|
If the xmp profile is stable (do test it imo, even if it's booting it might be throwing errors occasionaly) 3200cl14 is pretty good for Ryzen. I've fought for hours trying to get past 2933cl14. If you're on the 2700x from a year ago, my understanding is PBO is where you're going to see the best gains, but I don't have any first hand experience with zen+.
|
# ? Jun 20, 2019 18:54 |
|
|
# ? May 15, 2024 04:48 |
|
Alright I just got a 9900k and an MSI Z390 Carbon Pro AC motherboard and I’m screwing around with some overclocking, but it’s a lot less forgiving than my old 6600k was. I’ve got the multiplier set to 50 and voltage set in override mode (not adaptive or anything yet) to 1.33v and I can pass an hour in AIDA64 without crashing but it the CPU package temp gets up to 99c. AIDA shows 1% throttling and hwinfo confirms that it did throttle a few times on a couple of cores but not really by much and not for long at all. That being said I’d like to get the temps under more control. 1.33v is the minimum voltage I’ve had stability with but I’ve never messed with LLC before. Is this scenario something that might be useful for? Lower the voltage but set LLC at a higher level? As for cooling I’m using a Corsair h115i pro 280mm AIO. Also I had 2 exhaust fans, one on the top (140) and one on the back (120) but as a precaution to give the VRM some airflow I turned the top fan around so it blows down so now I have 2 140mm fans at the front of my case blowing through the radiator, one 140mm fan blowing down from the top for the VRMs and one 120mm fan on the back for exhaust I also know it should be a good idea to switch to adaptive voltage at some point but I don’t know how to figure out what I should set the offset to. GutBomb fucked around with this message at 02:26 on Jul 6, 2019 |
# ? Jul 6, 2019 02:19 |
|
GutBomb posted:Alright I just got a 9900k and an MSI Z390 Carbon Pro AC motherboard and I’m screwing around with some overclocking, but it’s a lot less forgiving than my old 6600k was. I’ve got the multiplier set to 50 and voltage set in override mode (not adaptive or anything yet) to 1.33v and I can pass an hour in AIDA64 without crashing but it the CPU package temp gets up to 99c. AIDA shows 1% throttling and hwinfo confirms that it did throttle a few times on a couple of cores but not really by much and not for long at all.
|
# ? Jul 6, 2019 02:31 |
|
Indiana_Krom posted:My experience from fiddling with my 9900k on a custom loop water cooler is that the package seems to have a pretty hard limit at about 225w power consumption where no amount of traditional cooling will be able to prevent throttling. The only way to get it cooler is either reduce the power consumption, or delid and lap it. It’s in the high 60s/lower 70s in games so I’m not too worried about it since I’m not running a “stress test only” machine, but I still want to try to squeeze as much as I can out of it as efficiently as possible.
|
# ? Jul 6, 2019 02:44 |
|
GutBomb posted:Alright I just got a 9900k and an MSI Z390 Carbon Pro AC motherboard and I’m screwing around with some overclocking, but it’s a lot less forgiving than my old 6600k was. I’ve got the multiplier set to 50 and voltage set in override mode (not adaptive or anything yet) to 1.33v and I can pass an hour in AIDA64 without crashing but it the CPU package temp gets up to 99c. AIDA shows 1% throttling and hwinfo confirms that it did throttle a few times on a couple of cores but not really by much and not for long at all. I'm running a 9900k at 1.33v also under the same H115i Pro (mounted on the top of my case as an exhaust) and under p95 AVX workloads it gets into the 90s also. I haven't tried AIDA64. I think that's about as good as you'll get. The most demanding I've ever seen it in real world stuff is in Battlefield 5 it'll get up into the highish 80s. I consider this fine and so should you! Depending on your motherboard if you've left LLC at some default and have set a specific Vcore you might not be getting that at the processor. Can you google around and see which HWinfo64 heading on your motherboard gives you the best idea of your actual delivered vcore? You want to set your LLC so that your VID and Vcore are the same but if you leave it to adjust itself it might be giving more than 1.33v to the cpu by being too aggressive with LLC.
|
# ? Jul 6, 2019 04:22 |
|
VelociBacon posted:I'm running a 9900k at 1.33v also under the same H115i Pro (mounted on the top of my case as an exhaust) and under p95 AVX workloads it gets into the 90s also. I haven't tried AIDA64. I think that's about as good as you'll get. The most demanding I've ever seen it in real world stuff is in Battlefield 5 it'll get up into the highish 80s. I consider this fine and so should you! I have LLC on level 3 (MSI Claims “Auto” is 3 and mine is set to auto). I was just playing Dark Souls III for a couple of hours and I left hwinfo running and it says I maxed at 84c during that time, and that’s not a particularly demanding game. But then this morning I went and looked at my cooling setup and realized my boneheaded mistake. One of the fans on my rad was backwards. I turned it around so they are both blowing into the radiator and now stress tests top out at 85c on the package with no throttling on any cores. Edit: then I replaced the Corsair paste with some kryonaut and my temps dropped even more. I didn’t expect a 5 degree drop (under load) with a paste change but it happened. That Corsair paste must suck. GutBomb fucked around with this message at 23:47 on Jul 6, 2019 |
# ? Jul 6, 2019 15:07 |
|
GutBomb posted:I have LLC on level 3 (MSI Claims “Auto” is 3 and mine is set to auto). I was just playing Dark Souls III for a couple of hours and I left hwinfo running and it says I maxed at 84c during that time, and that’s not a particularly demanding game. That's interesting because I've got kryonaut sitting on my desk and I'm using the stock paste. Thanks for sharing.
|
# ? Jul 7, 2019 00:02 |
|
VelociBacon posted:That's interesting because I've got kryonaut sitting on my desk and I'm using the stock paste. Thanks for sharing. 5.2ghz crashes cinebench immediately upon starting the CPU benchmark at 1.38v and I don't really want to go any higher than that.
|
# ? Jul 7, 2019 20:41 |
|
OJ posted this info in the Parts thread that I was looking at about OCing Ryzen 3000: https://forums.somethingawful.com/showthread.php?threadid=3774409&pagenumber=728&perpage=40#post496689448 Just wondering if this is still the right way for my 3600 after seeing GN's video on PBO: https://www.youtube.com/watch?v=B7NzNi1xX_4 tl;dw via Reddit Precision Boost (PB) - AMD's boost algorithm, decides core clocks based on power (voltage¤t) and temperature. Can only boost to max stock clocks. Precision Boost Overdrive (PBO) - Increases the power that can be provided to the CPU by the VRMs, doesn't change clocks. AutoOC - Increases the value of the max clock by 200Mhz. DOES NOT mean the CPU WILL now boost to max clock+200Mhz, just that it CAN if Precision Boost (PB) decides there's enough power and thermal headroom to do so. PBO and AutoOC don't do much right now because throwing more power at the CPU is useless if the CPU doesn't have the thermal headroom to clock higher. IF you have an LN2 setup, then go right on ahead, the CPU clocks scale linearly all the way to -56C.
|
# ? Jul 21, 2019 17:23 |
|
Whats the consensus for max safe 24/7 VCCSA/VCCIO voltages for CFL chips? My Micron E-dies are coming and I would wanna drive them at 3600C16.
|
# ? Jul 22, 2019 05:49 |
|
1.2V-1.25V is what most people think is the max safe 24/7 setting, though many motherboards set them much higher with high frequency XMP profiles and there aren’t many reports of degradation yet. 3600 shouldn’t require more than 1.1-ish Volt though. Be aware that setting these voltages too high can also cause instability. TLDR:
|
# ? Jul 22, 2019 06:31 |
|
So I haven't really overclocked seriously since my celeron 300a days. And that was more just out of interest than anything. So fast forward to today and I'm trying to play with my RX570 8Gb using afterburner. What I've quickly noticed is that bumping the core clock up doesn't really have any affect on frame rate, even with a 12% increase (stable) but putting up the mem clock without touching the core sees an immediate improvement . Is that normal? This is only using the stability testing thing that comes with afterburner, I haven't tested it in a game yet as I was doing it late last night and that'll be tonight's project. Edit: as far as I understand it the 8gb version of my card was more targeted towards mining, so I'm wondering if the memory bandwidth might not be the greatest and that's why I'm seeing improvement from bumping that up as opposed to the core?
|
# ? Jul 25, 2019 23:21 |
|
My understanding is that most modern GPUs can get a bit of a free boost from increasing the memory clock although like with most overclocking it's a bit of a lottery. One thing to bear in mind is that your optimum clock speed will generally be considerably below your maximum stable speed - this is because your GPU is doing error correction, which means resending data, and at a certain point the increased speed is offset by the increase in errors.
|
# ? Jul 26, 2019 09:43 |
|
I want to get my RAM to run at it's specified timings. RAM: https://www.newegg.com/g-skill-16gb-288-pin-ddr4-sdram/p/N82E16820232306?Item=N82E16820232306 MOBO: ASUS ROG Strix x470-i CPU: 3700x So, my computer won't post anything if I set the DOCP profile. I have to reset the bios by jumping the pins on the motherboard, I've tried it 3 times. So, I manually set the voltage to 1.3 and the RAM speed to 3600. Great! It posted. A week later my dumb self realized I forgot to set any of the timings manually, so while it was running at 3600... the timings were set to something like 25-25-25-35. So I go back into BIOS and manually set the timings to 16-16-16-35 and a 1.35v this time. It posts, no crashes, everything seems stable. I haven't tried the 15-15-15-35 yet, but am going to this weekend. The only difference between these manual settings and the DOCP is the BCLK frequency. I currently have it set to auto, but the DOCP sets it to "100." I know nothing about BCLK, could that be causing the no post? Will I be safe to set the timings down to 15-15-15-35 and leave the BCLK at auto? Does ASUS just need to update BIOS to make the DOCP stable?
|
# ? Jul 26, 2019 15:44 |
|
Seems like the latest AGESA version AMD sent to vendors is fairly flaky at the moment, a fair number of folks are having RAM issues right now. Probably will need to wait for a bios fix.
B-Mac fucked around with this message at 18:29 on Jul 26, 2019 |
# ? Jul 26, 2019 18:26 |
|
eames posted:1.2V-1.25V is what most people think is the max safe 24/7 setting, though many motherboards set them much higher with high frequency XMP profiles and there aren’t many reports of degradation yet. 3600 shouldn’t require more than 1.1-ish Volt though. My new 2x16GB Micron E-die came and runs stable on 3500C16 @ 1.32V, under stock vccsa/vccio to boot too. Waaaay better than my old 2x16GB Hynix AFR (Corsair LPX Vengeance 3200C16) that took monstrous vccsa/vccio volts to even run at the rated XMP speed. Palladium fucked around with this message at 06:42 on Jul 27, 2019 |
# ? Jul 27, 2019 04:40 |
|
B-Mac posted:Seems like the latest AGESA version AMD sent to vendors is fairly flaky at the moment, a fair number of folks are having RAM issues right now. Probably will need to wait for a bios fix.
|
# ? Jul 27, 2019 17:04 |
|
1usmus should be publishing his ram calculator and MEMbench the 29th as well
|
# ? Jul 28, 2019 02:19 |
|
Thom P. Tiers posted:I want to get my RAM to run at it's specified timings. Does yours cold boot correctly with manually entered timings?
|
# ? Aug 3, 2019 07:02 |
|
GRINDCORE MEGGIDO posted:Does yours cold boot correctly with manually entered timings? Nope. I tried higher voltages, CL14 timings with a lower frequency, etc etc. Only thing I can get it to boot at is the 3600 CL16 1.35v. Which is fine I guess. Maybe ASUS will come out with a BIOS update that helps it run at the 15 DOCP. Oh well.
|
# ? Aug 3, 2019 15:51 |
|
Thom P. Tiers posted:Nope. I tried higher voltages, CL14 timings with a lower frequency, etc etc. Only thing I can get it to boot at is the 3600 CL16 1.35v. Which is fine I guess. Maybe ASUS will come out with a BIOS update that helps it run at the 15 DOCP. Oh well. So it cold boots at 3600 c16 ok? Interesting. Do you think you could post a screenshot of ryzen master, showing the memory timings? GRINDCORE MEGGIDO fucked around with this message at 01:47 on Aug 4, 2019 |
# ? Aug 4, 2019 01:45 |
|
I just downloaded MSI afterburner to start playing with my vid card. But before I started tweaking I decided to run the Heaven Benchmark and see what MSI reports my card at. I bought a Giggabyte 2070 Super Gaming OC advertised at 1815. the MSI "pointer" on the dial shows at about what I would guess is 1850. But when I run Heaven, the actual reported GPU clock in real numbers is 1980 -1995.. Did I get super lucky on my card? Kinda wondering if I should just leave it alone.
|
# ? Aug 9, 2019 07:18 |
|
Kilazar posted:I just downloaded MSI afterburner to start playing with my vid card. But before I started tweaking I decided to run the Heaven Benchmark and see what MSI reports my card at. All these cards automatically clock themselves up, you can't go by the box at all. It's still worth overclocking generally.
|
# ? Aug 9, 2019 07:48 |
|
VelociBacon posted:All these cards automatically clock themselves up, you can't go by the box at all. It's still worth overclocking generally. Awesome thanks, I will start pushing it then. One other question, uningene artifacts/flashes really bad on the closing static screen when you "quit" and it's doing that without an OC. Is that just bad programming on that exit program splash screen?
|
# ? Aug 9, 2019 13:37 |
|
Kilazar posted:Awesome thanks, I will start pushing it then. Grab the 3dmark demo on steam and run the timespy benchmark and see if you get weird flashing in there. For me when I was too overclocked it was flash/flicker-heavy.
|
# ? Aug 9, 2019 14:18 |
|
My EVGA 2070 non-super does not artifact in 3DMark Time Spy, but if too heavily overclocked will just not start the demo at all. If it can load it, it doesn't seem to have an issue the whole time. Too-ambitious overclock causes some game engines to crash for me too. But I have not yet seen artifacting like I would have expected from it, just go-until-it-crashes behavior in games 'til I bumped the clocks down a bit to their current values.
Agreed fucked around with this message at 14:51 on Aug 9, 2019 |
# ? Aug 9, 2019 14:49 |
|
I've found a good quick 5 min stability test is to ride a horse around in Witcher 3 with everything maxed.
|
# ? Aug 9, 2019 14:51 |
|
Are there any programs that can load up a GPU while the PC is being controlled over RDP? It would be nice to stress test (and monitor) while I'm not home.
|
# ? Aug 20, 2019 21:41 |
|
Never thought about it before, but I tried it just now and FurMark over RDP works fine.
Eletriarnation fucked around with this message at 22:51 on Aug 20, 2019 |
# ? Aug 20, 2019 22:49 |
|
Oh weird. Furmark not working is literally what inspired my post. It detects my 1080 as GPU 1 and the RDP adapter as #2, but errors stating it needs a OpenGL 2.0 device when I try to start the test.
|
# ? Aug 20, 2019 22:54 |
|
When I start it over RDP, I only see the regular GPU - in this case, a Radeon 7750 since it's my work desktop. The stress test seems to work fine. Maybe there's a special requirement for RDP to use hardware acceleration? Both the source and destination PCs in my case are Windows 10, which might matter.
|
# ? Aug 20, 2019 23:54 |
|
I’m running w10 Pro 1903 so who knows. It looks like OCCT is happy to do it over RDP as long as the furry donut is in the foreground.
|
# ? Aug 21, 2019 00:32 |
|
I have an odd occurrence while messing around with my 9900k on an ASRock z370 taichi, maybe someone can shed some light on it. While messing around with undervolting, fixed mode vcore of 1.25 with LLC 1 will put 1.216V into the cpu with a cpu package power of 250W or so in Prime95 small FFT with AVX. If I set voltage to offset mode with a negative 40 mV offset LLC 3 I still get 1.216V as reported hwinfo64 but packager power is closer to 300W. I confirmed the difference using a kill-a-watt meter at the wall and it’s accurate. I have no idea why offset mode would draw so much power versus fixed at the same voltage output. B-Mac fucked around with this message at 05:36 on Aug 21, 2019 |
# ? Aug 21, 2019 05:32 |
|
Sounds to me like the voltage reporting between the two modes and/or LLC levels is inconsistent.
|
# ? Aug 21, 2019 06:20 |
|
Just tried my first OC on my 9600k/Asus Z390 build. Noctua cooler and case fans. I set 48 multiplier at 1.25v to start. Set LLC to 6. Looks like I idle at 35-36c now and I’m seeing 77c max during Realbench/Cinebench. Sound about right? Still learning.
|
# ? Aug 23, 2019 06:12 |
|
If anybody is looking for fans with high static pressure (heatsink, radiator, case with restricting filter) I recommend looking into the Arctic P12 and Arctic P14. They were recommended on various forums and I ordered two because they seemed ridiculously cheap (4€ for one simple 120mm fan). The blade design looks similar to the Servo Gentle Typhoon. Build quality isn't great however when normalized to be quiet (~1300rpm) they match the Noctua NF-F12s (~1000rpm) in my case with temperatures even being slightly lower. The non-PWM version spins up at 350 RPM which is also decent. They're quite loud at max RPM though. My only concern is bearing longevity but I already kind of regret buying the Noctuas.
|
# ? Aug 30, 2019 09:38 |
|
Just to add confirmation, for Ryzen 3000 series find your infinity fabric maximum and run your RAM at that speed using whatever CL is stable. I moved from 3200 CL16 to 3600 CL 20 and reduced by nearly 10ns latency while gaining a huge amount of memory bandwidth
|
# ? Sep 2, 2019 13:19 |
|
The rumour that your hardware won't last as told is true afterwards.
|
# ? Sep 15, 2019 22:30 |
|
attempting to continue the discussion from the GPU thread here...Paul MaudDib posted:SiliconLottery's historical binning stats are here. Note that they typically use a 200 MHz AVX offset. 1.437V is high but can still be very safe if you run a low/safe LLC on Z370, because it ends up around 1.34V under load but without transient spikes. Arguably its much safer than 1.4V with high LLC (=what SL recommends, but they only give one year of warranty... ) because high voltages at low loads are not as harmful as high voltages (or -spikes) under load. This posting by an ex-Asus R&D engineer has nice graphs that illustrate the effects of LLC, you’ll want to look at the spike at the end of the dip: https://www.overclock.net/forum/6-intel-motherboards/1638955-z370-z390-vrm-discussion-thread-412.html#post28022104 Note that the transient loads pictured are from a simple interrupt while the UEFI is idle, you can probably imagine how it compares to a core going from full AVX load to idle. The big unknown is wether it is „safer“ to run high Vcores with low LLCs or medium Vcores with medium LLCs... but considering many people are running high Vcores with high LLCs and haven’t degraded their CPU it may not matter at all. eames fucked around with this message at 07:41 on Oct 8, 2019 |
# ? Oct 8, 2019 07:35 |
|
|
# ? May 15, 2024 04:48 |
|
eames posted:attempting to continue the discussion from the GPU thread here... I know that coffee lake comprises more than just the 9900k but I'd be pretty surprised if many people can run at 1.4v with the 9900k even with doing some serious custom loop cooling and even then I feel like you'd need to remove the solder TIM and replace it with liquid metal etc etc. Using the 'flattest' LLC you can still seems the best option with incremental increases of vcore unless I'm misreading that post?
|
# ? Oct 8, 2019 14:33 |