[kithylin]

I suppose I'll write here and perhaps I can get some comments. I'll apologize up front if anyone has covered this.. it's a big thread.

Short story time. I had owned an Intel X99 system for a long time since 2015. It was acting up on me a lot, more and more.. motherboard dying. I needed to replace it and at the time I had $300 max at the time and it was black friday 2018 so I went with the cheapest AMD setup I could afford that was equivalent to the I7-5820K setup I had. So my first modern AMD system since 2008 ended up being a Ryzen 5 2600, and a MSI Gaming X470 motherboard. At that point I just needed a computer that worked at all and didn't die on me. This system worked okay for a while and everything was fine. Then Stimulus 2020 came and I made an educated guess based on AMD's history that they would be releasing "Some sort of big bang to end out AM4" like they did with piledriver on AM3+ and I gambled that "whatever big chips that AMD finishes off AM4 with" will likely support 4000+ Mhz ram easily. So I bought a set of 4x8GB Patriot Viper DDR4-4400 with Stimulus. All 4 sticks are single-rank samsung B-Die ram. Initially I got this ram to run in my X470 board with the R5-2600 @ 3533 Mhz stable @ 14-14-14 with some tuning of volts and everything was great. I was also able to run windows 7 64-bit in this system dual-boot and everything was fine. I had a basic system: Single 1080 Ti + NVME SSD and that's the only storage I had and the only expansion cards. No SATA anything, everything else USB.

Well.. this is where it gets to why I'm writing in this thread and where I am now. With Stimulus I wanted other things. I wanted to go nvidia SLI (That's a different subject I don't want to discuss here) so I wanted an AMD board that was SLI compatible and something with a big VRM. The MSI X470 Gaming board was fine but I was nearly overloading it's VRM with an overclocked R5-2600. The VRM's would often run around 80c even with direct airflow after a few hours of crunching data with Adobe Premiere. I was looking on ebay and I happened to find someone selling a ASRock X370 Taichi for $75 free shipping with all the original accessories and retail box and everything. I jumped on it and it's mine now. But I've had a number of problems with this system ever since I switched to this board.

Problem #1: Memory speeds. I know this ram and this CPU could do 3533 @ 14-14-14 with all 4 sticks of 8GB installed (full 4x8GB) and it was stable for around 7 months in the X470 board. But here on this X370 Taichi it won't even POST at all at those speeds. I've even tried it with 1 ram stick installed, nope. 2 sticks installed, nope. The fastest I can get this system to run stable is 3200 @ 12-13-13 with either 2 sticks installed or 4 sticks installed. If I run at this speed, it's perfectly fine. It's sad I can't run at the fast speeds I could previously but.. at least it works. It's been stable for several months @ 3200-12-13-13. I guess it's a limitation of this board. I don't even know. And I've had numerous friends spend hours and days helping me with this. No manner of SoC volts, DRAM volts, bios option tuning, timings, nothing will let this CPU+RAM combo run at 3533 Mhz in this board for any reason. Any thoughts on #1?

Problem #2: General instability (Mostly fixed). When I first got this board from ebay, the general thing most people do is "Let's update the motherboard to the latest bios! That's the best!" so the seller shipped it to me on the latest bios, which for this board was 6.20 at the time. Well I didn't at first know what it was and I was struggling for months to figure out what all the crashing and random blue screens was about. Finally I saw the asrock bios page about the disclaimer not to use 6.20 with my R5-2600. So I back-flashed down to bios 5.60 with AGESA Combo-AM4 1.0.0.1, and after switching there MAGICALLY all the instability melted away and I can run full 4150 Mhz all-core OC again and everything's been perfectly fine and stable for months now. I suspect using an even older bios may result in even higher stability for #2 and may get me my ram speed back I had on X470 before. But I can't figure out a way to reliably flash back to anything older without bricking the board. Any thoughts on Back-flashing? I do have access to a eeprom USB flasher and know how to use it. I think the bios chips are soldered on this board though (sadly) and that's a bit of a bear to get flashed sometimes. I have the clip-on thing but it doesn't work 90% of the time.

And Problem #3: Minor and I do understand up front that I'm probably the only one to ever care about this, but.. I can't run windows 7 on this system. Even with all the drivers loaded and slipstreamed it for USB-3 and NVME on install it will completely at random blue screen and it's different every time. Sometimes at desktop sometimes in a game, but it always blue screens and never works. Win7 was flawless on the X470 board. It ran all the time and never crashed and never blue screened for months. But it just won't run on this board. Does anyone have any input on #3?

Otherwise: The X370 Taichi is a fantastic board and the big VRM is much better for my overclocked R5-2600, peaking at 51c under heavy workloads vs the mid-80's C with that MSI board I had at first. I hope and pray that ASRock releases a bios for 5000 series for this board some day. I really don't want to give it up. The VRM's on the X470 and X570 taichi aren't as big as they are on this X370 Taichi and I'd like to go up to an overclocked 5950X some day.

 

[kithylin]

MSI X470 Gaming all have L-shape layout heatsinks and they look huge, where the VRM and temps so high I'm surprised, would new thermal compound help? additional thermal pads surround them too???

Problem #1: I think is related to T-topology versus the suspected Daisy-chain in X470. In short their behavious is completely different, all the settings must be re-searched, re-find, re-lookup between them, consider them no parameters share. quick and dirty fix is tRP and tRAS +2 each. or more at times, look up history pages for wisdoms.

Taichi x370 somehow requires more voltages to push ram, I think. Again something to do with T-topology thing, I think.

Problem #2: Am using zen2 and like you 4x8GB, and random reboot even when idling doing nothing versus gaming heavy load whole day no problem, crazy stupid motherboard. bios 6.2 was very bad for me, but other users are fine, might be a buggy flash, then I up to 6.2A and all much better. There's official 6.4 now, I highly recommend it. I'm not simply because warranty is over and I don't want to risk it.

What's your voltages? mind sharing zen timing and all other usual screen shots?

I run VDDCR_SOC 1v only inside the indepth voltages, then run uncore SOC 1.25v and 1.2v respectively
cldo_VDDP is 0.9v
both VDDG is 0.95v

oh, you're zen+, different, nevermind.

I could post some photos if you want I suppose. I recently had a few minor instability issues and I've stepped it up from 12-13-13 @ 3200 -> 14-14-14 @ 3200 now and seems to be better.

I knew it! I have not found a single user who has managed to confirm passing Prime95 with those settings on a Zen 2 CPU on the X370 Taichi. This mobo is not stable with Zen 2 at all!
This should ring alarm bells for any Zen 2 user on this thread, people. Instability at stock settings is inadmissible.

Most likely no one is reporting prime95 stability for anything in 2020 because most sane people stopped using prime95 for actual stability testing several years ago.

 

[kithylin]

I would like to know why you think that.
Prime95 is one of the fastest ways to find instability in CPUs. It's updated to squeeze every little bit of performance it can from Zen 2 cores and it fails me where other stress testing tools pass. This tells me Prime95 is better than other tests at finding instability.
Those "sane" people you talk about that are afraid to run Prime95 tests are the same people that think a PC is stable just because it can run games without crashes. Then a few months later find corrupt files in the file system and can't find the cause or when it happened. I use my PC for more than just games, so stability is a must, and this mobo is failing me on that.

The point of a stability test is to determine if the system will crash when doing a specific task that you normally perform with your computer. If you play games, play 5-8 different games for 4-10 hours a day and see if crashes playing games. If you do adobe premiere rendering then load up a render that takes 5 days to complete, then run that and see if it makes it through. In the case of Prime95 there is nothing that exists that loads a computer in the same way that Prime95 does. There's not even math calculations that will ever load the computer in the same way as how prime95 loads the system. Therefore there's no reason to use Prime95 because it's not actually "Testing anything". It's a fake test that just generates an unrealistic load for no reason what so ever. It doesn't matter if our computers are unstable in Prime95 because we'll never use our computers in the same way that Prime95 works anyway. You could have a computer that is completely and perfectly stable in games and adobe rendering and every other "common daily task" we could do with our computer but fail in 1 second in prime95. That doesn't mean the computer is unstable. The computer is stable. It's just not stable in some old fake test that doesn't represent anything.

 

[kithylin]

Now, if you don't use your PC for anything important and don't mind any of the above, sure, there are no big consequences in running unstable systems.

People using computers for important tasks like that aren't overclocking in the first place. If stability is that critical they won't risk it. Stock computer components won't have corruption issues and there's no reason to stress test them. Overclocking is only for random home computer users that play games or record gameplay footage to post on youtube or other non-critical tasks. It's good enough that home computers being overclocked are only stable in 99.99% of applications.

 

[kithylin]

Sure. Nothing wrong with that, as long as you don't store important files there.

I still stress test my systems, I just don't use prime95. 15 runs of cinebench R20 multi + 1-2 hours of converting videos in handbrake + play games for a few hours. If it passes all that it's good enough. Running Prime95 today is like taking your vehicle, putting it in park and rev'ing the motor to redline and holding it there for 4 hours. Will the vehicle survive? Probably. Is it a good idea? Probably not. Even if it did survive, would you be doing that to your vehicle on a regular basis? Probably not. The only time prime95 makes sense is if someone with a home computer is going to be doing some sort of mathematical computational task where they deal with millions of prime numbers on a daily basis. If that's what they intend to do with their computer, then sure use prime95. If they're just a gamer that watches youtube and streams and will never deal with computational tasks involving prime numbers then it's pointless.

 

[kithylin]

This is a really bad analogy.
Rev'ing an engine to redline for 4 hours is harmful for the engine. Running Prime95 on a CPU is a harmless operation. CPUs are designed to run these tasks 24/7 for YEARS.
If my system fails Prime95 consistently in less than 10 minutes. Then something is wrong with my system.

Prime95 artificially causes the computer's components to heat up and run hotter than any normal program could ever possibly heat up the computer. Nothing else gets the system as hot as prime95 does. It's a perfect analogy.

Or if they have important files they don't want to lose.
Or if they do work on that computer.
Or if they want an operating system that won't break down a few months down the line inexplicably with some corrupt file.

Just so you are aware: Prime95 does not test the stability of file storage or I/O with files, and it does not test any other program's functionality other than computing prime numbers. You could have a computer that's perfectly stable in prime95 but still crash in handbrake, adobe, compiling programs, playing games, recording video, etc. Just because your computer is stable for 24 hours in prime95 does not mean it still won't crash. It just won't crash computing prime numbers. People need to understand that about what they are testing. The reason I say all of this and the reason I stopped using prime95 years ago is because I had a computer several years ago back in 2010, a 6 core I7 system and It would sit there and run 24-hours in Prime95 non-stop with no issues, no crashing and it passed perfectly. Yet at the same time I could run the same computer on the same overclock through handbrake converting a video and the program would crash and end after 1-2 hours of conversion. And even though it was "Prime95 stable", it would repeatedly cause ARK: Survival Evolved to crash to desktop completely at random. Prime95 is not the "be all end all" test of stability. Just because something is stable in Prime95 doesn't mean jack squat.

 

[kithylin]

Running Prime95 for hours on a stock system is completely harmless.

A stock system yes I agree with you there. The issue is people are overclocking the system well beyond it's rated design parameters and then running prime95 on an overclock and that's where it can be dangerous sometimes. In general Prime95 (stock or overclocked) is going to cause all components to run at least +10c and sometimes +20c hotter than the system would ever see under any other type of "normal program". Sure stock chips can handle it and have protections built in. But why expose a chip to 90c constantly for a few hours when it will only ever max out at 75c under normal programs? It's just creating heat and making the system run super hot for no useful reason other than "Let's see how hot we can run it". It's also playing a gambling game of "Let's hope and pray that my processor's safety features actually work and it'll shut off the system before it sustains damage". What about people that buy low end AM4 boards with no heatsink on the VRM's and then plop in a 12 core chip and run prime95 on it and run the VRM's @ 125c for 4 hours. No one tells them that it's dangerous for their board to do that.

You don't seem to understand how a semiconductor works electrically. This has nothing to do with specific operating system tasks. The only thing that matters is if all the transistors are switching on time correctly before the next clock cycle comes. If a single one of the billions of transistors is not switching correctly, you have a potentially unstable system.

Don't forget that it's not just transistors and switching them. There's also the cache pools, caching algorithms, branch prediction, different instruction sets, etc. Every different game, program and process with the computer loads the processor and it's caching system differently. My issue is with people I see in the forums that ONLY test with Prime95. They run a system through Prime95 for 3-4 hours and claim "Oh it's Prime stable, everything in the system is stable." and then they stop testing and go do everything with the system. That's very common. At least 90% of people posting on OCN and other forums completely rely on prime95 exclusively. Prime95 is just one test (and not even a really great one). It's just.. I don't even know anymore. Prime95 has just become entrenched in people's minds as the only thing that matters with a computer and it's the only test everyone seems to talk about when testing a computer and processor when I'm reading the forums. :/

 

[kithylin]

Any news about X470?

Ryzen 5000 CPU support for older AMD motherboards has begun rolling out ahead of schedule

ASRock is first out of the gate, and Asus will follow suit in a couple of weeks.

www.pcgamer.com

Sent from my iPhone using Tapatalk

EDIT: NEVERMIND! I found a cached version of the news page. I have no idea why ASRock would delete the news article but they did. But google cached it first before it could be deleted so here we go: ASRock > News

So this IS coming and it is a reality.

 

[kithylin]

So here's my experience so far with the beta bios for this board: It's unstable as heck and don't use 5000 series in this board. The chip is running significantly reduced memory performance as much as -30% to -40% of what it should be capable of. The memory performance is so bad that these chips are no faster at ram speed than a 3000 series XT chips are. Which is terrible. The 5000 series chips should be quite a bit faster at ran performance in a proper (X570) motherboard. The PCIE top most slot runs at 8x instead of 16x and as if that wasn't bad enough the onboard audio (digital output or analog output, either one) switches to "banshee screaming at us from the void" completely at random and the only way to get sound back to normal is disable the sound output in windows sound control panel then re-enable it. If we go with a manual all-core overclock around 4.7 Ghz at 1.288 <-> 1.300 volts then the onboard audio is always in "wailing banshee mode" and no amount of disabling-enabling the sound output will help so that's not even usable at all.

I ran my 5800X at 0.940v for SoC and that enables me to run the ram up to 3933 Mhz @ 16-16-16-33-1T for my ram kit (single sided samsung b-die). Which according to a friend of mine with a 5800X and a ram kit that can do 4600 mhz, in a gigabyte X570 board he tried 3933 @ 16-16-16-33-1T too and saw 59-60 GB/s in all 3 (read, copy, write) for ram performance and latency of 49ns instead of the 64 ns I get in this X370 taichi board.

See here. I get almost half of his ram performance. Also even if I manually match the FCLK to the memory clock in bios, it still runs the ram on a divider instead of 1:1. This board will never run ram 1:1 for the ryzen 5000 series. I've tried everything and it just won't do it.

In general: If you want to run a 5000 series chip then you're wasting money and significantly crippling it's performance by using a 5000 series chip in this motherboard, even with the beta bios. Just spend the money and buy even a cheap budget X570 board instead.

EDIT: Here's my bios settings for voltages if anyone's wondering.

This is running Automatic mode + PBO with all limits maxed out.

 

[kithylin]

Check my previous post, just above yours.

Same mobo, same mem sticks, same profile, but the difference is huge.

Check your settings, something is probably messing up everything.

To back up my earlier claims with actual data now I have a post to share with folks in this thread. I went ahead and bought a X570 motherboard for my new Ryzen 5800X processor and right now I have it on my test bench in the other room playing with some older video cards and aiming at hwbot world records before I mount it into my main case with the 1080 Ti. But you can at least look at this and see the difference in CPU score / Physics score. You can go to the bottom and expand "General" and see but the summary is: ASRock X370 Taichi right side, new MSI Prestige X570 Creation left side. Same ram (other than 2 sticks in the X570 and 4 sticks in the X370). 3933 ram speed in the X370, 3800 in the X570. This demonstrates how the X370 platform is crippling the 5000 series processors if you use them in this board. It's running a whole +42.3% faster in the X570 board which is huge. Result

 

[kithylin]

You post is pointless cos we know that with this agesa IF is not able to get 3933,put taichi to 3800 and you will have 1:1 and same results.

That's not what I saw in my experience with the board. I ran the same processor at 3933 Mhz ram and 1967 Mhz FCLK for the X370 score above (Which is 1:1 by the way) and it was still -42% slower. I'm only trying to share facts and information but it's obvious now that no one here will actually listen to facts. I'm just trying to educate people that while 5800X does work on the X370 taichi with the beta bios the processor can not and will not run at it's full potential performance while on the X370 taichi with that bios. Apparently there's a reason why it's not posted to the public on the ASRock website. But whatever. I've shared the information. If people choose to ignore it then that's on them.

 

[kithylin]

So it turns out I can set my FCLK above 1600mhz without problems, when i set MCLK to anything above DDR-3200 is when i start getting the cold reboots...

When I had my 5800X in my X370 Taichi motherboard I was able to run it at 1967 Mhz FCLK no problem. Are you sure your ram can do 3200 mhz or higher? Are you giving it enough voltage? I had my CPU voltage @ 1.325 with LLC @ 3 for CPU + SOC @ 1.10v

 

[kithylin]

Would you say it's fairly stable to run with a 5600X and not worry about it?
I've got the Asrock X370 Killer SLI, and have the CPU on its way lol

I can not and will not say anything about voltage because I won't be liable if your chip dies on you because of something I said. However several people I have talked to have run their 5000 series chips (5800X & 5900X) @ between 1.30v <-> 1.325v for several months and they haven't reported to me any issues yet. The chips are only a few months old though so no one really knows if voltage above 1.30v will degrade them long-term or not.

 

[kithylin]

I'm only inquiring about overall stability of the BIOS on X370 with a 5000 series CPU, nothing more.
If it's a flakey experience then I'll probably sell the 5600X and stick to my 3600 for now.

On the X370 Taichi board with beta bios I upgraded from a R5-2600 -> 5800X and at first I had terrible stability. The entire system would do the "Turn off -> Turn on -> Turn off -> Turn on -> Turn off -> Turn on -> Turn off -> Turn on -> Turn off -> Turn on -> thing about 20 times and then finally decide to run (even at stock clocks / all AUTO for ram and CPU) Any attempt at overclocking anything resulted in a lot more Turn off -> Turn on -> Turn off -> Turn on stuff until eventually I had a bios tuned with the right voltage and it would turn on and run every time. There definitely was some initial instability and I didn't think the board would ever boot and run with the chip but I didn't give up and just waited it out for it to "Do it's thing" and eventually it came on and decided to run. Once I got all the bugs worked out and the system overclocked it would come on and start up and run when I pressed the power button though but it took a few days of tinkering with it to get it there.

 

[kithylin]

Huh, that's weird, I didn't have that issue at all. Just threw the sticks in (3733 b-die) and it POSTed first try. Did it's typical first boot bootloop that this board has done for every CPU, just restarts once as soon as it gets to the main screen and then its good.

Try moving the sticks around a bit, I find the memory being more stable in my board if the ram sticks are in A1/B1 vs A2/B2 like asrock recommends.

In my case moving the memory wouldn't of helped anything as it was perfectly stable for 1 year with the ram as-is with my R5-2600 chip. I would of never touched the ram or even tried to move it at all. If it has any sort of self-resetting issues then it's an issue with the chip and bios, not the ram.

 

[kithylin]

Bit the bullet CPU came today works great!
Didn't have much time to tune RAM but it works!
No CPU offset voltage settings and PCIe runs at 8x like everyone else.
Still super cool that this works and I can run a 5000 series CPU on a X370 motherboard.

Do note like I tried to tell everyone in this thread before however: You are running that processor around -40% of it's true potential by using it in a X370 board instead of an X570 board though.

 

[kithylin]

Huh?

My 5600X at stock, with loosely tuned 3466/3533 gets to around 95% of the scores of every 5600X (also stock) owner I've seen on any B550/X570.

I'm sure that I can't oc/uv the cpu as efficiently as they can (curve optimiser is a sweet thing, for sure), but I'm hitting 4.650MHz sustained (Noctua D-15, dual).

All in all, we took a calculated risk...we'll sure be -10 or even -15% compared to a fine tuned 5000+X570 user, but -40%? No way!

I started with my 5800X in my X370 Taichi with the latest beta bios and I had it running a manual all-core overclock of 4725 Mhz @ 1.325v with LLC-2, 1966 Mhz IF clock @ 3933 Mhz memory. I believe it was 18-18-18 on the timings on the ram but that shouldn't make this big of a difference. Then later I upgraded to a MSI Prestige X570 Creation motherboard and I now run it at 4750 Mhz all-core overclock. The same 1.325v @ LLC-2, and this time the IF clock at 1900 Mhz for 3800 Mhz ram and I have it at 14-16-14 ram now. So slower IF and ram speed, but a little faster timings, but not that much. I recorded 3dmark benchmark runs on the X370 Taichi and my new system and the 3dmark physics score is as much as +40% faster on the X570 board vs the X370 Taichi board. I posted in the thread but folks seemed to respond thinking either I'm crazy or lying or both so.. whatever. No one listens to actual facts in here or they refuse to acknowledge facts and data. I came from a Ryzen 5 2600 @ 4.2 Ghz all-core and using the 5800X in the X370 board there were several games (Valheim and Carmageddon Reincarnation) that would still run at the choppy 40-50 FPS from the 5800X that I got out of my R5-2600. I didn't see any noticable improvement at all in most of the games I play (1080p). But yet here on the X570 board both games run maxed out flat at a minimum of 80 FPS at all times (my monitor's vsync max) with my 5800X. It's most definitely very noticeably faster in the X570 board. I thought I would try to educate folks about this sort of thing and share my experience.. but I'm mostly laughed at and told I'm crazy despite seeing it with my own eyes. Personally I'm glad I went with a X570 board instead of staying with that X370 board for my new chip though.

The only thing that changed was the motherboard. Same video card at the same clocks. Same ram sticks. Same custom water loop. Same case. Same power supply.

EDIT: Well, the operating system changed. I was on windows 10 64-bit before on the X370 Taichi and I'm now on Windows 7 64-bit on the X570 board so I'm not sure if that has much to do with it.

 

[kithylin]

In any case, a 5000cpu+X570mobo will definitely be better than a 5000cpu+X370mobo. Noone argued about that...but 40% clearly shows there was some other reason.

Or my thoughts: The beta bios for the X370 won't let the 5000 series chips run properly at their full speed / possibly isn't optimized for them. Personally I think that the beta bios just adds microcodes and nothing else then lets the 5000 series chips run as if they were 3000 series chips in the X370 Taichi. But that's just my thoughts.

 

[kithylin]

Hey, anyone tried to apply negative offset to a 5xxx on 6.61?

I'm not sure it works...

Any other way to keep the cpu cooler?

You could try Liquid Nitrogen. I hear it keeps CPU's really cold. :>