DISCLAIMER:
None of the below is intended to be a guide. It's for information and insight only. Any updates required to the BIOS must be cleared with PCS in order to avoid any warranty issues.
Now that it seems we have a relatively stable BIOS for the AM5 lineup I thought I would do some testing of the various profile options / configurations and then compare to a manual setup. This likely isn't for everyone, but I'm hoping that it will give some insight into the different choices. I've not gone to any great lengths with the testing, I just did a number of simple tests to see where there was an impact and where there wasn't. Mileage will obviously vary as it completely comes down to silicone for how you will get on.
As much as this isn't a PCS configuration, we are always offering up such builds so I wanted to be able to share my findings to add some context for future discussions etc.
The main hardware considerations are:
X670e Proart Motherboard running BIOS 1416 (Latest stable at time of writing)
AMD 7800X3D CPU
Kingston Fury 6400CL32 RAM
My first episode in this little chain is on the Memory. With the SoC voltage issues and warranty concerns that have been going on over the past few weeks/months I think this is the most pertinent starting point. I'm not going to go into discussion with regards to all this in here, this is purely a show and tell. I will say that it looks like all the profile options are now safe to use (latest BIOS revisions only), so you can have a play around with them and see how you get on.
So, on installation of the latest BIOS I didn't touch a thing. I then ran some tests to see how the system faired. The stock settings are very loose and are effectively set for any old ram. My findings, as they were in the past, is that there is actually slight instability for me. It totally depends on the RAM in use, but if the RAM in the system wants for slightly higher voltage for operation, the default setting may trip up. This happened below:
Initially I forgot to switch off the AA x16 in Nvidia Settings, hence the second run). During the 3DMark test, which was Firestrike, I got some brief pauses during the runs. This I have found to typically be instability. I've had it in the past when I was tweaking the PBO. It doesn't lead to a crash or anything, but it always highlights a problem to me. This happened with the 6400 trial just below.
All Stock (RAM Frequency 4800Mhz)
Cinebench Multi - 18360
Cinebench Single - 1821
Froze * 32276 (CPU 31135, GPU 36087) https://www.3dmark.com/3dm/94428956? (Forgot to turn off AA x16)
Froze * 32590 (CPU 31022, GPU 36443) https://www.3dmark.com/3dm/94429139?
SuperPi 6.985
EXPO I Trial (6400 Unrealistic & Broken)
Cinebench Multi - 18278
Cinebench Single - 1821
Froze * 32793 (CPU 31262, GPU 36678) https://www.3dmark.com/3dm/94430242?
SuperPi 6.953
EXPO I Trial (6000 Setting - Realistic)
Cinebench Multi - 18301
Cinebench Single - 1821
33185 (CPU 32380, GPU 36875) https://www.3dmark.com/3dm/94431525?
SuperPi 6.981
EXPO II Trial (6000 Setting - Realistic)
Cinebench Multi - 18345
Cinebench Single - 1818
33091 (CPU 32475, GPU 36811) https://www.3dmark.com/3dm/94432555?
SuperPi 7.201
EXPO Extreme Trial (6000 Setting - Realistic)
Cinebench Multi - 18331
Cinebench Single - 1824
33163 (CPU 32081, GPU 36899) https://www.3dmark.com/3dm/94433204?
SuperPi 7.452
The interesting note so far, is that the SuperPi result seems to be getting worse with the more extreme RAM settings. I think this highlights how sensitive the RAM can be with AM5. The Cinebench scores aren't really changing much, gather this is a CPU intensive operation rather than RAM. For the Firestrike scores I think this is likely the most interesting for most owners of the 7800X3D. You do see a significant gain across the profiles. As expected 1000 points is gained with EXPO/DOCP, with margin of error after that. I think that goes to show, at the very least for me, that the best available option for most is going to be EXPO I. You get the benefits, with no real down side from the tightening of the timings. Overall the best result was from EXPO I in the above.
The next post will be on my manual tweaks. I'm still fine tuning this right now. I'm getting a bit more performance but nothing that's blowing me away so far.
None of the below is intended to be a guide. It's for information and insight only. Any updates required to the BIOS must be cleared with PCS in order to avoid any warranty issues.
Now that it seems we have a relatively stable BIOS for the AM5 lineup I thought I would do some testing of the various profile options / configurations and then compare to a manual setup. This likely isn't for everyone, but I'm hoping that it will give some insight into the different choices. I've not gone to any great lengths with the testing, I just did a number of simple tests to see where there was an impact and where there wasn't. Mileage will obviously vary as it completely comes down to silicone for how you will get on.
As much as this isn't a PCS configuration, we are always offering up such builds so I wanted to be able to share my findings to add some context for future discussions etc.
The main hardware considerations are:
X670e Proart Motherboard running BIOS 1416 (Latest stable at time of writing)
AMD 7800X3D CPU
Kingston Fury 6400CL32 RAM
My first episode in this little chain is on the Memory. With the SoC voltage issues and warranty concerns that have been going on over the past few weeks/months I think this is the most pertinent starting point. I'm not going to go into discussion with regards to all this in here, this is purely a show and tell. I will say that it looks like all the profile options are now safe to use (latest BIOS revisions only), so you can have a play around with them and see how you get on.
So, on installation of the latest BIOS I didn't touch a thing. I then ran some tests to see how the system faired. The stock settings are very loose and are effectively set for any old ram. My findings, as they were in the past, is that there is actually slight instability for me. It totally depends on the RAM in use, but if the RAM in the system wants for slightly higher voltage for operation, the default setting may trip up. This happened below:
Initially I forgot to switch off the AA x16 in Nvidia Settings, hence the second run). During the 3DMark test, which was Firestrike, I got some brief pauses during the runs. This I have found to typically be instability. I've had it in the past when I was tweaking the PBO. It doesn't lead to a crash or anything, but it always highlights a problem to me. This happened with the 6400 trial just below.
All Stock (RAM Frequency 4800Mhz)
Cinebench Multi - 18360
Cinebench Single - 1821
Froze * 32276 (CPU 31135, GPU 36087) https://www.3dmark.com/3dm/94428956? (Forgot to turn off AA x16)
Froze * 32590 (CPU 31022, GPU 36443) https://www.3dmark.com/3dm/94429139?
SuperPi 6.985
EXPO I Trial (6400 Unrealistic & Broken)
Cinebench Multi - 18278
Cinebench Single - 1821
Froze * 32793 (CPU 31262, GPU 36678) https://www.3dmark.com/3dm/94430242?
SuperPi 6.953
EXPO I Trial (6000 Setting - Realistic)
Cinebench Multi - 18301
Cinebench Single - 1821
33185 (CPU 32380, GPU 36875) https://www.3dmark.com/3dm/94431525?
SuperPi 6.981
EXPO II Trial (6000 Setting - Realistic)
Cinebench Multi - 18345
Cinebench Single - 1818
33091 (CPU 32475, GPU 36811) https://www.3dmark.com/3dm/94432555?
SuperPi 7.201
EXPO Extreme Trial (6000 Setting - Realistic)
Cinebench Multi - 18331
Cinebench Single - 1824
33163 (CPU 32081, GPU 36899) https://www.3dmark.com/3dm/94433204?
SuperPi 7.452
The interesting note so far, is that the SuperPi result seems to be getting worse with the more extreme RAM settings. I think this highlights how sensitive the RAM can be with AM5. The Cinebench scores aren't really changing much, gather this is a CPU intensive operation rather than RAM. For the Firestrike scores I think this is likely the most interesting for most owners of the 7800X3D. You do see a significant gain across the profiles. As expected 1000 points is gained with EXPO/DOCP, with margin of error after that. I think that goes to show, at the very least for me, that the best available option for most is going to be EXPO I. You get the benefits, with no real down side from the tightening of the timings. Overall the best result was from EXPO I in the above.
The next post will be on my manual tweaks. I'm still fine tuning this right now. I'm getting a bit more performance but nothing that's blowing me away so far.
