|
ijyt posted:I’m not even sure how I managed my new build at RRP to be honest. Teach us your ways sensei.
|
#
?
Dec 14, 2020 16:27
|
|
|
|
| # ? Apr 24, 2024 11:47 |
|
|
my bh 5900x is being delivered today! salvation at last
|
|
#
?
Dec 14, 2020 18:34
|
|
|
CommieGIR posted:AMD just benchmarked their 128 core Epyc: https://www.notebookcheck.net/128-core-AMD-EPYC-Milan-server-CPU-scores-an-eye-watering-87878-on-CineBench-R23.509223.0.html There are no 128 core Epycs, that's a benchmark of a server with two 64 core CPUs. The actual improvement versus the last gen is only around 10% for multithreaded workloads, though single threaded ones benefit more.
|
|
#
?
Dec 14, 2020 21:41
|
|
|
ConanTheLibrarian posted:There are no 128 core Epycs, that's a benchmark of a server with two 64 core CPUs. The actual improvement versus the last gen is only around 10% for multithreaded workloads, though single threaded ones benefit more. True, I stand corrected.
|
|
#
?
Dec 14, 2020 21:42
|
|
|
repiv posted:tomshardware benchmarked the cyberpunk ryzen SMT fix OK, that was not "one weird trick", that was actually really important. It should clearly make its way in to a patch, maybe with a check to enable it on the lower count chips vv Yeah, it doesn't seem to be a panacea, but it's something definitely worth investigating HalloKitty fucked around with this message at 23:11 on Dec 14, 2020 |
|
#
?
Dec 14, 2020 22:06
|
|
|
HalloKitty posted:OK, that was not "one weird trick", that was actually really important. It should clearly make its way in to a patch, maybe with a check to enable it on the lower count chips orcane fucked around with this message at 23:09 on Dec 14, 2020 |
|
#
?
Dec 14, 2020 23:03
|
|
|
orcane posted:On the other hand in Computerbase's test the 8+ core Ryzen 5000s lost 5-10% in average FPS and <5% minimum FPS with the "fix", while the 5600X gained 10% FPS but lost 15% minimum FPS. CPU usage went up for all of them but the game doesn't actually run better (the 5600X will be noticeably worse). So maybe it's not as cut and dry. I didn't want to crap up this thread, but I was really surprised that the game would be able to effectively load the extra 8 threads that SMT on an 8-core CPU gets you. You run into increasing synchronization overhead as thread counts go up.
|
|
#
?
Dec 14, 2020 23:10
|
|
|
orcane posted:On the other hand in Computerbase's test the 8+ core Ryzen 5000s lost 5-10% in average FPS and minimum FPS went down by a few percent with the "fix", while the 5600X gained 10% average FPS but minimum FPS dropped by 15%. CPU usage went up for all of them but the game doesn't actually run better (the 5600X will be noticeably worse). So maybe it's not as cut and dry. The curious bit is that Tom's Hardware found that the 5600X benefited from the hack in all situations, both in average and min frames (albeit basically by 0% at 4k because of being GPU limited). I wonder if Computerbase was using different settings for them to end up getting such opposite results?
|
|
#
?
Dec 14, 2020 23:13
|
|
|
I'm sure this thread has covered it but holy cow the 5900x runs hot. Playing with a few EVE online clients, CPU runs at 85C. This is on a EK water block. I think I was getting low 60s on a 3600x Edit Running 9 fans on full with 100% pump brings it down to 73 lmao MeruFM fucked around with this message at 00:30 on Dec 15, 2020 |
|
#
?
Dec 15, 2020 00:22
|
|
|
Shades of Intel. I guess when you pump the freq you get heat no matter 7nm.
|
|
#
?
Dec 15, 2020 03:27
|
|
|
Its at the point where liquid metal makes sense to start using, even more so with zen and the chiplets having such a small surface area to move heat through
|
|
#
?
Dec 15, 2020 03:47
|
|
|
overclocking my 5600x with PBO+ 200mhz offset to 4.8ghz with curve optimizer to undervolt the worst cores has me sitting pretty cool at around 79-81c at max AVX2 extreme testing with air-cooling (FUMA2).c urrently sitting around 4700/630 CB20 scores which is good and good enough tempwise. granted that's 1 CCX as opposed to 2 and there's a reason the 5600x has been outperforming 5800/5900/5950 in gaming benchmarks because thermals are less favorable there and just extra heat generation without much being used up in multi-threading (which games suck at using). i think a problem with liquid metal paste is that it'll etch off the engraving for warranty but if you don't care, go for it. from what I understand of people who're using liquid cooling, they get about the same temps as I do with air-cooling. i think something about the chips is just generating a high heatspot that no matter the cooling system won't really wick away and there's only so fast it can transfer heat out which is why no matter what's being used seems people are hitting around ~80c under max load. Xaris fucked around with this message at 04:05 on Dec 15, 2020 |
|
#
?
Dec 15, 2020 04:00
|
|
|
Xaris posted:overclocking my 5600x with PBO+ 200mhz offset to 4.8ghz with curve optimizer to undervolt the worst cores has me sitting pretty cool at around 79-81c at max AVX2 extreme testing with air-cooling (FUMA2).c urrently sitting around 4700/630 CB20 scores which is good and good enough tempwise. granted that's 1 CCX as opposed to 2 and there's a reason the 5600x has been outperforming 5800/5900/5950 in gaming benchmarks because thermals are less favorable there and just extra heat generation without much being used up in multi-threading (which games suck at using). i think a problem with liquid metal paste is that i,t'll etch off the engraving for warranty but if you don't care, go for it. Well its more the fact that the 5600x/5800x are a single CCX so your not suffering the IF latency penalties that the 5950x/5900x will have by juggling workloads between the CCX's Also who bought me this custom text, show yourself coward thats hilarious
|
|
#
?
Dec 15, 2020 06:51
|
|
|
MeruFM posted:I'm sure this thread has covered it but holy cow the 5900x runs hot. That sounds... Pretty high. Granted the 3600x is pulling much less power but still. How much radiator space is that with? Is the CPU the only part in the loop? What's your coolant temp sitting at under that load? Are those CPU temps from ryzen master or another program?
|
|
#
?
Dec 15, 2020 14:53
|
|
|
amazon has had the 5800X in stock for like the past hour if anyone's looking for one https://www.amazon.com/dp/B0815XFSGK/
|
|
#
?
Dec 15, 2020 15:29
|
|
|
pmchem posted:amazon has had the 5800X in stock for like the past hour if anyone's looking for one Amazon keeps showing it can't ship to my location or one of their lockers. I'm assuming this is some manifestation of it being out of stock but the page not being updated yet. e: finally worked after deleting some old addresses. So strange. doomisland fucked around with this message at 15:54 on Dec 15, 2020 |
|
#
?
Dec 15, 2020 15:51
|
|
|
MeruFM posted:I'm sure this thread has covered it but holy cow the 5900x runs hot. So weird, you're 10c cooler then my friend using an nhd15s with two fans. (5900x)
|
|
#
?
Dec 15, 2020 15:57
|
|
|
Cambridge MA Microcenter shows 14 5600x in stock
|
|
#
?
Dec 15, 2020 17:28
|
|
|
VorpalFish posted:That sounds... Pretty high. Granted the 3600x is pulling much less power but still. How much radiator space is that with? Is the CPU the only part in the loop? What's your coolant temp sitting at under that load? They're HWinfo numbers, CPU Die (average). The CCD1 and CCD2 numbers are 10-15 degrees lower but the Tctl number is slightly higher. I don't use Ryzen master because I need the virtualization. I have 720mm radiator space and my coolant sits at 42C, but I do run my fans and pump at 25% until coolant hits 45C, that's why I also mentioned the 100% everything temps. It does go through a 3080 first though, but that shouldn't make a huge diff. I remember the cpu being only a few degrees cooler when I didn't have the 3080 yet. GRINDCORE MEGGIDO posted:So weird, you're 10c cooler then my friend using an nhd15s with two fans. (5900x) I guess that's fine and sounds about right. Water should be cooler than air right?
|
|
#
?
Dec 15, 2020 18:56
|
|
|
MeruFM posted:They're HWinfo numbers, CPU Die (average). The CCD1 and CCD2 numbers are 10-15 degrees lower but the Tctl number is slightly higher. I don't use Ryzen master because I need the virtualization. 42c for coolant honestly sounds pretty high for 720mm of rad space and ~500w of heat load unless ambient is pretty high as well tbh but I'm not gonna claim to be any kind of water cooling expert, and I guess 25% can mean very different things depending on what fans you're using.
|
|
#
?
Dec 15, 2020 19:49
|
|
|
VorpalFish posted:42c for coolant honestly sounds pretty high for 720mm of rad space and ~500w of heat load unless ambient is pretty high as well tbh but I'm not gonna claim to be any kind of water cooling expert, and I guess 25% can mean very different things depending on what fans you're using. Ambient here is around 25c. But I have 3 noctua intakes and 6 generic fans out. Maybe something's just constantly exploding and I don't realize. Or maybe the water temp monitor is wildly inaccurate. This is my 2nd water system but the 1st was maybe 12 years ago so I don't have much experience on water. I do set my fans to the point where I cannot hear them and subsequently get angry at any electrical noise coming out of the mobo/gpu. Anyways, I guess zen 3s are just hot af. edit: just stuck a meat thermometer into my water  it's accurate within 2 degrees so shrug
MeruFM fucked around with this message at 23:00 on Dec 15, 2020 |
|
#
?
Dec 15, 2020 22:52
|
|
|
42C doesn't sound all that high to me with fans set real low if that's a benchmark/synth load temp--that's still a delta T under 20C. Tons of rad space only does so much if you're barely running the fans and keeping pump speeds low to reduce the noise. As a general rule you want slightly more intake airflow than exhaust airflow in order to help keep dust out--3 v 6 is pretty unbalanced.
|
|
#
?
Dec 15, 2020 23:16
|
|
|
A nearly 20C delta between ambient and the coolant on 720mm worth of radiator is not that great even for 500w of load but if the fans are <500 RPM it is probably not too unusual. I've got 375w worth of load on a 420mm radiator (3x140) and a ~15C delta is about the limit where anything short of stopping the fans entirely dumps all the heat before it can get any warmer though, but bigger fans. You should be able to feel heat just pouring out of the radiators with that kind of delta. That being said, the heat spreader on modern CPUs is basically an insulator. The tiny size of the die combined with the high power consumption means they all run hot and there isn't much that can be done about it short of popping the spreader off and doing direct die cooling. CPUs regularly approaching 90C even on exceptionally good cooling is not uncommon or even a sign of a problem as long as they aren't throttling right away by a huge amount. As long as they sustain the performance/power output, your cooling is working normally and you shouldn't worry about it.
|
|
#
?
Dec 15, 2020 23:19
|
|
|
MeruFM posted:Anyways, I guess zen 3s are just hot af. I don't know about the 12 and 16 core parts, but I remember HW unboxed found the 5800x was on average 3-5 degrees hotter than the 3700x for the same power draw.
|
|
#
?
Dec 15, 2020 23:40
|
|
|
DrDork posted:Try taking both RAM sticks out and putting back only one of them, and then letting it turn on and run for 5+ minutes. Some of these boards seem real wonky with RAM, and forcing a re-detection of it like that sometimes helps. OK this was the correct answer, this motherboard model is beyond insane about ram. I bought a second one of the X470D4U2-2T motherboards, then: 1) put in the 3000-series CPU with a single stick of ram in slot A1, and waited... and waited... by default this loving thing takes several minutes to show anything whatsoever over VGA. Finally made it to the BIOS though! 2) Flashed the 5000-series beta BIOS, and after I enabled fast boot and disabled "wait for BMC" whatever that is, which shaves minutes off the boot and brings it down to around 10 seconds. After that I put in two sticks of 32GB in slots B1 and B2 (says this is an allowed configuration as "Priority 2" whatever that means) and left the two slots closer to the CPU cooler empty 3) Now it makes it all the way to Windows! Except CPU-Z says that the ram (32GB sticks of Samsung non-ECC sodimm) are all running at 1199.7 mhz... uh? 4) The BIOS has settings for increasing RAM overclock all the way up to 4000mhz. I tried turning it up to 2000mhz and it boots. I tried turning it up further to the 3200 which the RAM is rated for and... after existing BIOS it can no longer boot. It can't even boot to BIOS! To get anything whatsoever I have to power it off and pull the BIOS coin battery to get rid of the overclock, seems to be the only way to get back to BIOS and booting. I don't seem to see anything about Intel XMP in the BIOS though, yet it lets me set frequency/timing/voltage on the ram and the infinity fabric. This motherboard is distilled ASRock jank. I think my favorite part is that it's a server motherboard and has out-of-band management, yet it has a BIOS toggle for LN2 mode... ok then.
|
|
#
?
Dec 16, 2020 00:27
|
|
|
Zero VGS posted:OK this was the correct answer, this motherboard model is beyond insane about ram. What voltage settings do you have? Normally I would right away say try setting 3200 mhz on the ddr frequency, 1600 on memory clock, 1600 on infinity fabric, set VSOC voltage to 1.05V (may need to go to 1.1V) and DRAM voltage to 1.35V (may need to go to 1.4) and see what happens. May also need to adjust VDDG voltages, Zen3 seems to prefer lower VDDG/P (im using 900mV on VDDP and 950mVs on the VDDG CCD/IOD) but very hungry on the VSOC (i'm personally running 1.138v on VSOC). However, I'm not sure if SO-DIMM behaves any differently. May need to up your VSOC/DRAM Voltage a bit in increments. I'm not sure what stock is on that board but I would guess you could at least start with 1.0V/1.3V safely and add +0.05V increments. Xaris fucked around with this message at 00:42 on Dec 16, 2020 |
|
#
?
Dec 16, 2020 00:37
|
|
|
Zero VGS posted:OK this was the correct answer, this motherboard model is beyond insane about ram. Glad it helped! CPU-Z reports the DRAM's internal clock, which for DDR4 gets doubled when you're talking the number on the box. So 1200Mhz x 2 = 2400, which incidentally is a common JEDEC spec for DDR4. This is the value you should expect it to fall back to every time you clear CMOS / change the RAM up. So to hit 3200 you'd be looking for CPU-Z to report 1600Mhz. I've had similar experiences as you with other ASRock boards, where pushing RAM too far results in an unbootable board requiring an aggressive CMOS clear, despite it purportedly having a "self test / self recovery" ability. A quick google suggests that the board does not support XMP profiles, but not all is lost. If you look in CPU-Z's Memory tab, it should have listings of XMP values. Try giving those a shot.
|
|
#
?
Dec 16, 2020 01:00
|
|
|
Indiana_Krom posted:A nearly 20C delta between ambient and the coolant on 720mm worth of radiator is not that great even for 500w of load but if the fans are <500 RPM it is probably not too unusual. I've got 375w worth of load on a 420mm radiator (3x140) and a ~15C delta is about the limit where anything short of stopping the fans entirely dumps all the heat before it can get any warmer though, but bigger fans. You should be able to feel heat just pouring out of the radiators with that kind of delta. I mean lowering the coolant temp will bring the die temp down, but you're right it's probably nothing to worry about. I guess I was expecting 10-12 above ambient just cause they have such a huge amount of rad space. I also sometimes forget it's not winter everywhere, I was expecting lower ambient temps as well.
|
|
#
?
Dec 16, 2020 01:19
|
|
|
DrDork posted:Glad it helped! LOL ok when I was trying to do test 3200mhz I was actually requesting 6400mhz then, that explains it. I set both the ram and infinity fabric clock to 1600 and now it successfully boots to desktop. As you said, CPU-Z shows 1600mhz but Task Manager shows ram is in fact 3200 (it seemed to automatically change the cas latency to 22-22-22) I ran Prime95 blend and it ran fine for a few minutes, but the second I exited it got the classic "WHEA UNCORRECTABLE ERROR", maybe that was just because the system didn't know how to reallocate the 8GB of ram that Prime95 was hoarding. Super bogus that you can't even get back to the BIOS after going too far on RAM. The coin battery "lock" gets blocked by the PSU cable when inserted so I have to unplug that and then the coin battery all just to test a new clock. I'll see if there's maybe a jumper somewhere to clear CMOS. Still, this is a much better position than I've been. The coolest part is the motherboard actually has a built-in RAMDISK setting. It actually lets you drill down into the M.2 SSD directory and find a single file to copy to a ramdisk upon boot. That was one of the main things I wanted to do with this PC is have a 30GB ramdisk of the content package file for Path of Exile because it is notorious for hitching when loading assets, even with an NVME drive.
|
|
#
?
Dec 16, 2020 02:08
|
|
|
Zero VGS posted:LOL ok when I was trying to do test 3200mhz I was actually requesting 6400mhz then, that explains it. You can use like a paper clip or screwbridge to bridge the CMOS pins. just anything conductive. probably even a penny. WHEA can be a lot of things, could be not getting enough voltage. You'll have to manually change your CAS timings as well.
|
|
#
?
Dec 16, 2020 03:17
|
|
|
Xaris posted:You can use like a paper clip or screwbridge to bridge the CMOS pins. just anything conductive. probably even a penny. The manual says there's a CMOS reset pad but that you still need to remove the battery anyway Waiting 5 seconds after just removing the battery works fine for me so the pad seems pointless unless the manual is lying and it works without battery removal. Speaking of voltage, I noticed under load that the 5800x is naturally going above 1.4vcore. Isn't that like, really high? I thought going above 1.4 on an Intel chip is like accelerated electromigration damage zone? Another weird thing is that the CPU hits 75c with Prime95 blend which is supposed to be an unrealistically tough stress test, yet if I load up Dark Souls 3 with no discrete GPU it hits 91c and stays there. I figured that makes sense if it has to use integrated graphics cores, except it apparently doesn't have integrated graphics! (motherboard has a built in VGA chip since it is server console grade)
|
|
#
?
Dec 16, 2020 06:27
|
|
|
DrDork posted:As a general rule you want slightly more intake airflow than exhaust airflow in order to help keep dust out--3 v 6 is pretty unbalanced. Yep, that's why the noctuas are intakes, and I run them slightly faster for the coveted positive pressure. Tested with state of the art toilet paper streamers. The heat coming out of the radiators feel like a space heater on low. My cats like to lie on the shelving because of that, where the side mount blows.  I did a prime95 bench and it never went past 78C, pure CPU ofc. Maybe space mining simulators just push systems to their limit. MeruFM fucked around with this message at 06:37 on Dec 16, 2020 |
|
#
?
Dec 16, 2020 06:34
|
|
|
Dramicus posted:I don't know about the 12 and 16 core parts, but I remember HW unboxed found the 5800x was on average 3-5 degrees hotter than the 3700x for the same power draw. I have both the 5800X and the 5950X, the latter runs 20-25C cooler with the same cooling setup  .
|
|
#
?
Dec 16, 2020 06:35
|
|
|
MeruFM posted:The heat coming out of the radiators feel like a space heater on low. My cats like to lie on the shelving because of that, where the side mount blows. working as intended
|
|
#
?
Dec 16, 2020 06:40
|
|
|
MaxxBot posted:I have both the 5800X and the 5950X, the latter runs 20-25C cooler with the same cooling setup The 5800X has the same power limit at half the die area. Goons have had good success limiting ppt to ~120W without losing any performance and you can probably squeeze out even more efficiency with PBO 2.
|
|
#
?
Dec 16, 2020 08:06
|
|
|
MaxxBot posted:I have both the 5800X and the 5950X, the latter runs 20-25C cooler with the same cooling setup My 5800x is air-cooled with an nh-d15 and it never goes above 74 under normal loads. YMMV.
|
|
#
?
Dec 16, 2020 10:08
|
|
|
https://www.antonline.com/AMD/Computers/Electronic_Components/Microprocessors/1414834 5600 in stock. just like the 5800 they had before there appears to be a lot of stock.
|
|
#
?
Dec 16, 2020 18:39
|
|
|
Sorry noob question, what programs do people recommend to bench their ryzen cpus? Want to see if it's boosting correctly, temps, etc.
|
|
#
?
Dec 16, 2020 18:46
|
|
|
What 32 GB ram kit should I get for a 5600x? No RGB, thx.
|
|
#
?
Dec 16, 2020 19:10
|
|
|
|
| # ? Apr 24, 2024 11:47 |
|
|
Fauxtool posted:https://www.antonline.com/AMD/Computers/Electronic_Components/Microprocessors/1414834 What is this retailer? Is this a legit site?
|
|
#
?
Dec 16, 2020 19:12
|
|
