not for me, I’m comfortable doing that stuff. mine will be used for live music on sometimes hot stages, stability is priority 1.
Run UEFITool 0.26.0 and open the BIOS file (eg. X3MSTX_1.58) and do a search for something like "OC Tweaker" and it should come up with something like:
Double click on it to jump to the "PE32 image section" and right-click on it and select "Extract as is…" and save it as something like X3MSTX_1.58_Setup.sct
Now run Universal IFR Extractor to extract into a human readable form.
You’ll need to search around a bit to get a tutorial on making BIOS mods but be very careful as you can brick your machine.
You can use my custom compiled version of flashrom to flash a modified BIOS.
I’m scared of bricking my motherboard so I’ve tried throwing some money at it again. I’ve ordered an alpenfohn blackridge with a notcua na-12x15 i found on ebay for £50 so I’ll try a couple of different fan configurations and let you know the results when its delivered
cool, I haven’t had time to play with my 12x15 yet as I’m still multi-tasking (installed custom LEDs + building a portable case for my setup), I’ll report back when I have too.
yes I can’t get it stable beyond 3466 at stock CL16 timings / 1.35V (have not yet tweaked timings but that is next). Dram calculator shows I’d need over 1.4V to reach that so that is pretty good. but they do get very hot - do yours too?
what specific kit are you running? I’m running Crucial Ballistics DDR4-3200 16x2 (BL2K16G32C16S4B) and managed 3533 16-17-18-36 1T but I can’t get 3600 stable at any timings.
you can try to steal mine if it helps but my account is too new to post links, apparently. hopefully cheating the system doesn’t break the rules
i.imgur.com/2d5D6cj.png
they’re not that impressive but DRAM calculator was useless for me too, I just did everything manually.
Exact same kit, and same hard ceiling 3533mHz at stock CL16 timings. I haven’t played with the timings yet but I’ll check yours out thanks (I have Ryzen DRAM Calculator but haven’t had time yet).
In fairness pretty good for 1.35V. If we could have 1.4V we should be able to push harder. Does anyone know of other voltage hacks like this, ie. has it been done, maybe on another board? Should be possible but bricking risk is a danger.
BTW does your RAM get really hot too?
I don’t see a temperature sensor in HWInfo so I’ve got no idea on RAM temps, but it doesn’t get hot enough to have stability issues. ¯_(ツ)_/¯
I actually signed up for an account here in the hopes someone already made a modded BIOS that allows >1.35v, but tbh I’m not even sure how well Micron B-Die scales with voltage. I’ve only ever played with Samsung B-Die and a few Hynix kits.
same (literally Hynix and B-Die recently), are we the same person? do you ever see us in the same room?
for our E-die (I assume that’s what you meant), I’m going by DRAM Calc which suggests we should be running at over 1.4V for 3533Mhz (and it won’t even give any higher freqs right). So I assume there’s a bit more in it with 1.4V.
re. temp, I have an IR (non-contact) thermometer that read over 74degC, which I’m not happy with. Like yours it seems to be stable, but … Not sure that reading was accurate, but it did seem extra hot to touch.
same (literally Hynix and B-Die recently), are we the same person? do you ever see us in the same room?
for our E-die (I assume that’s what you meant), I’m going by DRAM Calc which suggests we should be running at over 1.4V for 3533Mhz (and it won’t even give any higher freqs right). So I assume there’s a bit more in it with 1.4V.
re. temp, I have an IR (non-contact) thermometer that read over 74degC, which I’m not happy with. Like yours it seems to be stable, but … Not sure that reading was accurate, but it did seem extra hot to touch.
I hope for your sake you’re not me
fairly certain my kit is Micron B-Die unless I’m reading it wrong.
i.imgur.com/j0Bv3kT.png
RAM gets more temperature sensitive when it’s OCed but I have a hard time imagining 1.35v gets you anywhere near aggressive territory. DDR4 by spec is fine until 70c, at least, and "hand hot" is only ~40-50C, anyway.
Question guys probably a stupid question but does this mean we won’t get ryzen 5000 support on our motherboards aswell?
https://hothardware.com/news/amd-prevent…000-cpu-on-x370
I haven’t heard anything definite, but that link probably isn’t relevant to us. The 300 series desktop motherboard chipsets are pretty old, while the X300 mobo is fairly recent. Also my understanding is that the A300/X300 Deskmini boards don’t actually have chipsets, but run entirely on the APU (with perhaps some supporting circuitry, but not at the full chipset level). That’s why eg. they’re limited to the PCIe lanes the APU supports. So the ‘300’ in the DeskMini boards is not comparable do the 300 desktop chipsets.
I suspect 5000 support will come eventually to DeskMini, has anyone heard anything?
Thanks, this is supposed to fix USB issues, direct link: https://download.asrock.com/BIOS/AM4/X300M-STX(1.60)ROM.zip
Just a heads up, unlike the 1.46 BIOS the 1.60 BIOS does not apply FCLK frequencies above 2000 correctly, you will end up in 2:1 mode.
@The_Pook Those sticks you think are Micron RevB, are they 8Gigabit or 16Gigabit modules on them? As far as I am aware the 8s are very poor overclockers with the 16s being ok.
According to this thread, 1.2.0.1 (Patch A) may not be ready for prime time, 1.2.0.2 seems to be the non-beta version (when we get it): https://www.reddit.com/r/Amd/comments/md…atch_a_and_1202
Has anybody tried it? @ Monk, how’s it running for you besides the mem clock issue?
Memory stability has regressed compared to 1.46. With a set of Crucial BL2K8G32C16S4B and a 4650G I was able to validate (800% HCI Memtest and hours of Linpack 6GB) 4133 20-22-22-44 1T no problem whereas now the same and even looser timings at 4000 are resulting in BSOD and system resets. I have not tried lower clocks yet, I should have time this weekend to take a look at that. I managed to downgrade to 1.46 without issue. Going to revalidate the 4133 settings to make sure I was not the problem and have damaged something.
XMP on the same kit works just fine on 1.60 as it did on 1.46 with no measurable difference in performance on the few benchmarks that spit a score out that I have run, primarily Firestrike and Timespy. Those XMP tests were run at fixed clocks/voltage (3.8 @ 1.3) to disable PBO/XFR and stop them from affecting the results.
I have been fortunate in that I have not experienced any of the reported USB / Audio issues so cannot comment on that.
Overall from the 3 BIOS versions I have used 1.46G has been the best with 1.60 the worst and 1.40 the device shipped with being closer to 1.46 than 1.60 as that managed 4066 at the same timings I was running for 1.46.
I’m amazed you got clocks that high (even with loose timings). Did you check inside Windows that they really were as requested in the BIOS? Wondering if it down-clocked or changed timings behind your back. How does it compare latency-wise?
I got the USB issues on my X570 desktop board (external USB drives disconnecting randomly). I got rid of all my external drives (not for that reason, but it helped : ). From reading up on it, it seems the cause was the PCIe bus and USB traffic getting too high, causing uncorrectable transfer errors that caused the USB controller to reset. This was supposed to be worse on PCIe Gen 4 due to the doubled bandwidth, that probably explains why we have fewer (no?) issues on Gen 3.
Here’s a quick and dirty comparison for 3200(XMP) / 4000 / 4066 / 4133 on my system. CPU was locked at 3.8GHZ 1.3v, VSOC 1.15v except under XMP where I left it at auto. The latency should be lower if I allow the CPU to run at auto as it is my understanding that AIDA’s latency benchmark is affected by clockspeed. (I suppose this is something else to validate 
)
imgur /a/YbyE4Dm (mods: I hope this is acceptable)
As you can see there is a regression at 4133 in both AIDA and Firestrike but I am confident that the reported clocks are accurate as it was ZenTimings along with the GPU Memory Clock in HWINFO that alerted me to the issue under 1.60 with clocks not setting correctly. I should also note that it seems I was using BIOS L1.46 not 1.46G as I thought originally. I will flash 1.46G when I have some free time and see how that goes, along with doing some more thorough benchmarking.
I’ve just figured out how to look at your pics :). direct link: https://imgur.com/a/YbyE4Dm
That’s impressive, I didn’t expect your latency to be that good with the loose timings. I will have to try something similar.
BTW, I was getting regular random reboots on my 4750G under multi-core load. I’ve had little time to play with the system recently as I’m building a whole portable rig around it, but of course suspected the BIOS, PSU, the memory, or even a faulty CPU (which would suck as I imported it from Korea).
Turns out this may have been a Radeon driver issuer. There’s an update that my existing installation hadn’t detected. If you have random reboots, give it a shot: https://www.amd.com/en/support/apu/amd-r…amd-ryzen-7-pro
source: https://www.reddit.com/r/ASRock/comments…kmini_x300_and/
EDIT: yep, seems fixed (yay).
For what its worth, you should NEVER expect tweaked Memory timings to work across ANY AGESA/Microcode and MOST BIOS updates.
Considering change done to the interface and clock handling or countless other interconnected modules, its a pipe dream.
It never was a thing on Intel, and it will never be a thing on AMD.
Countless times I went super crazy and manually tightened down each and every primary, secondary and tertiary timing’s only to have it bork on the next BIOS/AGESA/Microcode update.
Expect to have to manually tune after each update, for those times when it just works, that’s LUCK!
Yes I have had that luck myself, but I do NOT EXPECT it and have always PLANED on having to retune my timings after each update.
With that out of the way, once I figure out how to flash the X300 BIOS onto my A300 I want to give a 5xxxG a spin, I look forward to it.