MPG X870E CARBON WIFI Beta BIOS

Sort by date Sort by votes
The latest bios has degraded my PBO offset significantly, somehow. I am using a Ryzen 9 9950X3D. I was able to do a -40 CCD0 and -26 CCD1 stable, and now with A33, I'm only able to get a -30 on CCD0 and -18 on CCD1. Has anyone else experienced this weird occurrence?
 
Upvote 0 Downvote
90 % will not do anything over -20 it will work but it will not be 100 stable run some mem test and video card test you will see it . Some people do a lot of tweaking to get it to work and it will works sometimes with more voltage. For example mind while go to - 23 -24. All core ill work but soon as I do a mem test and other tests it will crash and give errors.
 
Upvote 0 Downvote
dlomond said:
90 % will not do anything over -20 it will work but it will not be 100 stable run some mem test and video card test you will see it . Some people do a lot of tweaking to get it to work and it will works sometimes with more voltage. For example mind while go to - 23 -24. All core ill work but soon as I do a mem test and other tests it will crash and give errors.
Click to expand...
I don't have this issue. My cores on CCD1 have a huge variance for it to be stable though. I've been constantly testing it with multiple tests and it's now stable. The lowest core is -18 and the highest core is -34. Here are the numbers I have for both CCDs.

CCD0
-50, -37, -36, -37, -50, -44, -50, -49

CCD1
-25, -18, -24, -22, -34, -33, -31, -34

This is super stable on my system with all core and single core using an LLC of Mode 7 and a 900KHz switching frequency on the CPU and the SOC is set to LLC of Mode 4 at 700KHz. SOC is set to 1.245v.

I am running 4 dimms of 24GB of RAM with pretty tight timing at 6200mhz 30, 38, 38, 96 Nitro of 1, 2, 0. I set nitro 8x on memory training options to get this stable. It runs on 1.430v for the dimms and the others are 1.380v. UCLK=MCLK too so I feel like I'm lucky in that regard tbh.

As you can probably tell, I got pretty involved in tinkering the bios over all.
 
Upvote 0 Downvote
syntax20515d602ed said:
I don't have this issue. My cores on CCD1 have a huge variance for it to be stable though. I've been constantly testing it with multiple tests and it's now stable. The lowest core is -18 and the highest core is -34. Here are the numbers I have for both CCDs.

CCD0
-50, -37, -36, -37, -50, -44, -50, -49

CCD1
-25, -18, -24, -22, -34, -33, -31, -34

This is super stable on my system with all core and single core using an LLC of Mode 7 and a 900KHz switching frequency on the CPU and the SOC is set to LLC of Mode 4 at 700KHz. SOC is set to 1.245v.

I am running 4 dimms of 24GB of RAM with pretty tight timing at 6200mhz 30, 38, 38, 96 Nitro of 1, 2, 0. I set nitro 8x on memory training options to get this stable. It runs on 1.430v for the dimms and the others are 1.380v. UCLK=MCLK too so I feel like I'm lucky in that regard tbh.

As you can probably tell, I got pretty involved in tinkering the bios over all.
Click to expand...
Hello
Can you also present what you explained in a video and share it here?
 
Upvote 0 Downvote
uxexes said:
Hello
Can you also present what you explained in a video and share it here?
Click to expand...
I'm new to all of this myself. I also don't have much equipment to do a video. Why would you want a video? It's meh results. My point being i had better on previous bios versions.
 
Upvote 0 Downvote
syntax20515d602ed said:
I don't have this issue. My cores on CCD1 have a huge variance for it to be stable though. I've been constantly testing it with multiple tests and it's now stable. The lowest core is -18 and the highest core is -34. Here are the numbers I have for both CCDs.

CCD0
-50, -37, -36, -37, -50, -44, -50, -49

CCD1
-25, -18, -24, -22, -34, -33, -31, -34

This is super stable on my system with all core and single core using an LLC of Mode 7 and a 900KHz switching frequency on the CPU and the SOC is set to LLC of Mode 4 at 700KHz. SOC is set to 1.245v.

I am running 4 dimms of 24GB of RAM with pretty tight timing at 6200mhz 30, 38, 38, 96 Nitro of 1, 2, 0. I set nitro 8x on memory training options to get this stable. It runs on 1.430v for the dimms and the others are 1.380v. UCLK=MCLK too so I feel like I'm lucky in that regard tbh.

As you can probably tell, I got pretty involved in tinkering the bios over all.
Click to expand...
How did you test stability of all this?
Try running Aida64 stability test over night with cpu, cache, fpu checked to check your CO as it is really good at catching errors for CO.
As for ram, again, what did you test that with? Running 6200 on 4 dimms of 24gb seems unlikely to be stable as it stresses the IMC too much? Karhu, testmem5 with anta777 X3D, ycruncher VT3?

If all those pass or if you already have done them and passed then very impressive.
 
Upvote 1 Downvote
syntax20515d602ed said:
I've done karhu, ycruncher and occt tests for hours on end. No errors. I haven't done aida64 or testmem5.
Click to expand...
Test Aida and testmem5 then to ensure 100% stability. Many people reported that they are fully stable with their CO values but when they ran Aida64 with cpu, cache, fpu checked, errored pretty quick. If tm5 and Aida pass overnight, you are golden 😁
 
Upvote 1 Downvote
syntax20515d602ed said:
I've done karhu, ycruncher and occt tests for hours on end. No errors. I haven't done aida64 or testmem5.
Click to expand...
If it provides stability for what you do, that's the only thing that truly matters.
I also use AIDA, but based on my experience, it wouldn't be the first test to crash. With a bad CO curve, you're much more likely to crash quickly with Cinebench R23 or Y-Cruncher VT3 perhaps not immediately, but it would eventually. OCCT is also a solid option. AIDA is more of a good benchmarking tool for memory speed IMO.
For example, with a bad memory setting, AIDA might run for an hour or longer without crashing my system, whereas VT3 would fail within 5-10 minutes. I never test for more than an hour, there’s no real purpose since I wouldn’t subject my components to that level of stress in daily use, whether in gaming or software. Some insist on one-hour tests, others say 'overnight.' But by that logic, why not claim you need to stress your hardware for an entire week before calling it stable? It's absurd.
 
Last edited:
Upvote 0 Downvote
vic123159d02e5 said:
If it provides stability for what you do, that's the only thing that truly matters.
I also use AIDA, but based on my experience, it wouldn't be the first test to crash. With a bad CO curve, you're much more likely to crash quickly with Cinebench R23 or Y-Cruncher VT3 perhaps not immediately, but it would eventually. OCCT is also a solid option. AIDA is more of a good benchmarking tool for memory speed IMO.
For example, with a bad memory setting, AIDA might run for an hour or longer without crashing my system, whereas VT3 would fail within 5-10 minutes. I never test for more than an hour, there’s no real purpose since I wouldn’t subject my components to that level of stress in daily use, whether in gaming or software. Some insist on one-hour tests, others say 'overnight.' But by that logic, why not claim you need to stress your hardware for an entire week before calling it stable? It's absurd.
Click to expand...
I guess it's up to the person that's testing, then, really. There is so much conflicting information out there about testing methodologies. I just do an in-between. Hydra Pro helps with testing through its validation. It's much more rigorous than other tests I've found, and it makes use of Cinebench, Kharu and Y-cruncher VT3. Then, after that pass, I do a 4-hour test with OCCT with a variable workload. As for memory, AIDA likes to catch some errors, memtest using hydra does that well too IMO. OCCT is kinda meh in that area from what I've seen.

I have also found that setting the right VDDG for each core is a huge factor for stability and getting further gains in the Curve Offset, so people who are struggling should look into that. Just do it in a step increment at a time. Higher seems to be better for CCD1, and lower seems to be better for CCD0. Mine are 1.060 for CCD1 and 0.980 for CCD0. It really helps. Also, this new BIOS seems to help with stability greatly from the last ones' AGESA version.
 
Upvote 0 Downvote
Back
Top