[RaZieL980]

Hey folks,
Currently have this MoBo on BIOS 4602 with a 2600X. Grabbing a 5800X as an upgrade and will obviously need to update the BIOS.

Question is; Can I jump straight to 5837 (latest BIOS for this board, AGESA 1.2.0.0 ) from 4602 or shall I do a few incremental jumps? Anyone done the same upgrade as I'm planning and had any issues?

Thanks!

 

[MishelLngelo]

Hey folks,
Currently have this MoBo on BIOS 4602 with a 2600X. Grabbing a 5800X as an upgrade and will obviously need to update the BIOS.

Question is; Can I jump straight to 5837 (latest BIOS for this board, AGESA 1.2.0.0 ) from 4602 or shall I do a few incremental jumps? Anyone done the same upgrade as I'm planning and had any issues?

Thanks!

Yes.

 

[xeizo]

Strange find on the new bios 5837, sleep works as it should again! Keeps all OC, IF/mem 1:1:1! Boots really fast from sleep
I discovered by coincidence, I was running Windows Insider on my 3700X build and the latest Insider 21318 kills L3 totally. More than three times slower than it should be(hope this is a bug and not a new "feature"). So I decided to go back to normal Windows which is silly faster, a clean install meant I kept sleep as it is default. And it works without a hitch, has been a problem on Zen for me since Zen+.

Bios 5837 is clearly the best so far imho, at least for Zen 2, haven't tested Zen 3 as I have the CH8 for that.

Only con for bios 5837 is it undervolts vcore quite heavily which hurts performance, a little + offset on vcore and the 3700X performs better than a 3800X using PBO

 

[MishelLngelo]

Strange find on the new bios 5837, sleep works as it should again! Keeps all OC, IF/mem 1:1:1! Boots really fast from sleep
I discovered by coincidence, I was running Windows Insider on my 3700X build and the latest Insider 21318 kills L3 totally. More than three times slower than it should be(hope this is a bug and not a new "feature"). So I decided to go back to normal Windows which is silly faster, a clean install meant I kept sleep as it is default. And it works without a hitch, has been a problem on Zen for me since Zen+.

Bios 5837 is clearly the best so far imho, at least for Zen 2, haven't tested Zen 3 as I have the CH8 for that.

Only con for bios 5837 is it undervolts vcore quite heavily which hurts performance, a little + offset on vcore and the 3700X performs better than a 3800X using PBO

I still can't make offset more than -0.03v without occasional bugout but that's probably due to my very early 3700x. In any case. 1.326v is most I have seen which is much better than 1.4v+ before.
One thing that still does almost nothing is fast BOOT, upon a restart it just doesn't show post screen but spends same time as blank(black) screen.

 

[xeizo]

I still can't make offset more than -0.03v without occasional bugout but that's probably due to my very early 3700x. In any case. 1.326v is most I have seen which is much better than 1.4v+ before.
One thing that still does almost nothing is fast BOOT, upon a restart it just doesn't show post screen but spends same time as blank(black) screen.

It is made to have 1.4V+, ideally it goes up to 1.5V in idle, under load it automatically goes down to 1.25-1.33V.

If you do minus offset, like you have done, you kill single core performance. The CPU needs high vcore in idle to be able to boost, that's why I use +(plus) offset as the Prime board undervolts by almost 0.05V. If you have good enough cooling multi wont be affected, I have a Dark Rock Pro 4 which cools perfect. Max temp under CB is 75C. Best core boosts to 4475MHz which is nice for a 3700X.

And yes, sleep works flawless now

 

[MishelLngelo]

It is made to have 1.4V+, ideally it goes up to 1.5V in idle, under load it automatically goes down to 1.25-1.33V.

If you do minus offset, like you have done, you kill single core performance. The CPU needs high vcore in idle to be able to boost, that's why I use +(plus) offset as the Prime board undervolts by almost 0.05V. If you have good enough cooling multi wont be affected, I have a Dark Rock Pro 4 which cools perfect. Max temp under CB is 75C. Best core boosts to 4475MHz which is nice for a 3700X.

And yes, sleep works flawless now

That's more like in theory because silicone lottery which I obviously didn't win (CTR says it's bronze sample) and it's one of first batch. It barely touches 4.4GHz, At least 1.4v is needed for all core OC to 4.3GHz. Not a great loss in SC performance but boost gets smaller and smaller with every degree in temps over 70c and controlling voltage is only way.
This is what I'm getting now with best settings for PBO etc,

That's for 70c max, as soon as it gets over that I loose (by my calculations) 10-20 points for every degree over 70c.

 

[xeizo]

That's more like in theory because silicone lottery which I obviously didn't win (CTR says it's bronze sample) and it's one of first batch. It barely touches 4.4GHz, At least 1.4v is needed for all core OC to 4.3GHz. Not a great loss in SC performance but boost gets smaller and smaller with every degree in temps over 70c and controlling voltage is only way.
This is what I'm getting now with best settings for PBO etc,
View attachment 2479759
That's for 70c max, as soon as it gets over that I loose (by my calculations) 10-20 points for every degree over 70c.

It's a matter of preference, I prefer single core performance as multi already are great as is. This is with current settings, now CBR23 is very demanding so in lighter applications the single boost are even better. As you said, temps comes in when the load is as steady.

Not that the difference is big, this is 3.33% faster single core, but anyways

edit. I have a old sour 3700X as well, I was a very early adopter, later samples are much better from what I've heard

 

[MishelLngelo]

It's a matter of preference, I prefer single core performance as multi already are great as is. This is with current settings, now CBR23 is very demanding so in lighter applications the single boost are even better. As you said, temps comes in when the load is as steady.

Not that the difference is big, this is 3.33% faster single core, but anyways

edit. I have a old sour 3700X as well, I was a very early adopter, later samples are much better from what I've heard

View attachment 2479781

Yes, what I have seen later ones (3rdQ) are at least 100-200MHz better across the board @ about 0.1v less but that translates to very few points in benchmarks and even less in real life. That's what 3800x was supposed to be.

 

[xeizo]

I see you are on build 21301, beware that 21318 kills L3 in Ryzen. 3.5 times higher L3 latency and 6 times lower bandwidth. I had to go back to normal 19042 as I couldn't stand such a large regression.

 

[MishelLngelo]

I see you are on build 21301, beware that 21318 kills L3 in Ryzen. 3.5 times higher L3 latency and 6 times lower bandwidth. I had to go back to normal 19042 as I couldn't stand such a large regression.

I have both, regular 21H1 and insider 21318 dev builds but CPU and RAM scores are practically identical.

 

[MishelLngelo]

21H1

And this one is from 21318 dev build

 

[xeizo]

21H1
View attachment 2479814
And this one is from 21318 dev build
View attachment 2479815

LoL! They are not identical, L3 cache latency is 3.5 times higher and bandwidth is 6-12 times lower! Just as I said. You have to read all the boxes, not just the first row. And yes, performance nosedives. 21318 is a abomination for Ryzen.

In essence, 21318 makes a Celeron out of Ryzen(Celeron = crippled cache)

 

[MishelLngelo]

21H1
View attachment 2479814
And this one is from 21318 dev build
View attachment 2479815

21H1
View attachment 2479814
And this one is from 21318 dev build
View attachment 2479815

Yes I can see that L3 cache is much worse just not sure how much and where it matters. Both tests are on same machine but just noticed I tested with different Aida versions, here's one with same version on both windows.

 

[xeizo]

Yes I can see that L3 cache is much worse just not sure how much and where it matters. Both tests are on same machine but just noticed I tested with different Aida versions, here's one with same version on both windows.
View attachment 2479877

N-body and Image benchmarks in GB5 takes a hefty nosedive as does CPU in TimeSpy, it sure affects general performance. Anyway, it made me quit insider on that machine. I still use it in my 2700X rig so I can see if/when the problem is fixed.

 

[MishelLngelo]

N-body and Image benchmarks in GB5 takes a hefty nosedive as does CPU in TimeSpy, it sure affects general performance. Anyway, it made me quit insider on that machine. I still use it in my 2700X rig so I can see if/when the problem is fixed.

I have insifer on separate and small SSD, just to keep up with it's development with very few stuff in it. Real work and play is on NVME. Only performance problem I have with it is dismal performance with Dx11, it's been like that with last few builds. All SW and drivers are same as with "regular" builds.

 

[radier]

https://twitter.com/i/web/status/1364550020540768260

 

[MishelLngelo]

That means yet another BIOS version. When we'll see it ? Btw, there's something wrong with Asus Prime site. Drivers and utilities doesn't open.

 

[radier]

https://twitter.com/i/web/status/1364971430337740804