TC McQueen
Member
Unless they do something that radically alters how the chip interacts with the mobo, it should work.
At worst, you might have to update the chipset driver before swapping out the CPU.
-
Hey, guest user. Hope you're enjoying NeoGAF! Have you considered registering for an account? Come join us and add your take to the daily discourse.
AMD Ryzen Thread: Affordable Core Act
- Thread starter ·feist·
- Start date
At worst, you might have to update the chipset driver before swapping out the CPU.
According to what was said so far, AM4 [and most likely the TH4] will remain in use for 3-4 next years.
It's likely. AMD is much better about maintaining socket and motherboard compatibility than Intel.
Harbinger of War
Member
Hey, glad to hear everything is up and running. Any issues getting that Corsair LPX memory working on the B350-F? Do you have it running at stock speeds/voltages or did you have to underclock it? And would you mind sharing your early thoughts on the B350-F itself?
Soooooooo........I put together a new Asus X370-Pro/Ryzen 1800x build the other day and am having sporadic issues with hard machine lockups and restarts. Memory is the Corsair Vengence RGB 3466 8GB x2 kit (CMR16GX4M2C3466C16) - I believe it's on the mobo's QVL list. Power supply is a Corsair RM650x. OS (Fresh W10 Pro install) is running on a new Samsung 960 Pro 512GB SSD and I've got one older 1TB Samsung HD + Optical drive installed also.
I did enable DOCP in the BIOS and the memory is running at 3200MHZ. I also ran the basic overclocking process from the BIOS and it kicked the CPU speed up to 3.85 Ghz. PC posts fine and I was able to run Windows 10 and use the machine without issues for the first day or so. Then it started to crash/lockup - display suddenly shuts down and rig becomes completely unresponsive. Fans continue to run and RAM continues to cycle through the default LED patterns. But it takes a reset of the power supply switch to get the pc's power switch to respond afterwards - Windows event viewer reports nothing but an unexpected shutdown. Machine will boot and will remain responsive for a period of time after that, but this issue has occurred at random points - even during sleep. I initially thought it was related to Hybrid sleep, so I shut that off. But I've had it pack in while I'm doing casual web browsing. It did also keel over when I tried to run a Prime 95 test today, but that resulted in a Windows crash instead of a hard lockup of the Mobo and generated a dump file. So that feels like a different issue.
Also weird - I ran the Thaiphoon utility, but it failed to identify the RAM manufacturer. I believe the sticks above are using Samsung B-Die chips, but I find it odd that the utility won't tell me one way or the other. It also took ages to run - like 25+ minutes for DIMM # 1 the first time I ran it. So also not sure if that is normal.
Only other thing I can think to mention is that the memory positioning in DIMM slots 1 & 3 was tight when using the Scythe Fuma cooler I have installed. As in, the memory in DIMM slot 1 may actually be making incidental contact with the Fuma cooler. I just touched the cooler however and it's not hot to the touch, so I'm not sure if that makes a lick of difference. I could move the memory to DIMM slots 2 & 4, but can't recall if that will incur a performance hit?
Anyway, any troubleshooting suggestions would be appreciated. Again, the random nature of the problem is somewhat aggravating, so I'm not sure if it's related to heat, voltages or what.
I did enable DOCP in the BIOS and the memory is running at 3200MHZ. I also ran the basic overclocking process from the BIOS and it kicked the CPU speed up to 3.85 Ghz. PC posts fine and I was able to run Windows 10 and use the machine without issues for the first day or so. Then it started to crash/lockup - display suddenly shuts down and rig becomes completely unresponsive. Fans continue to run and RAM continues to cycle through the default LED patterns. But it takes a reset of the power supply switch to get the pc's power switch to respond afterwards - Windows event viewer reports nothing but an unexpected shutdown. Machine will boot and will remain responsive for a period of time after that, but this issue has occurred at random points - even during sleep. I initially thought it was related to Hybrid sleep, so I shut that off. But I've had it pack in while I'm doing casual web browsing. It did also keel over when I tried to run a Prime 95 test today, but that resulted in a Windows crash instead of a hard lockup of the Mobo and generated a dump file. So that feels like a different issue.
Also weird - I ran the Thaiphoon utility, but it failed to identify the RAM manufacturer. I believe the sticks above are using Samsung B-Die chips, but I find it odd that the utility won't tell me one way or the other. It also took ages to run - like 25+ minutes for DIMM # 1 the first time I ran it. So also not sure if that is normal.
Only other thing I can think to mention is that the memory positioning in DIMM slots 1 & 3 was tight when using the Scythe Fuma cooler I have installed. As in, the memory in DIMM slot 1 may actually be making incidental contact with the Fuma cooler. I just touched the cooler however and it's not hot to the touch, so I'm not sure if that makes a lick of difference. I could move the memory to DIMM slots 2 & 4, but can't recall if that will incur a performance hit?
Anyway, any troubleshooting suggestions would be appreciated. Again, the random nature of the problem is somewhat aggravating, so I'm not sure if it's related to heat, voltages or what.
RumblingRosco
Member
Soooooooo........I put together a new Asus X370-Pro/Ryzen 1800x build the other day and am having sporadic issues with hard machine lockups and restarts. Memory is the Corsair Vengence RGB 3466 8GB x2 kit (CMR16GX4M2C3466C16) - I believe it's on the mobo's QVL list. Power supply is a Corsair RM650x. OS (Fresh W10 Pro install) is running on a new Samsung 960 Pro 512GB SSD and I've got one older 1TB Samsung HD + Optical drive installed also.
I did enable DOCP in the BIOS and the memory is running at 3200MHZ. I also ran the basic overclocking process from the BIOS and it kicked the CPU speed up to 3.85 Ghz. PC posts fine and I was able to run Windows 10 and use the machine without issues for the first day or so. Then it started to crash/lockup - display suddenly shuts down and rig becomes completely unresponsive. Fans continue to run and RAM continues to cycle through the default LED patterns. But it takes a reset of the power supply switch to get the pc's power switch to respond afterwards - Windows event viewer reports nothing but an unexpected shutdown. Machine will boot and will remain responsive for a period of time after that, but this issue has occurred at random points - even during sleep. I initially thought it was related to Hybrid sleep, so I shut that off. But I've had it pack in while I'm doing casual web browsing. It did also keel over when I tried to run a Prime 95 test today, but that resulted in a Windows crash instead of a hard lockup of the Mobo and generated a dump file. So that feels like a different issue.
Also weird - I ran the Thaiphoon utility, but it failed to identify the RAM manufacturer. I believe the sticks above are using Samsung B-Die chips, but I find it odd that the utility won't tell me one way or the other. It also took ages to run - like 25+ minutes for DIMM # 1 the first time I ran it. So also not sure if that is normal.
Only other thing I can think to mention is that the memory positioning in DIMM slots 1 & 3 was tight when using the Scythe Fuma cooler I have installed. As in, the memory in DIMM slot 1 may actually be making incidental contact with the Fuma cooler. I just touched the cooler however and it's not hot to the touch, so I'm not sure if that makes a lick of difference. I could move the memory to DIMM slots 2 & 4, but can't recall if that will incur a performance hit?
Anyway, any troubleshooting suggestions would be appreciated. Again, the random nature of the problem is somewhat aggravating, so I'm not sure if it's related to heat, voltages or what.
Reset everything to the non overclocked state. See if the lockups still continue. If they stop, it may be due to an overclocking issue.
Also, if you're running Windows 10, make sure you've applied all updates. The Windows install was very unstable until I applied the latest 1703 update.
I've had some issues with long term stability after applying an OC. I've backed back down to the defaults to see if it was an OC issue or some other Windows/Ryzen compatibility issue. The last day are so have been pretty stable....which I'll take over a modest speed increase any day. I just hope that future driver/BIOS updates allow me to push my system a little further.
This is from your motherboard's manual (and applies to nearly all Ryzen motherboards with a standard configuration):
http://dlcdnet.asus.com/pub/ASUS/mb...O/E12344_PRIME_X370-PRO_web_only_20170712.pdf
http://dlcdnet.asus.com/pub/ASUS/mb...77_PRIME_X370-PRO_UM_v2_web_only_20170712.pdf
After changing your memory to the 2nd and 4th DIMM slots, reset your UEFI/BIOS to stock settings as mentioned above by Soka. Rerun Thaiphoon Burner and you should have your values displayed correctly.
Also, as FireCould stated, be sure to have Windows and your motherboard BIOS up to date:
https://www.asus.com/uk/Motherboards/PRIME-X370-PRO/HelpDesk_Download/
So......interesting development. I had installed the Corsair Link software the other day to play with the LED patterns on the Vengeance RAM. While doing some troubleshooting, I happened to notice the Link processes were memory resident despite my having not launched the software. So I un-installed it, thinking it might be interfering with the Thaiphoon scan and possibly even contributing to the system instability.
One reboot later and damned if the Thaiphoon scan isn't working correctly now (See below) and completing in the fraction of the time to boot. So I re-applied the DOCP setting in the BIOS to clock the memory back up to 3200MHZ and we'll see if the stability problems persist.
·feist·;246513094 said:
Missed this initially - thx. As mentioned above, I think Corsair Link was causing issues. But I'm going to relocate the RAM anyway.
RumblingRosco
Member
I've heard others with issues regarding RGB RAM and Ryzen. I don't have RGB RAM so I never looked into it further, but you might want to Google it a bit.
Thanks for the update. Depending on how intel handles the 8700k i may create a Ryzen build instead. Especially if I can reuse my motherboard for a better cpu. Thinking about going with an 1600 or 1700 and then replacing this with the Ryzen gen 2/3 in a couple years when its out.
What motherboards reliably hit 3200mhz speed for memory? Tough to find a mobo without complaints from people saying they can't hit 3200.
What motherboards reliably hit 3200mhz speed for memory? Tough to find a mobo without complaints from people saying they can't hit 3200.
From what I gathered (and from personal experience with my Asus B350 Plus) is that it's important to look at the QVL for memory. I bought 3200 MHz ram from corsair because Asus's QVL said it would do 3200 Mhz after the bios update, and sure enough it did, first try. To be safe, to pair random mother boards and memory, make sure you do your research.
Wozzer
Member
With the new RAM my Ryzen build finally seems to be stable, even at 3200mhz. I am however experiencing very noticable stutters when playing games.
Thanks to the Gsync FPS counter I'm seeing dips from a stable 144fps on CS:GO down to 90's and back up again. It's not being picked up by other FPS monitors, which I can only assume being because they're not refreshed as often.
It appears to get worse as I do more on the machine; and that's not to say heavy multi tasking but literally opening a 3rd browser tab :S
Thanks to the Gsync FPS counter I'm seeing dips from a stable 144fps on CS:GO down to 90's and back up again. It's not being picked up by other FPS monitors, which I can only assume being because they're not refreshed as often.
It appears to get worse as I do more on the machine; and that's not to say heavy multi tasking but literally opening a 3rd browser tab :S
What's the best way to determine why a system is crashing randomly? I'm in the middle of finishing a build for my girlfriend and the system is crashing/locking up randomly, like, after it's been on and running fine for 20 - 30 minutes.
When this crash happens, my case lights freeze up and stay like that, but my keyboard/mouse lights shut off, and my monitor loses all signal. Hitting the power button/reset button on my case do nothing, and I'm forced to turn off my PSU manually. Upon booting back up, the system has trouble starting up 2-3 times before finally booting into Windows successfully on the 3rd/4th attempt.
If someone wants to help me/can help me figure out why this is happening, I'd appreciate the help, cause it could be so many different things and I have no idea where to start looking. Here are the specs of the computer:
When this crash happens, my case lights freeze up and stay like that, but my keyboard/mouse lights shut off, and my monitor loses all signal. Hitting the power button/reset button on my case do nothing, and I'm forced to turn off my PSU manually. Upon booting back up, the system has trouble starting up 2-3 times before finally booting into Windows successfully on the 3rd/4th attempt.
If someone wants to help me/can help me figure out why this is happening, I'd appreciate the help, cause it could be so many different things and I have no idea where to start looking. Here are the specs of the computer:
Captain Chaos
Member
I have a MSI B350 Gaming Plus matched with R5 1400 and 16GB Corsair RAM. Machine is rock solid with CPU running at 3.6 and RAM at 3200.
Sound like the motherboard is trying to reset something during boot up since power delivery might be insufficient for some component to operate correctly.
First thing to check would probably be the RAM since that's the weakest link for Ryzen
Don't overclock anything
Update to the latest BIOS
Set the RAM to JEDEC standard, not XMP
Run Windows Memory Diagnostic or Memtest86+
I reset my RAM back to default speeds and ran memtest: no problems. I should also mention I had this happen once before and once after setting my RAM to XMP settings.
So......interesting development. I had installed the Corsair Link software the other day to play with the LED patterns on the Vengeance RAM. While doing some troubleshooting, I happened to notice the Link processes were memory resident despite my having not launched the software. So I un-installed it, thinking it might be interfering with the Thaiphoon scan and possibly even contributing to the system instability.
One reboot later and damned if the Thaiphoon scan isn't working correctly now (See below) and completing in the fraction of the time to boot. So I re-applied the DOCP setting in the BIOS to clock the memory back up to 3200MHZ and we'll see if the stability problems persist.
Missed this initially - thx. As mentioned above, I think Corsair Link was causing issues. But I'm going to relocate the RAM anyway.
I have the Asus B350-Plus Mobo with the 1700x and the exact same RAM. I've been having the same issues with instability and have messed with timings and voltage to no avail. It's been mighty frustrating.
I reread the QVL this week and it says it only supports this RAM in a 1 DIMM setup (8GB vs 16GB). Feeling a bit foolish now because I missed that detail on this RAM and I'm not sure if I can return it now. Hopefully it will be supported fully soon? I don't want to have to run this kit at minimum speed to have stability.
SmokedMeat
Gamer™
If everything goes through (selling my CPU/MB tomorrow) I'll be looking to buy a new setup asap.
What's the safest Motherboard/ram combo to pick for a Ryzen setup? I just want something that works with no hassle.
What's the safest Motherboard/ram combo to pick for a Ryzen setup? I just want something that works with no hassle.
could be the harddrive since memtest ran without problems and the system didn't lock up
also try to run "sfc /scannow" in the command window to check windows for corrupt files
Early impressions with my rig after the changes I described look promising, but I'm gonna have to give it a few days to see how it goes. I did just run a Prime95 run for about an hour and it completed, albeit with some rounding errors noted in the log. Not sure if that's normal or not, but I'm happy it didn't crash within a few minutes as it did yesterday pre-changes. Probably worth noting that my 1800X CPU is only clocked at 3.6GHZ presently also. I'll probably bump that up gradually once I've experienced a stable period.
I hear you though on the frustration front - this is my first AMD build in years and the first (to the best of my knowledge) where I've had to mess with overclocked memory. My Intel builds were largely hassle free, so having to deal with system instability is a PITA. I did go into it with my eyes open however knowing about the Ryzen/memory issues, so I wasn't surprised to run into problems.
I would recommend you run that Thaiphoon utility against your DIMM's - it had the added benefit beyond memory identification in that it flagged the Corsair Link issue with my rig. Maybe it will highlight an issue with your setup also.
The RAM situation seems really frustrating with these AM4 motherboards. Yet if I can find something that works I'll be ok because it's really appealing to me to not have to buy another mother board to simply replace my CPU.
One thing that confuses me about AMD is why did they only include the Wraith Spiral fan on the base 1600 and 1700? This cooler seems legit if all you are interested in is going to 3.6-3.7 ghz, while these chips tend to max out around 4.0 ghz. I can buy a 1600 for 200 dollars, have a decent cooler included, and replace it two years with an 8 or 12 core ryzen gen 2 chip without having to buy a new motherboard. Almost seems too good to be true.
Pick a motherboard and then go to this motherboards web page and check the memory quality vendor list (QVL). It should give you the memory manufacturer, speed, and part number of a memory chip(s) that will work. In doing some research it seems like any speed over DD4 3000 is slim pickings on these motherboard memory QVL lists. Don't forget you'll probably have to update your motherboard bios to make certain memory on the QVL list works as intended.
Which Corsair ram did you buy?
One thing that confuses me about AMD is why did they only include the Wraith Spiral fan on the base 1600 and 1700? This cooler seems legit if all you are interested in is going to 3.6-3.7 ghz, while these chips tend to max out around 4.0 ghz. I can buy a 1600 for 200 dollars, have a decent cooler included, and replace it two years with an 8 or 12 core ryzen gen 2 chip without having to buy a new motherboard. Almost seems too good to be true.
Pick a motherboard and then go to this motherboards web page and check the memory quality vendor list (QVL). It should give you the memory manufacturer, speed, and part number of a memory chip(s) that will work. In doing some research it seems like any speed over DD4 3000 is slim pickings on these motherboard memory QVL lists. Don't forget you'll probably have to update your motherboard bios to make certain memory on the QVL list works as intended.
Which Corsair ram did you buy?
Prime95 will not report errors if everything works correctly.
I never had errors when I tested my RAM with Prime95, my system just shut down
Prime95 ran error free when I finally found stable timings for my RAM
Prime95 will not report errors if everything works correctly.
I never had errors when I tested my RAM with Prime95, my system just shut down
Prime95 ran error free when I finally found stable timings for my RAM
I may end up having to tweak the RAM settings some more then. I'll see how it goes in terms of my original system lockup problem for now however.
This is so weird.
The following ALWAYS crashes my system. It's just a crash though.
Still, don't try this unless you're ready for a possible system crash. Or even worse.
Right Click Taskbar Time -> Adjust Date and Time -> Set Time Automatically
Toggle on and off repeatedly as fast as possible = Crash.
I don't know if it's a Ryzen issue, a ThreadRipper issue, a Motherboard issue, or even a Windows 10 issue.
The following ALWAYS crashes my system. It's just a crash though.
Still, don't try this unless you're ready for a possible system crash. Or even worse.
Right Click Taskbar Time -> Adjust Date and Time -> Set Time Automatically
Toggle on and off repeatedly as fast as possible = Crash.
I don't know if it's a Ryzen issue, a ThreadRipper issue, a Motherboard issue, or even a Windows 10 issue.
Louis Cyphre
Banned
I've never been able to get my ram above 3000. I have the Corsair Vengeance LPX - CMK16GX4M2B3200C16 but I have not tried the latest BIOS for my MSI Carbon Fiber.
I don't see it on the list here:
https://www.msi.com/Motherboard/support/B350-GAMING-PRO-CARBON.html#support-mem-2
If that is your motherboard. Now that I see the compatibility I might go with MSI B350 Pro Carbon because there is a big amount of compatibility with DD4 3200 memory.
Edit: I think I found your memory here on the list:
CMK16GX4M2B3200C16W
That's a shame it can't go above 3000.
https://www.msi.com/Motherboard/support/B350-GAMING-PRO-CARBON.html#support-mem-2
If that is your motherboard. Now that I see the compatibility I might go with MSI B350 Pro Carbon because there is a big amount of compatibility with DD4 3200 memory.
Edit: I think I found your memory here on the list:
CMK16GX4M2B3200C16W
That's a shame it can't go above 3000.
RumblingRosco
Member
What does the clock in your BIOS say? Also, how the heck did you even find that this crashes things?
Ran a test and it came back clean. I've now had the crash happen a third time, this time in the middle of a cleaning utility going through my system and cleaning out junk files. So maybe you're onto something with it being my SSD. I really hope not though.
I have Corsair vengeance rgb installed, works perfectly. The gskill trident rgb I bought first on the other hand died because of corrupted spd.
FingerBang
Member
Yeah, I ended up going with the base 1600 instead of the x version because of that. I OC'd it to 3.7 with a voltage of 1.25 or 1.27 (don't remember) and it runs stable with the Wraith Spiral fan. I know that my chip can reach 3.8 and even 3.9 but the temps are a little too high for the stock fan to handle... and I don't really need the extra speed.
I don't know if I'm going to need extra CPU power, but if AMD can make Ryzen better and reach 4.5ghz or even 5ghz I might upgrade to zen 2 or 3 in a couple of years. Something that right now seems impossible with Intel
mordecaii83
Member
Just wanted to check in after spending the last few days trying to get my Threadripper system stable...
I'm currently sitting on a memory clock of 3333 (bought new RAM, couldn't get my old set above 2800), and the 1950x is sitting at stock clocks. I can run it at 4.0GHz, but it gets too hot if I run a stress test like Intel Burn Test or Prime95. It does however remain cool enough under Cinebench and WPrime, but I decided it just wasn't worth it. I haven't ran into any games or anything else where stock clocks are holding it back.
I may try to increase the RAM speed to 3433 or 3600 (my memory is rated at up to 3866), but otherwise I'm finally happy with the system. I was having crashing issues that took me forever to track down, but I've been stable for over 24 hours now no matter what I've thrown at the system. It's pretty awesome to see games like Watch Dogs 2 using all 32 threads!
I'm currently sitting on a memory clock of 3333 (bought new RAM, couldn't get my old set above 2800), and the 1950x is sitting at stock clocks. I can run it at 4.0GHz, but it gets too hot if I run a stress test like Intel Burn Test or Prime95. It does however remain cool enough under Cinebench and WPrime, but I decided it just wasn't worth it. I haven't ran into any games or anything else where stock clocks are holding it back.
I may try to increase the RAM speed to 3433 or 3600 (my memory is rated at up to 3866), but otherwise I'm finally happy with the system. I was having crashing issues that took me forever to track down, but I've been stable for over 24 hours now no matter what I've thrown at the system. It's pretty awesome to see games like Watch Dogs 2 using all 32 threads!
opticalmace
Member
Nice. What cooler are you using?
mordecaii83
Member
I'm using Thermaltake's Water 3.0 Ultimate. It keeps the CPU around 55C max even under stress testing unless I OC, then it goes up rather quickly. Can't really be surprised with 16 cores running 100% though!
I haven't tried any OC besides 4.0GHz so I may try 3.9 and see what temps are like on that, but I'm still planning on leaving it stock for daily use.
The Bios clock is fine now but it was wrong when I found out about triggering that crash.
The computer crashed at night. When I turned it on the time was off. I toggled the switch to refresh it and I guess my connection was messed up at that time since I had to do it several times. On the 5th toggle or so (10 clicks) it just died.I tried it again on reboot and this time it crashed on the second toggle.
My guess based on what I saw was that Windows update triggered a time update which crashed the computer that first time. It hasn't crashed randomly since I just turned off the auto update time thing.
·feist·;246398252 said:
Yup, just saw that Computerbase.de has been playing with the Noctua full-coverage air and it seems to be pretty good even compared to AIO solutions. I wonder how full-coverage AIO will perform.
I've pushed mine to 4GHz it could probably go to at least 4.1Ghz too, but temps are too high in all core stress tests with my 240mm radiator.
Hmm, 3866 mhz kit, so I'm guessing single rank? If so, I can see you pushing that to 3466 or 3600 at least with some minor voltage tweaking. I've got 3333 on my dual rank, but 3466 is a limit to far at the moment.
mordecaii83
Member
I thought I had 3466 stable, but after about an hour of memtest86+, I got 2 errors. I've tried voltages all the way up to 1.45, but I just can't get it to be stable. I don't think my CPU was 100% stable at that time so it's possible that's what caused the errors, but I don't think so.
What stress test are you using? I've been doing a mix of Intel Burn Test, WPrime, Prime95, and Cinebench. I can pass WPrime and Cinebench pretty easily but IBT and P95 cause my temps to go way higher than the other two.
Edit: Also, just a complaint I have... I set the pump on my AIO to always be running at 100% by turning off "Smart Fan" in the bios because that's how it's supposed to be set up, but it will randomly reset itself to be pwm and I can't figure out why. It's really annoying because I constantly have to check and see if it reset itself whenever I reboot the system.
I've noticed quite a few posts from people with new builds frustrated about the memory situation.
I understand that frustration, but don't immediately overclock your system to the max right out of the box, and there are no guarantees above 2933MT/s on most boards.
Ryzen officially supports up to DDR4-2666. Depending on your configuration (dual/single rank memory, 2 or 4 sticks) even that can be considered an overclock - though most people don't seem to encounter problems running at that speed now.
If you're overclocking the CPU, IntelBurnTest was a good immediate stress test. At the maximum stress level, it would generally fail before 15-20 loops could complete on my system if it was unstable.
Prime95 producing the maximum heat seemed to be a better test of long-term stability. Even once the CPU was passing IBT, Prime95 would still fail eventually until I increased the voltage another couple of ticks in the UEFI. But IBT failed much faster, so I wouldn't start with Prime95.
Memory overclocking requires different tools for stability testing:
HCI MemTest (run it to at least 1000% coverage), TPU MemTest64, or Google stressapptest (which requires Linux).
You need to leave these tests on overnight after testing stability to confirm that your memory overclock is not producing errors.
People also recommend leaving HWiNFO64 running when stress testing, as it will report WHEA Errors. These errors may not crash the system, but if you get any WHEA Errors, your overclock is not stable.
Please monitor the heat that your CPU is putting out with these tests.
Though Ryzen is not the hottest CPU, they will put out maximum heat and really stress your system's cooling solution.
Make sure that your CPU voltage under load is no higher than 1.425V - though I would generally target around 1.35-1.40V for a long-term overclock.
I would suggest maybe starting with RAM at 1.35V and 2666MT/s, and the CPU around 1.375V. Push up the CPU clockspeed as high as it will boot. Then gradually drop it by 100MHz at a time until it can pass 20 loops of IBT.
Once it can pass 20 loops of IBT, move on to Prime95. There, you might have to either adjust the voltage or adjust the clockspeed in 25MHz increments to ensure stability.
Finally, memory testing.
If it's passed the above tests, I would be surprised if it then failed memory testing at 2666MT/s.
I'd still want to test it personally but I would probably see what the fastest memory speed is the system will POST at with the CPU overclock dialed in (and please make a note of what that is, as your system may reset everything if it can't POST) and work backwards from that until it can pass 1000% of HCI MemTest or run one of the other tests overnight.
Apparently Google's stressapptest is the fastest tool for checking memory errors but it requires booting into Linux.
Of course this is all if you want a 100% stable system.
I won't ever tolerate random crashes/reboots, while I've seen a lot of people on gaming forums say that if it can pass 10 loops of IBT and runs a couple of games without crashing, they're happy and don't concern themselves about anything else.
If it does eventually crash, they'll just bump up the voltage or drop the clockspeed a tick, and repeat that until it stops happening.
I would also be running these tests on a pre-built system or a system that was not being overclocked too. Any new hardware is subjected to at least a few days if not a week of testing. Hard drives alone can take a whole week to do full surface testing now that their capacity has grown so much faster than their speed. (I'd never put data on an HDD without doing that first)
I understand that frustration, but don't immediately overclock your system to the max right out of the box, and there are no guarantees above 2933MT/s on most boards.
Ryzen officially supports up to DDR4-2666. Depending on your configuration (dual/single rank memory, 2 or 4 sticks) even that can be considered an overclock - though most people don't seem to encounter problems running at that speed now.
If you're overclocking the CPU, IntelBurnTest was a good immediate stress test. At the maximum stress level, it would generally fail before 15-20 loops could complete on my system if it was unstable.
Prime95 producing the maximum heat seemed to be a better test of long-term stability. Even once the CPU was passing IBT, Prime95 would still fail eventually until I increased the voltage another couple of ticks in the UEFI. But IBT failed much faster, so I wouldn't start with Prime95.
Memory overclocking requires different tools for stability testing:
HCI MemTest (run it to at least 1000% coverage), TPU MemTest64, or Google stressapptest (which requires Linux).
You need to leave these tests on overnight after testing stability to confirm that your memory overclock is not producing errors.
People also recommend leaving HWiNFO64 running when stress testing, as it will report WHEA Errors. These errors may not crash the system, but if you get any WHEA Errors, your overclock is not stable.
Please monitor the heat that your CPU is putting out with these tests.
Though Ryzen is not the hottest CPU, they will put out maximum heat and really stress your system's cooling solution.
Make sure that your CPU voltage under load is no higher than 1.425V - though I would generally target around 1.35-1.40V for a long-term overclock.
I would suggest maybe starting with RAM at 1.35V and 2666MT/s, and the CPU around 1.375V. Push up the CPU clockspeed as high as it will boot. Then gradually drop it by 100MHz at a time until it can pass 20 loops of IBT.
Once it can pass 20 loops of IBT, move on to Prime95. There, you might have to either adjust the voltage or adjust the clockspeed in 25MHz increments to ensure stability.
Finally, memory testing.
If it's passed the above tests, I would be surprised if it then failed memory testing at 2666MT/s.
I'd still want to test it personally but I would probably see what the fastest memory speed is the system will POST at with the CPU overclock dialed in (and please make a note of what that is, as your system may reset everything if it can't POST) and work backwards from that until it can pass 1000% of HCI MemTest or run one of the other tests overnight.
Apparently Google's stressapptest is the fastest tool for checking memory errors but it requires booting into Linux.
Of course this is all if you want a 100% stable system.
I won't ever tolerate random crashes/reboots, while I've seen a lot of people on gaming forums say that if it can pass 10 loops of IBT and runs a couple of games without crashing, they're happy and don't concern themselves about anything else.
If it does eventually crash, they'll just bump up the voltage or drop the clockspeed a tick, and repeat that until it stops happening.
I would also be running these tests on a pre-built system or a system that was not being overclocked too. Any new hardware is subjected to at least a few days if not a week of testing. Hard drives alone can take a whole week to do full surface testing now that their capacity has grown so much faster than their speed. (I'd never put data on an HDD without doing that first)
What coolers are my fellow ThreadRippers using? And what temps and frequencies are you getting on prolonged (1-2+ hour) workloads?
1950X here. I'm using an EVGA CLC 280mm with stock thermal paste and I'm seeing low 60s and typically 4.2GHz with the stock auto OC. But that's based off preliminary tests. I'll be benching the system hard later today.
1950X here. I'm using an EVGA CLC 280mm with stock thermal paste and I'm seeing low 60s and typically 4.2GHz with the stock auto OC. But that's based off preliminary tests. I'll be benching the system hard later today.
Captain Chaos
Member
mordecaii83
Member
Stock I'm seeing temps in the high 50's, but I've only ever seen my processor boost up to 4.175, and only for a short time. I see it boosting up to 4.1 all the time until it starts doing something more demanding.
RumblingRosco
Member
I might be wrong on this, but I like to run the Prime95 Blend Test to do some of my memory stability tests. It is also a nice test to run IMO once everything is "stable" to do a double-check. To be totally honest, I don't really understand what any of the Prime95 tests actually do, I've just always used Blend as my very last test.
So, my post is somewhat pointless, but just a thought.
Prime95 is mainly a CPU stress test that hits some amount of memory and checks for stability.
I would use Prime95/IBT to see if the system crashes with your memory overclock. (after your CPU overclock is completely stable)
But these tools do nothing to test for memory errors. Memory needs to always return back exactly whatever data was placed in it. Memory errors are when it does not.
Memory errors can range from being trivial things that have no apparent effect at all, to crashing applications, to crashing the system, and worst of all, to corrupting data. Just depends what data is in that bit of memory, and when it happens.
Now you cannot guarantee against memory errors happening 100% of the time without using error-correcting memory, but even ECC memory can only correct for a single bit-flip at a time. If there is multi-bit corruption, you're still out of luck. (at least without true server-grade hardware)
That's why people have said that Ryzen doesn't have "true" ECC support. It can correct single bit errors, but will only report multi-bit errors. It does not halt ("crash") the system when a multi-bit error is detected to prevent potential data corruption from spreading.
For most people though, all you need to worry about is whether the system can read & write to all locations in memory without any errors occurring due to an overclock.
Errors caused by other factors are comparatively rare - especially if you don't keep the system running 24/7 (where errors might accumulate over time), so if one of these memory tests fail, it's pretty safe to say that the issue is being caused by an unstable overclock.
SmokedMeat
Gamer™
Just got back from Microcenter!
Ryzen 1600
Asus Prime B350 Plus
G-Skill Ripjaws DDR4-3200
I was planning on getting the Corsair DDR4-3000 memory, but the clerk told me he's using the G-Skill in his setup. Plus it was less money.
Hopefully I'll be able to install everything tomorrow after work.
Ryzen 1600
Asus Prime B350 Plus
G-Skill Ripjaws DDR4-3200
I was planning on getting the Corsair DDR4-3000 memory, but the clerk told me he's using the G-Skill in his setup. Plus it was less money.
Hopefully I'll be able to install everything tomorrow after work.
RumblingRosco
Member
Thanks for the very thorough explanation. I'm learning a lot in this thread and it has been a ton of fun/very helpful.
Congrats. I've always been jelly there isn't a microcenter near me. So many good computer hardware deals today which are in store only.
Got my 1600x, Asus Strix B350 build put together. With updated bios I can get 16gb Corsair Vengence CMK16GX4M2B3000C15 to 2666 with DOCH settings. Anything higher and I get 3 beeps and occasional crash during startup. I'm going to try Dominator CMD16GX4M2B3000C15 (Samsung B-die) tomorrow hoping for stable 2933.