Personal computing discussed

Moderators: renee, Flying Fox, Thresher

 
UnfriendlyFire
Gerbil Team Leader
Topic Author
Posts: 285
Joined: Sat Aug 03, 2013 7:28 am

RAM is stable with 3200 MHz XMP, fails to boot when I slightly tighten timings even with volt increase

Sun Sep 15, 2019 6:54 am

CPU: Ryzen 1600

Motherboard: Asrock B450M Pro4 (BIOS version updated to 3.50)

RAM: ADATA XPG GAMMIX D10 (2x8GB), what Thaiphoon burner reported: https://imgur.com/a/NQideja

I tried CL16-16-16-16 with 1.40V DRAM, 1.10V SOC, Gear Down enabled and Spread Spectrum disabled, still wouldn't boot to Windows: https://imgur.com/a/aH9AVVg

But the XMP 3200 MHz with CL16-18-18 is stable at 1.35V DRAM and auto SOC voltage. Is there something I'm missing? I could bump the SOC voltage up, but I've read that many motherboards may deliver a higher voltage than what the BIOS and software utilities report (Claim: https://www.gamersnexus.net/guides/3251 ... fe-voltage and someone refuting it after suspecting that the voltage measurements was done incorrectly: https://www.reddit.com/r/Amd/comments/8 ... e/dv03gjy/ ).

Update: I decided to load up the XMP profile, bump the speed to 3333 MHz with the same XMP timings, set DRAM voltage to 1.40, SOC voltage to 1.1 and enable Gear Down. I'm able to boot into Windows and currently conducting stability tests. If it works, I then plan on disabling Gear Down and run the tests again.
 
Vrock
Gerbil God
Posts: 22385
Joined: Tue Jan 15, 2002 7:00 pm
Location: The Land of the Looney Lolcats

Re: RAM is stable with 3200 MHz XMP, fails to boot when I slightly tighten timings even with volt increase

Wed Sep 18, 2019 10:44 am

I have this same board (for all of 3 days now) and some Vulcan Z DDR4 3200. It absolutely refuses to run at the XMP setting, in fact I can't get the RAM over PC2866 speed.

https://www.asrock.com/mb/AMD/B450M%20P ... cification

It seems that "PC3200" RAM is quite picky about motherboards, go figure.
I'm not altogether on anyone's side, because no one is altogether on my side.
 
just brew it!
Gold subscriber
Administrator
Posts: 53463
Joined: Tue Aug 20, 2002 10:51 pm
Location: Somewhere, having a beer

Re: RAM is stable with 3200 MHz XMP, fails to boot when I slightly tighten timings even with volt increase

Wed Sep 18, 2019 1:01 pm

Vrock wrote:
It seems that "PC3200" RAM is quite picky about motherboards, go figure.

Or vice-versa. TBH I'm not sure which component is being picky here.
Nostalgia isn't what it used to be.
 
dragontamer5788
Gerbil Elite
Posts: 516
Joined: Mon May 06, 2013 8:39 am

Re: RAM is stable with 3200 MHz XMP, fails to boot when I slightly tighten timings even with volt increase

Wed Sep 18, 2019 1:16 pm

just brew it! wrote:
Vrock wrote:
It seems that "PC3200" RAM is quite picky about motherboards, go figure.

Or vice-versa. TBH I'm not sure which component is being picky here.


Its Ryzen Gen1 (1600). Its a CPU that's known to be picky with RAM.

Ryzen+ was better with varieties of RAM. The motherboard provides power, so yeah its somewhat important. But the CPU contains the memory controller, which is vitally important to RAM handling.

3200 MHz was the best most people got on Ryzen Gen 1, with some people getting lucky with 3466 with high-quality Samsung B-dies. IMO, you should probably do an extended test to really make sure 3200 MHz is actually 100% stable with a 24-hour Memtest86.
 
Vrock
Gerbil God
Posts: 22385
Joined: Tue Jan 15, 2002 7:00 pm
Location: The Land of the Looney Lolcats

Re: RAM is stable with 3200 MHz XMP, fails to boot when I slightly tighten timings even with volt increase

Wed Sep 18, 2019 2:08 pm

just brew it! wrote:
Vrock wrote:
It seems that "PC3200" RAM is quite picky about motherboards, go figure.

Or vice-versa. TBH I'm not sure which component is being picky here.

Well said, it's odd to me that in this day and age, this is still a thing. I mean, in the day of VIA KT266 and Nforce, I would have expected this. ;)
I'm not altogether on anyone's side, because no one is altogether on my side.
 
The Egg
Gold subscriber
Minister of Gerbil Affairs
Posts: 2920
Joined: Sun Apr 06, 2008 4:46 pm

Re: RAM is stable with 3200 MHz XMP, fails to boot when I slightly tighten timings even with volt increase

Wed Sep 18, 2019 2:31 pm

You can call me a fuddy-duddy, but that's part of why I bought RAM that runs dead-nuts 2666mhz on an H370 chipset that doesn't allow OC'ing. A small amount of performance is lost with the slower RAM, but I can just select the XMP profile and go, no screwing around, no worrying about stability (though I still test). The CPU's built-in turbo frequencies are also good enough for what I'm doing. *shrug*
 
just brew it!
Gold subscriber
Administrator
Posts: 53463
Joined: Tue Aug 20, 2002 10:51 pm
Location: Somewhere, having a beer

Re: RAM is stable with 3200 MHz XMP, fails to boot when I slightly tighten timings even with volt increase

Wed Sep 18, 2019 2:46 pm

I KNOW I'm a "fuddy-duddy". For the past decade or more I've been running bog-standard stock settings on any system I actually care about. With ECC DIMMs (where supported).
Nostalgia isn't what it used to be.
 
just brew it!
Gold subscriber
Administrator
Posts: 53463
Joined: Tue Aug 20, 2002 10:51 pm
Location: Somewhere, having a beer

Re: RAM is stable with 3200 MHz XMP, fails to boot when I slightly tighten timings even with volt increase

Wed Sep 18, 2019 2:46 pm

I KNOW I'm a "fuddy-duddy". For the past decade or more I've been running bog-standard stock settings on any system I actually care about. With ECC DIMMs (where supported).
Nostalgia isn't what it used to be.
 
dragontamer5788
Gerbil Elite
Posts: 516
Joined: Mon May 06, 2013 8:39 am

Re: RAM is stable with 3200 MHz XMP, fails to boot when I slightly tighten timings even with volt increase

Wed Sep 18, 2019 3:16 pm

The Egg wrote:
You can call me a fuddy-duddy, but that's part of why I bought RAM that runs dead-nuts 2666mhz on an H370 chipset that doesn't allow OC'ing. A small amount of performance is lost with the slower RAM, but I can just select the XMP profile and go, no screwing around, no worrying about stability (though I still test). The CPU's built-in turbo frequencies are also good enough for what I'm doing. *shrug*


The InfinityFabric speed on Zen and Zen+ chips is synchronized with the RAM clock. So a 2666 MHz RAM will run the Infinity Fabric at 2666 MHz, which limits the speed of cross-CCX communication.

Its a quirk of AMD, but a RAM overclock effectively overclocks part of the chip itself. Its way more important to get fast RAM on AMD than it is to get on Intel.

just brew it! wrote:
I KNOW I'm a "fuddy-duddy". For the past decade or more I've been running bog-standard stock settings on any system I actually care about. With ECC DIMMs (where supported).


Once a system passes a 24-hour test, I think its probably going to be pretty stable :-) But yeah, it does take a bit of work to run Memtest86 for that long.
 
Vrock
Gerbil God
Posts: 22385
Joined: Tue Jan 15, 2002 7:00 pm
Location: The Land of the Looney Lolcats

Re: RAM is stable with 3200 MHz XMP, fails to boot when I slightly tighten timings even with volt increase

Wed Sep 18, 2019 4:08 pm

just brew it! wrote:
I KNOW I'm a "fuddy-duddy". For the past decade or more I've been running bog-standard stock settings on any system I actually care about. With ECC DIMMs (where supported).

It's the ECC RAM that puts you squarely in fuddy duddy territory, you know. ;)
I'm not altogether on anyone's side, because no one is altogether on my side.
 
just brew it!
Gold subscriber
Administrator
Posts: 53463
Joined: Tue Aug 20, 2002 10:51 pm
Location: Somewhere, having a beer

Re: RAM is stable with 3200 MHz XMP, fails to boot when I slightly tighten timings even with volt increase

Wed Sep 18, 2019 4:24 pm

Vrock wrote:
just brew it! wrote:
I KNOW I'm a "fuddy-duddy". For the past decade or more I've been running bog-standard stock settings on any system I actually care about. With ECC DIMMs (where supported).

It's the ECC RAM that puts you squarely in fuddy duddy territory, you know. ;)

That and the RAID-1 /home area. :wink:
Nostalgia isn't what it used to be.
 
UnfriendlyFire
Gerbil Team Leader
Topic Author
Posts: 285
Joined: Sat Aug 03, 2013 7:28 am

Re: RAM is stable with 3200 MHz XMP, fails to boot when I slightly tighten timings even with volt increase

Wed Sep 18, 2019 6:53 pm

Vrock wrote:
I have this same board (for all of 3 days now) and some Vulcan Z DDR4 3200. It absolutely refuses to run at the XMP setting, in fact I can't get the RAM over PC2866 speed.

https://www.asrock.com/mb/AMD/B450M%20P ... cification

It seems that "PC3200" RAM is quite picky about motherboards, go figure.


I did run Memtesthelper on the 3200 XMP for about 8 hours, another ~13 hours after adjusting some of the sub-timings, and another ~8 hours after adjusting two last set of sub-timings without errors. Any attempts at using 15-18-18-36 or 16-17-18-18-36 caused memory errors.

I guess when I upgrade to a Zen 2 or 3 in the future, maybe I could get better RAM speed/timing.
 
UnfriendlyFire
Gerbil Team Leader
Topic Author
Posts: 285
Joined: Sat Aug 03, 2013 7:28 am

Re: RAM is stable with 3200 MHz XMP, fails to boot when I slightly tighten timings even with volt increase

Wed Sep 18, 2019 7:21 pm

dragontamer5788 wrote:

IMO, you should probably do an extended test to really make sure 3200 MHz is actually 100% stable with a 24-hour Memtest86.


I've been running 8 hour stability test sessions with the memtesthelper (usually gets to around 1500%-1700% passes by that point by spamming multiple instances of HCI MemTest on each CPU thread and giving them around 2GB separate chunks of RAM to test for a total of 14.4GB of tested RAM)
 
dragontamer5788
Gerbil Elite
Posts: 516
Joined: Mon May 06, 2013 8:39 am

Re: RAM is stable with 3200 MHz XMP, fails to boot when I slightly tighten timings even with volt increase

Wed Sep 18, 2019 7:34 pm

UnfriendlyFire wrote:
dragontamer5788 wrote:

IMO, you should probably do an extended test to really make sure 3200 MHz is actually 100% stable with a 24-hour Memtest86.


I've been running 8 hour stability test sessions with the memtesthelper (usually gets to around 1500%-1700% passes by that point by spamming multiple instances of HCI MemTest on each CPU thread and giving them around 2GB separate chunks of RAM to test for a total of 14.4GB of tested RAM)


That's better than nothing, but Memtest86 is nice because it is its own operating system. So all your RAM is tested with Memtest86.

The RAM that the Windows Kernel is using won't be tested with HCI MemTest, or other tools. The main benefit of HCI MemTest is that you can use your computer while waiting for the result. Its a good quick-and-dirty, but a full Memtest86 would be more exhaustive for sure.
 
synthtel2
Gerbil Elite
Posts: 954
Joined: Mon Nov 16, 2015 10:30 am

Re: RAM is stable with 3200 MHz XMP, fails to boot when I slightly tighten timings even with volt increase

Wed Sep 18, 2019 10:01 pm

I haven't had great experiences with any memory test for OC confirmation; error rates bad enough to make the system unusable long-term make it through even when all tests say it's good. What I try to do now is have all tests pass something significantly more stressful than the 24/7 config. Any given timing might not be the limiting factor, so frequency and voltage are the good variables to use (so if 3333 1.4V is as far as it'll go at current timings then the 24/7 options keeping the same timings would be 3200 1.4V or 3333 1.45V, or potentially much further off than that if you think you're frequency-bound rather than latency-bound).

If you're really going for maximum performance, you might consider loosening timings a touch to get better clocks at a given DRAM voltage. The latency in nanoseconds stays about the same when you do this, aside from the quantization error you get because timing adjustments are so coarse.

SOC voltage doesn't affect the timings you can get and 1.1V should be plenty for 3200 on most dies. If you're worried about it (and if you're going for 3333+ it's much more likely you should be) you can drop it in isolation until it fails to see how much margin you've got.

Who is online

Users browsing this forum: No registered users and 5 guests
GZIP: On