Overclock.net banner
7681 - 7700 of 8526 Posts

·
Registered
Joined
·
18 Posts
For me 3702 is more stable so far, I was able to push my crappy bronze 5900x from CO -15, -15, -30* to -20, -20, -30* without any issues.
3601 was prone to idle reboots at -20, -20, -30* with this CPU.
2hrs of corecycler (small, avx2) + OCCT + normal usage - rock solid so far, time will tell.
However I get slightly lower CB20 mt score 8611 (3702) vs 8639 (3601) with the same settings, all the settings were dialed in manually just in case.
 

·
Registered
Joined
·
1,120 Posts
Maybe not so stable, just had my first idle reboot in several_months! Judging from that, 3501/3601 was more stable, also booth booted with a more wide selection of memory settings. This one was finicky.

Same Curve Optimizer, only thing I changed was the three aforementioned memory settings
This one accepted all my prior settings.

For me 3702 is more stable so far, I was able to push my crappy bronze 5900x from CO -15, -15, -30* to -20, -20, -30* without any issues.
3601 was prone to idle reboots at -20, -20, -30* with this CPU.
2hrs of corecycler (small, avx2) + OCCT + normal usage - rock solid so far, time will tell.
However I get slightly lower CB20 mt score 8611 (3702) vs 8639 (3601) with the same settings, all the settings were dialed in manually just in case.
For me, for now, i went from -10 all to -15 all. Increased all core effective frequency from ~ 4450 to ~ 4550. Single core to ~ 4965. This is currently bringing me to 11413 multi and 639 single in R20 (with some background tasks running)


On a side note, ive noticed the act of setting just PBO to enabled or manual reduces single core effective frequencies until you bring them back up with CO. i believe that was the case with 3602 as well, tho i didnt pinpoint the cause then. With the system totally default, single core effective frequency with 3602 maxxed just at 4899mhz (grr), but with 3702, it has reached 4913mhz. the act of enabling pbo for either stripped off roughly 100mhz from the single core boost effective frequency.

Boost clock override still doest do anything, and its likely because the 5950x is already floating at the max set voltage for single core. My 5800x ran single core with less set voltage, roughly 1.43v. enabling a 200mhz overide brought that uo to the 1.48 i expected, anlong with the effective frequency increases. That one sucked at fclk tho.
 

·
Registered
Joined
·
18 Posts
Boost clock override still doest do anything, and its likely because the 5950x is already floating at the max set voltage for single core.
Same here. 0, 200 or anything in between doesn't seem to matter. Best cores just won't go past ~4900 effective clock.
Temp stays at 60C when running CB20 single threaded with pinned affinity.
 

·
Old to Overclock.net
Joined
·
361 Posts
I‘ve just removed aourmory, aura and iCUE from the os cause this apps caused….COIL WHINE, somewhere in socket/lower vrm shoulder; rythmical I might say, and I would easily hear it even with all my fans.
So much for the trash rgb utility…

Like, *** was that? I know leds and rgb is trash of the trash of DIY PC, but to cause coil whine?
 

·
Registered
Joined
·
563 Posts
I don't care about beta status, but what does kinda raise a question mark for me is why we are being given old AGESA patches. Both MSI and Gigabyte are weeks ahead of ASUS with the latest patches from AMD.
Gigabyte is still on AGESA 1.2.0.2. So not sure where you are getting your view on "weeks ahead on the latest AMD patches" from.

MSI is ahead as they are say that they are using 1.2.0.3.b compared to Asus on 1.2.0.3a. But after flashing, when you check the version in Aida64, it still shows as 1.2.0.3.a! So who knows if this patch b is actually implemented in the MSI firmware, but I don't think Asus is behind the other motherboard vendors.

Edit: The claimed USB fixes with 1.2.0.3.b on the MSI bios do not work:
Code:
https://www.reddit.com/r/Amd/comments/o4or1w/msi_releases_agesa_1203b_bios_for_x570b550_boards/
 
  • Rep+
Reactions: stimpy88

·
Registered
Joined
·
563 Posts
Thanks Safedisk. Sadly this is yet another version which refuses to boot when I enable DOCP. Every single version of your BIOS versions work except 3702 and 3601. Seem to hate my Crucial Ballistix 3600 CL16 memory (2 x 32GB DIMMS).

Previous bios (3601) used to end the post attempt on code 07. This (3702) seems more random, but often ends on code 64. Very odd!

I should finally add that whilst I realise my memory is not on the QVL, I've never had a problem until 3601.
There might be a DRAM timing setting which is marginal and does not stabilize with additional voltage. Do you have a copy of your Zentimings to share?
 

·
Registered
Joined
·
563 Posts
I think one of the bug fixes might be related to cpu voltage. 3602 my single core voltage liked to be below more around 1.45, usually below before setting a voltage offset. all-core liked to be 1.25v before offset. Offset never really increased the set voltage by the desired offsert either, always slightly less.

with 3702, single core voltage is just below 1.47v while all core has bumped up to 1.3v before any offsets. partucularly handy for curve optimizer stability. Tho i am positive about the all core voltage increase, i will double check the single core by flashing back to 3602 when i get home. fairly sure theres been a slight increase tho.

5950x, Formula
It sounds like the new bios increases core voltages to address instabilities with AVX2 code. It should help really well with those crash to desktop reports with software like FS2020, Doom etc.
 

·
Registered
Joined
·
964 Posts
Gigabyte is still on AGESA 1.2.0.2. So not sure where you are getting your view on "weeks ahead on the latest AMD patches" from.

MSI is ahead as they are say that they are using 1.2.0.3.b compared to Asus on 1.2.0.3a. But after flashing, when you check the version in Aida64, it still shows as 1.2.0.3.a! So who knows if this patch b is actually implemented in the MSI firmware, but I don't think Asus is behind the other motherboard vendors.
Asus do use 1.2.0.3b though, I have it on my B550-F, has been very stable but with a 3900X
 

·
Registered
Joined
·
563 Posts
Maybe not so stable, just had my first idle reboot in several_months! Judging from that, 3501/3601 was more stable, also booth booted with a more wide selection of memory settings. This one was finicky.

Same Curve Optimizer, only thing I changed was the three aforementioned memory settings
Idle reboot is triggered by idle voltage though. It can be due to the CPU idle voltage being too low, or the PSU going to sleep mode due to the ATX power draw being too low. One thing you can test is the Power Supply Idle Control setting in bios, which can help prevent your PSU from going to sleep when it shouldn't. Otherwise, I've also used Corecycler to detect AVX2 failures and set using Curve Optimizer, which results in a higher voltage (compared to SSE tests) and also avoid idle reboots.


Asus do use 1.2.0.3b though, I have it on my B550-F, has been very stable but with a 3900X
Cool, any USB issues with your B550-F system?
 

·
Registered
Joined
·
563 Posts
Thanks safedisk! I just flashed it in. Running fine, about to start stability testing. I noticed at first reboot, it needed to retrain memory once. So there is something in 3702 that is more sensitive to timings (mine is tight for 4x16GB B-die)
 

·
Registered
Joined
·
42 Posts
I need some help, just got a new WD SN850 1TB @ M.2_1 socket
Interface speed is only gen4 x 2lanes @ about 50% performance



 

·
Registered
Joined
·
964 Posts
any USB issues with your B550-F system?
No, I have not had USB issues with any of my AM4 systems, I have four systems all running on at least AGESA 1.2.0.3A

I use USB keyboard, mouse, midi controllers, USB mixer, audio interfaces, usb switch etc everything works fine
 

·
Registered
Joined
·
36 Posts
I need some help, just got a new WD SN850 1TB @ M.2_1 socket
Interface speed is only gen4 x 2lanes @ about 50% performance



Move to other slot and compare. I don’t know your specific board, but one slot is via chipset and the other direct. Initially I had them around the wrong way round and strangely on my board the slots were in the opposite configuration than I expected.
 

·
Registered
Joined
·
3,105 Posts
I agree and know, but I was shaming them, and to be honest shaming them is the right thing to do here. Even if it results them not posting new biosses here to test. I mean who are we to test their random changes and then quessing what happened. We don't ow them anything they ow us, we bought the boards. :)
You realize how naive of logic that is yeah? You don't want to test it, don't d/l it, end of story.

You can just stick to d/l from their website, which is still on 3601.

An no, they don't owe you anything either. Grow up.
 
7681 - 7700 of 8526 Posts
Top