|
PirateBob posted:I'm not sure if I placed the thermal paste optimally on my 5600x + Scythe mugen 5 rev b. I put a decent sized blob in the middle, but when I placed the heatsink it was pushing down more on one side than the other. Seems pretty normal to me. A lot of these random thermal tests you find online are done on open benches under ideal conditions, and not really indicative of what you should expect in real life. You can mess around in AMD Curve Optimizer for better clock speeds at the same voltages but you're not harming your machine in any way at those level.
|
#
?
Feb 5, 2021 18:09
|
|
|
|
| # ? May 17, 2024 00:02 |
|
|
Some Goon posted:Dunno in terms of vs what it could be, but those temps are absolutely fine. Okay. Just realized a part of my post could be misunderstood: I meant to say that the heatsink was pushing down one one side *during the installation*, while getting the first screw in. Then it evened out after the other screw was in. I was just a bit concerned about whether the paste was squished over to the other side during that process.
|
|
#
?
Feb 5, 2021 18:09
|
|
|
PirateBob posted:Okay. Just realized a part of my post could be misunderstood: I meant to say that the heatsink was pushing down one one side *during the installation*, while getting the first screw in. Then it evened out after the other screw was in. I was just a bit concerned about whether the paste was squished over to the other side during that process. Everybody does, it's fine. It should more-or-less even out once it's all the way on, unless it squirted out from under the heatsink entirely.
|
|
#
?
Feb 5, 2021 18:59
|
|
|
Some Goon posted:Everybody does, it's fine. It should more-or-less even out once it's all the way on, unless it squirted out from under the heatsink entirely. ...and if you're really, really worried, just pop the HSF off for a second. I'm willing to bet that what you would find is a very evenly-squished layer of thermal interface. I may have done this myself once or twice.
|
|
#
?
Feb 5, 2021 20:00
|
|
|
On this topic, how do people feel about using a thermal pad rather than paste (with a 5600X, if it matters)? I'm going to try to complete my build this weekend, now that my last part just arrived on my doorstep, and bought one with the intent of using it, but now I'm getting cold feet about not using paste with a new processor that apparently runs hotter than others.
|
|
#
?
Feb 5, 2021 20:12
|
|
|
Sarcastro posted:On this topic, how do people feel about using a thermal pad rather than paste (with a 5600X, if it matters)? I'm going to try to complete my build this weekend, now that my last part just arrived on my doorstep, and bought one with the intent of using it, but now I'm getting cold feet about not using paste with a new processor that apparently runs hotter than others. Unless I had a compelling reason specifically to use a pad (sealed inside an industrial thing or something with no mounting points like a SBC) I wouldn't. Even the best pads don't come anywhere near the performance of paste.
|
|
#
?
Feb 5, 2021 20:23
|
|
|
Thermal pads are worse in every way except longevity But a 5600x only uses like 95W max so it probably doesn't really matter. I think it's like a 2 degree difference at that wattage
|
|
#
?
Feb 5, 2021 20:25
|
|
|
It occurs to me that the very fact of using a pad means that if I don't like how it's going after testing out the new build for a time it'll be fairly easy to switch to paste. I'm not planning on overclocking, so might as well try it to start and see how it goes.
|
|
#
?
Feb 5, 2021 20:46
|
|
|
redeyes posted:Which processor? RAM is the most likely thing. I found that Corsair 3200 is being a bitch with these boards and dropping to 2133 doesnt help. I was a 2600X, and while I was trying to guide them to proper troubleshooting someone else had them messing around with CPU voltage. :facepalm: Inept posted:Mine has been ok. Get them to change their XMP setting to auto, which should revert back to JDEC memory speeds. But if they don't want to answer your questions or respond, just let them deal with it. I tried getting them to check memory settings but from their most recent post I think they gave up and are going back to the store at this point.
|
|
#
?
Feb 5, 2021 22:46
|
|
|
There's a non-zero chance this is accurate:
|
|
#
?
Feb 6, 2021 03:17
|
|
|
Seems pretty accurate to me.
|
|
#
?
Feb 6, 2021 04:03
|
|
|
I ended up buying a Comet Lake i5 over a 3600 after the dealer gave me a really good deal on a CPU/board combo, but after trying to put it together this morning the system would get stuck in a boot loop. I brought it back to the store and they spent some more time swapping out RAM sticks and even tried with a completely new motherboard from a new box and it plain just wouldn't work, so I sent both back for replacement. Too bad. Anyway I ended up doing a full teardown in the process of swapping back to the 2400G and cleaned out all the fans, the mesh filters, the cooler fin stack, repasted the CPU, and that stuff, and it ended up dropping my temps by a good 10 degrees. I guess what I'm saying is clean your computers regularly.
|
|
#
?
Feb 6, 2021 15:24
|
|
|
Any suggestions to try before I go through the trouble of swapping processors? Have a 5600X on a X570 Aorus Elite w/ latest F33a (1.2.0.0) bios that hangs memtest86+ (the one included in most ubuntu usbs these days) in single core mode. Hangs on Test #2 in the 4G-6G address range, both for 3600 16G sticks and my old known good 3200 8G sticks, with a single stick of either size in either channel. Tried both XMP and stock w/ 1.35v settings. I tried locking the cpu at 3.7Ghz, didn't make any difference. Thankfully I have a 2700 in a B450 Aorus Elite so I can do some swapping, but I'd appreciate any tips on things to test/check before I start having to redo all my HSF paste and update my old board's bios for 5000 series. The fact that it's clearly the same address regardless of channel A or B, and regardless of memory size makes me really suspicious of the CPU rather than the motherboard, but maybe I'm overthinking it. I've heard a lot more reports of bad Ryzen 5000 series chips than I think I've ever heard of bad CPU in my life though. Rescue Toaster fucked around with this message at 23:25 on Feb 7, 2021 |
|
#
?
Feb 7, 2021 23:16
|
|
|
Rescue Toaster posted:Any suggestions to try before I go through the trouble of swapping processors? Have you tried rolling back the BIOS? I've had bad experiences with some Gigabyte beta BIOS (I have an Aorus Elite X570), and have had issues resolve by just going to a previous version. It looks F32 supports Ryzen 3, so I think I'd start there. Was the machine running stably before, or has all troubleshooting so far been on F33a?
|
|
#
?
Feb 7, 2021 23:25
|
|
|
CaptainSarcastic posted:Have you tried rolling back the BIOS? I've had bad experiences with some Gigabyte beta BIOS (I have an Aorus Elite X570), and have had issues resolve by just going to a previous version. It looks F32 supports Ryzen 3, so I think I'd start there. I just put it together, so no experience with it at all yet. It's certainly easier to q-flash a bios than swap CPUs so I'll try that first, thanks.
|
|
#
?
Feb 7, 2021 23:26
|
|
|
Zen/Zen+ CPU fun and weirdness from the past couple weeks: As I mentioned earlier, I swapped out my desktop's long-serving Ryzen 1600 (which had started throwing MCEs and rebooting of its own volition) for a 3600. The system has been perfectly stable since then -- which wouldn't be news, except that I had previously been convinced that I had an early production run RX 5700 GPU which was suffering from the problems that those are now notorious for. All the symptoms matched what people were reporting and diagnosing as problems with early RDNA1 cards -- but this machine has now been up for a week and I've been browsing, playing Grim Dawn and E:D, and running Docker containers with zero problems One of my compute nodes is still running a Ryzen 2700 (for now). It had some stability issues after it got upgraded with hand-me-down RAM from one of my 3900X nodes, but I pretty quickly remembered that the 2700 was definitely not qual'd for 3000MHz RAM (2400MHz was "the standard" when I bought that CPU). I turned it down to 2800MHz, and things seemed stable. Over the span of 3 to 5 days, however, the CPU would begin halting threads of execution. At boot it would be running 16 tasks. Four days later it might only be simultaneously running 15 or 14. Another week later and it would be down to maybe 10. This was something I had never seen before, and reboots brought it right back to 16 threads every time, but that's obviously not a satisfactory resolution. Digging into the logs turned up some really funky memory overrun errors of a type I was not familiar with. On a hunch, I slowed the RAM down again, to 2666MHz, and all my threads have been safe and sound so far. That machine is about to be rebuilt, so the problem is going away forever, but I thought it was a weirdness worth sharing. mdxi fucked around with this message at 23:44 on Feb 7, 2021 |
|
#
?
Feb 7, 2021 23:39
|
|
|
mdxi posted:Zen/Zen+ CPU fun and weirdness from the past couple weeks: I'm starting to think that being an early adopter on Zen 1 was an increasingly raw deal, since I have the exact same problems and keep having to drop memory timings to even boot, but I can't justify spending dosh to get off of this 1600 with how the market is, and I won't get onto a dead-end socket. I think I'm down to 2800 from 3200 now.
|
|
#
?
Feb 8, 2021 00:19
|
|
|
CaptainSarcastic posted:Have you tried rolling back the BIOS? I've had bad experiences with some Gigabyte beta BIOS (I have an Aorus Elite X570), and have had issues resolve by just going to a previous version. It looks F32 supports Ryzen 3, so I think I'd start there. Tried both F32 and F30 and memtest hangs in the exact same place. I guess I'll be ripping out some CPUs later. Yuck. EDIT: memtest86+ was suddenly failing in the exact same way on my old B450 + 2700 board... so somehow the particular version of it on my linux usb stick suddenly crapped the bed? I'm sure it was the same one I ran previously on the same B450/2700 machine and worked fine. Really bizarre. I got a copy of Passmark's memtest86 v8 and it's running now, at least it didn't hang right away. I did take my X570 back to F32 for now as well. Fingers crossed. Rescue Toaster fucked around with this message at 02:32 on Feb 8, 2021 |
|
#
?
Feb 8, 2021 00:23
|
|
|
SwissArmyDruid posted:I'm starting to think that being an early adopter on Zen 1 was an increasingly raw deal, since I have the exact same problems and keep having to drop memory timings to even boot, but I can't justify spending dosh to get off of this 1600 with how the market is, and I won't get onto a dead-end socket. I got a 1700 within a couple months of launch in 2017 and it's been chugging along fine this whole time at 2933 CL16 on some ECC B-die binned for 2400. Then again I think I got good silicon because it's rock solid stable at 4.0GHz on 1.35v. Too bad your 1600 is probably out of warranty at this point or I'd say try RMAing it. Have you had the SOC overvolted pretty high for those memory clocks? Mine needed 1.1v to get to 2933 but that's in the safe range as far as I have read.
|
|
#
?
Feb 8, 2021 02:38
|
|
|
https://twitter.com/9550pro/status/1358591672758661121 So Cezanne (Zen 3 APU) for desktop exists... but now we get to see if it will ever get a retail launch like Renoir never did.
|
|
#
?
Feb 8, 2021 03:06
|
|
|
SwissArmyDruid posted:I'm starting to think that being an early adopter on Zen 1 was an increasingly raw deal, since I have the exact same problems and keep having to drop memory timings to even boot, but I can't justify spending dosh to get off of this 1600 with how the market is, and I won't get onto a dead-end socket. They have so much of it, and several of them are so bad, a lot of people just gave up fixing the issues in software, because AMD decided not to fix them in hardware. Cygni posted:https://twitter.com/9550pro/status/1358591672758661121 
|
|
#
?
Feb 8, 2021 05:54
|
|
|
BlankSystemDaemon posted:Don't look up my posts about Zen1 errata in this thread, then. Kaveri and Bulldozer CPUs are still cheap on the second-hand market 
|
|
#
?
Feb 8, 2021 06:32
|
|
|
SamDabbers posted:I got a 1700 within a couple months of launch in 2017 and it's been chugging along fine this whole time at 2933 CL16 on some ECC B-die binned for 2400. Then again I think I got good silicon because it's rock solid stable at 4.0GHz on 1.35v. Yeah I just got a used 1800X and the memory setup was much smoother than expected, I just put in the 3200MHz "safe" settings on the Ryzen calc and it was stable on the first try.
|
|
#
?
Feb 8, 2021 07:38
|
|
|
My last follow up on this because I thought it was interesting. The memtest86+ 5.01 built for the debian/*buntu/mint family of distros of 2020 vintage (ubuntu 20.04, mint 20.1, etc..) appears to be faulty due to something in the toolchain. Older versions of those distros with the same 5.01 source have working memtest. Get your own memtest stick, don't use the one from those live usb distros. https://bugs.launchpad.net/ubuntu/+source/memtest86+/+bug/1876157
|
|
#
?
Feb 8, 2021 15:24
|
|
|
CaptainSarcastic posted:Any known gotchas with Asrock B450m Pro boards? I have an acquaintance dealing with a lot of crashing, and I'm gently trying to guide them toward troubleshooting. I'm guessing RAM, but trying to get them to respond to questions is difficult. Any tips or tricks I can pass along? Make sure they're putting the ram in the right slots. The board should have come with a printout for how to load it up depending on the number of dimms and and number of sticks. I was personally getting failure to boot and slow ram speeds until a took a closer look at that printout and re-positioned the sticks, and that fixed everything.
|
|
#
?
Feb 8, 2021 16:14
|
|
|
Yay, I finally found some affordable DDR4-3200 ECC UDIMMs with Micron E-Die. The Kingston KSM32ES8/8ME modules.
|
|
#
?
Feb 8, 2021 22:37
|
|
|
SwissArmyDruid posted:I'm starting to think that being an early adopter on Zen 1 was an increasingly raw deal, since I have the exact same problems and keep having to drop memory timings to even boot, but I can't justify spending dosh to get off of this 1600 with how the market is, and I won't get onto a dead-end socket. Late last month, I pulled half the memory out of my 1700 in order to get the system to stay stable for more than 12 straight hours. Dunno why it degraded so much the last few months, but I'm getting on the fully baked Ryzen so I hopefully won't have to deal with this crashy poo poo anymore and won't have to deal with teething problems of a new arch, socket and memory standard. To that end, I have a General Memory Question for memory knowers in the thread: I got two kits of 2x8GB G-Skill 3600 CL16 Ripjaws (single rank Hynix according to the BIOS) and put them in an ASRock Taichi X570. Both kits ran 8 rounds of Memtest86 just fine at the timings advertised on the packaging individually*. Then I installed both at the same time and got exactly two errors on different tests of different runs. That's weird because I'm used to getting 1000s of errors, consistently, when the part is obviously NFG. I took them out and swapped the banks they were in, effectively also re-seating them, and now they're through ~40 rounds with no errors. Am I being too paranoid in still thinking I should take them back and get a single 2x16 kit I suspect is dual-rank and trying that? I'm just doing a sanity check in case the last three years of random, infrequent crashes that became more frequent recently has me annoyed beyond reason, here. *after a BIOS update - nothing would so much as POST above board default timings before that 😣 Munkeymon fucked around with this message at 00:29 on Feb 9, 2021 |
|
#
?
Feb 8, 2021 22:54
|
|
|
Munkeymon posted:2x8GB G-Skill CL36 Ripjaws That's some slow-rear end memory, dude.
|
|
#
?
Feb 8, 2021 23:04
|
|
|
That DRAM Calculator from 1usmus, is that thing generally reliable? I've been checking what CAS values it spits out for my new upcoming modules, and it goes with CAS 14 at 3200MHz. Would be nice, but seems a bit too good.
|
|
#
?
Feb 8, 2021 23:25
|
|
|
Combat Pretzel posted:That DRAM Calculator from 1usmus, is that thing generally reliable? I've been checking what CAS values it spits out for my new upcoming modules, and it goes with CAS 14 at 3200MHz. Would be nice, but seems a bit too good.
|
|
#
?
Feb 8, 2021 23:29
|
|
|
SwissArmyDruid posted:That's some slow-rear end memory, dude. Conflated 3600 and CL16 
|
|
#
?
Feb 9, 2021 00:44
|
|
|
Combat Pretzel posted:Yay, I finally found some affordable DDR4-3200 ECC UDIMMs with Micron E-Die. The Kingston KSM32ES8/8ME modules. What motherboard are you using? Ryzen ECC support is A Thing and i would be interested to see if it actually corrects injected errors. Been considering replacing my Ivy Bridge HTPC/home server/NAS with Zen 3 + ECC memory.
|
|
#
?
Feb 9, 2021 01:04
|
|
|
B-1.1.7 Bomber posted:There's a non-zero chance this is accurate: Truth in advertising.
|
|
#
?
Feb 9, 2021 05:25
|
|
|
Cygni posted:What motherboard are you using? Ryzen ECC support is A Thing and i would be interested to see if it actually corrects injected errors. Been considering replacing my Ivy Bridge HTPC/home server/NAS with Zen 3 + ECC memory.
|
|
#
?
Feb 9, 2021 06:41
|
|
|
Combat Pretzel posted:That DRAM Calculator from 1usmus, is that thing generally reliable? I've been checking what CAS values it spits out for my new upcoming modules, and it goes with CAS 14 at 3200MHz. Would be nice, but seems a bit too good. Honestly depends on what RAM you have (Samsung b-die for instance). It works best and you will have good success with Samsung b-die, less so with other DIMM manufacturers because it's largely heuristic based on a lot of previous work/anecdotal evidence and some rough calculations of typical timing calculations. Most of this work will have come out of people OC-ing Samsung B-die because it's the most popular. And yes depends on your motherboards VRM, memory topology, CPU and RAM silicon lottery. But there's no reason not to give it a shot. The most foolproof way that works for any type is follow this guide https://github.com/integralfx/MemTestHelper/blob/master/DDR4%20OC%20Guide.md#finding-a-baseline to find a baseline with standard primary timings (16-20-20-40-16) and up your Infinity Fabric and Mem Clock and Memory Multiplier until it's stable/boots, then if it's all good then plop in the secondary timings that Ryzen DRAM Calc spits out and if it's not good then backoff a few important ones like tRFC. Xaris fucked around with this message at 06:52 on Feb 9, 2021 |
|
#
?
Feb 9, 2021 06:49
|
|
|
Combat Pretzel posted:That DRAM Calculator from 1usmus, is that thing generally reliable? I've been checking what CAS values it spits out for my new upcoming modules, and it goes with CAS 14 at 3200MHz. Would be nice, but seems a bit too good. What does the calculator say for 3533? The only real risk is no one tested them and it's estimating off of the bin which could mean it's inaccurate. If you see 3533 c14 then maybe it's not accurate. Post in the thread how it goes once they get there. 3200 c14 seems in the realm of possibility. People were getting similar with zen1 and samsung ecc sticks.
|
|
#
?
Feb 9, 2021 07:05
|
|
|
Khorne posted:How many sticks of RAM at what capacity? I saw micron with ecc. The calculator is less reliable the less common the sticks & configuration are. quote:What does the calculator say for 3533? The only real risk is no one tested them and it's estimating off of the bin which could mean it's inaccurate. —edit: I probably want to run them at some mellower timings like CL16. With what error correction being there for its namesake. If I select the bad bins preset, it gives me something like that. Combat Pretzel fucked around with this message at 08:33 on Feb 9, 2021 |
|
#
?
Feb 9, 2021 08:29
|
|
|
Combat Pretzel posted:Yay, I finally found some affordable DDR4-3200 ECC UDIMMs with Micron E-Die. The Kingston KSM32ES8/8ME modules.
|
|
#
?
Feb 9, 2021 16:25
|
|
|
https://twitter.com/momomo_us/status/1359160668893704197 https://d29b39n1yn0j3z.cloudfront.net/wp-content/uploads/2021/02/CPU-Bracket.mp4
|
|
#
?
Feb 9, 2021 16:29
|
|
|
|
| # ? May 17, 2024 00:02 |
|
|
I like that heatsink mounting mechanism, looks like they copied noctua but used thumbscrews.
|
|
#
?
Feb 9, 2021 17:50
|
|
