggest it should have a note in that step like "if your board has 8-pin power connectors try connecting them even if the manual says not to."
  4. I guess we'll find out. Linus built a very similar system for his video editors recently and I liked it so much when I saw it that I used it as my starting point. He stated in the video that U14S w/ 1 fan was fine in his tests. I added a 2nd fan just in case. The fans are a little higher off the cpu than normal because the ram is so tall, but hopefully having a 2nd fan will more than counteract that. The other things I changed from his version were maxing out the RAM and using 3 SSDs instead of 1 Intel Optane (because I don't have a 10G ethernet w/ NAS drives like they do, so I need to render and cache locally).
  5. If they don't report per core temperatures, what would you see if half the cores were hot and half were ok? Do they report throttling per core?
  6. Video editing will probably be the most intense thing. I will also be programming and testing in VMs from time to time.
  7. Hi all, I just finished the initial physical build (my first one - didn't go smooth at all but seems ok now). Now that it finally boots into BIOS and sees all the components, I'm ready to stress-test the hell out of it. I'm not familiar with the programs that are used to stress & benchmark but I spent a lot of time and money on this machine and want to make sure I got good parts and don't have any physical mistakes affecting performance. Here are the components: P600S case: probably nothing to test other than making sure all the front ports are working ROG Zenith II Extreme Alpha Motherboard: probably should test all the back ports and wi-fi speed * NOTE: the chipset fans come on right on bootup but then don't seem to want to come on even when the temp is in the 60s. Normal? How do I stress the chipset? Figure out which USB ports go through it and do a large transfer? I think this thing has so many PCIe lanes that I could never stress it that way. 3970X cpu: want to test accuracy, speed and temperature of all 32 cores - I feel like the thermal paste on one side is a little thinner and I want to see if there is a noticeably hotter set of cores *what is the program for this nowadays? I heard Prime95 got the Pluto treatment lately. *NOTE: the system is air-cooled with a Noctua U14S (dual fan, pull-push), case has 3 fans rtx 2080ti: want to test accuracy, speed, and temperature - what is best program? Maybe benchmark it as well * NOTE: just like Linus pointed out in one of his videos, the U14S and rtx 2080ti are REALLY close to each other in the case (I can't even remove the GPU without taking off the cooler fans because I can't reach the PCIE locking clip) - I'd like to see if this causes a performance issue. * NOTE2: the card feels very hot when it's just showing a static 4K BIOS screen - but the fans aren't all the way spun up - I'd like to see if it's within its specs SSDs: I have 2 M.2 SSDs mounted to the board and 1 mounted to the DIMM.2 riser - I'd like to test the throughput and temperatures of all 3 separately. RAM: 256GB @ 3600 MHz (I'm about to kick off memtest for 24 hours, anything else?) Hard Disks: I put 2 WD Black 6 TB drives in an external enclosure w/ a 10 Gbps USB-c interface. I've never stress tested hard drives before. Should I? How would I? Any thoughts on programs, procedures, schedules, etc. would be much appreciated. I am willing to spend many days on this because I plan to own this PC for a very long time. I used my last one for 10 years. I specced it for video editing + extreme futureproofing Thanks!
  8. Yup...works now with all 3 power cables connected. All 256 GB of RAM running @ 3600 MHz, all 3 SSDs show up in BIOS. I ran out of thermal paste mid-application so one side of the cpu is a little thinner. Maybe it will spread out favorably. I will order more paste just in case. For now, idle temp is 49c, though I don't know the individual core temps, maybe that one side of the cpu is hotter. I'm going to run memtest for like 24 hours, and then install windows and test the cpu core temps & accuracy. What is the best tool for that these days? When I got my last PC 10 years ago it was prime95 but now I hear many derogatory things about that.
  9. omg I think I might know the problem....if true then on the one hand I feel stupid but on the other hand, I was just following what the manuals said and the manuals shouldn't assume you're an experienced system builder. So I noticed there was a persistent red LED that did not correspond to the QLED in the manual. I found that it corresponded to extra 8-pin power. The evga psu manual said the 24-bit pin was the motherboard power and the 8-pin cpu power was "optional: only if you will do EXTREME OVERCLOCKING or use TWO CPUs." I took that to heart and thought since I was using 1 CPU and not overclocking at all, this connection would do nothing. The Asus manual also did not mention needing this - just shows a bunch of power connections and says basically "here are the power connections - use what you want." So I decided since they went through the trouble of putting a red LED for that, maybe it was needed after all. So I connected it to board #2, where my CPU is right now, and it went past that step and showed a bunch of other things on the OLED screen, stopping at HDD (which makes sense since I don't have any storage of any kind mounted right now). I'm going to move the CPU back to board #3 and see what happens. I'll test the RAM and GPU on board #2 first. I would also point out that the error message "Memory Test" is wildly inaccurate and misleading for what the problem was. Why couldn't the OLED just say "no CPU voltage" or something? And why does the evga manual flat out lie to me? haha
  10. ok I think the PSU voltages are fine after further research. I put the CPU, cooler, and RAM on board #2 and plugged in the PSU to it. I got the exact same result as board #3. I also updated the BIOS just to make sure it was an equal test, and got the same result before and after the update. I also tested the voltages on board #2 and got pretty much the same results as board #3 to within a mV or so. I did not bother plugging in the GPU because I do not believe the system is going far enough into the sequence to care about that. I can try it just for the hell of it but I'm pretty sure it won't make a difference. So I guess at this point I don't know much of anything...because both motherboards could be bad, or the cpu or ram could be bad, or both, or one motherboard is bad plus one of the other things.
  11. Oh nevermind that last part...I have to reverse it because it's from the perspective of the board. Duh. Edit: I take it back - the tester from evga is shorting the 3rd and 4th pin while eatxpwr guides online say to short 4 and 5. The PSU and mobo are using different pinouts for power.
  12. I was about to pull everything off to put in board #2 and then I thought, why not test the voltages on the power supply connector. The voltages on the pins (with the plastic tester clip attached) match the chart for version 2.0 at https://www.smpspowersupply.com/connectors-pinouts.html However, in the manual for the board, it shows a totally different pinout for 24-pin power (EATXPWR). It didn't even occur to me that such an expensive PSU would have a totally different 24-pin connector as I thought PSUs were pretty standardized by this point. I don't see anything in the mobo manual or psu manual about this situation - is it a totally incompatible supply or is there some kind of adapter, jumper, extra cable, etc. I need to connect? This is the PSU: https://www.newegg.com/evga-supernova-220-p2-1200-x1-1200w/p/N82E16817438029?Item=9SIA0ZX4R45455
  13. I have confirmed that VCORE is not shorted to ground - just not outputting any voltage (or 1 mV, which might just be the margin of error of the meter)
  14. Actually no - MB_SOC is 0.89. I had it written correctly on paper but transcribed it on here wrong. Post edited
  15. Since I know I'm not keeping board #2, when I test the CPU on board #2, is it ok to just leave the current thermal paste on the cpu and cooler? I know if it were a long-term seating that would be bad, but for this test it can't make that huge a difference, right?
×