Ryzen performance gain from BIOS update?

D

DuronBurgerMan

[H]ard|Gawd
Joined
Mar 13, 2017
Messages
1,340
This dude put out some numbers showing some big improvements in certain games for Ryzen with the new BIOS updates that came out in the last couple of days (I presume these are the ones AMD has been teasing -- there's a new one for my board available now, too, will have to check it out later). Never heard of this guy before though, so no idea if he's legit or full of sh*t:



Thoughts?
 
Here is what stood out to me... he did not specify what his system was setup as, example being the if the CPU was at stock (doubt it, he mentions voltage tweaking) or what the RAM was running at before or after. The ONLY thing he said was he was running the "same Corsair RAM". So then, my thinking is he was one of those who wasn't able to get 3200MHz out of the gate, and this BIOS happened to update things enough to allow 3200 to function. Additionally, if he is overclocked, he hasn't stated how he went about it. If he was using the onboard "Game Boost" feature, while I haven't used it myself yet, it could very possibly increase the memory multiplier as well. Alternatively, if he had been using the A-XMP option then that could've also decided to auto-select 3200 when he may have been only running 2666 before.

Sadly, he also didn't state WHICH VERSION of the BIOS he updated to, as there have been a couple released within the last week. Namely was the Beta 1.41 which was quickly pulled, for reasons relating to the AGESA apparently. However, a day or two later they also updated the official version to 1.3 (which I suspect is based on the 1.3x Betas). Due to that, I personally cannot verify his claims, given I do own that board.

Provided he was on a huge deadline, where he would no longer have the board available to him, I can't help but think he overlooked something. :\ To me, memory speed seems the most likely culprit, particularly since his CPU-Z had shown no change. On the same token, AMD had stated that sub-timings have been tweaked and allowing for more performance, so actual memory clock might not be changing, just the sub-timings had been significantly tightened (all things we can't see in the BIOS [without effort through various programs in Windows]) which has lead to certain games having performance increases as they were likely mem-bandwidth starved on the CPU side.

I do have BIOS v1.41 Beta, and 1.3 Official, so theoretically I could test this. Problem is I lack access to very many (useful) 3D benchmarks. I'm still running shipping BIOS due to it running perfect at DDR4-3200.

[/2cents]

EDIT: Grammar error fix. Originally was "doubt it mentions voltage tweaks".
 
Last edited:
Formula.350 said:
Here is what stood out to me... he did not specify what his system was setup as, example being the if the CPU was at stock (doubt it mentions voltage tweaking) or what the RAM was running at before or after. The ONLY thing he said was he was running the "same Corsair RAM". So then, my thinking is he was one of those who wasn't able to get 3200MHz out of the gate, and this BIOS happened to update things enough to allow 3200 to function. Additionally, if he is overclocked, he hasn't stated how he went about it. If he was using the onboard "Game Boost" feature, while I haven't used it myself yet, it could very possibly increase the memory multiplier as well. Alternatively, if he had been using the A-XMP option then that could've also decided to auto-select 3200 when he may have been only running 2666 before.

Sadly, he also didn't state WHICH VERSION of the BIOS he updated to, as there have been a couple released within the last week. Namely was the Beta 1.41 which was quickly pulled, for reasons relating to the AGESA apparently. However, a day or two later they also updated the official version to 1.3 (which I suspect is based on the 1.3x Betas). Due to that, I personally cannot verify his claims, given I do own that board.

Provided he was on a huge deadline, where he would no longer have the board available to him, I can't help but think he overlooked something. :\ To me, memory speed seems the most likely culprit, particularly since his CPU-Z had shown no change. On the same token, AMD had stated that sub-timings have been tweaked and allowing for more performance, so actual memory clock might not be changing, just the sub-timings had been significantly tightened (all things we can't see in the BIOS [without effort through various programs in Windows]) which has lead to certain games having performance increases as they were likely mem-bandwidth starved on the CPU side.

I do have BIOS v1.41 Beta, and 1.3 Official, so theoretically I could test this. Problem is I lack access to very many (useful) 3D benchmarks. I'm still running shipping BIOS due to it running perfect at DDR4-3200.

[/2cents]
Click to expand...

There's a new BIOS out for my board, as of 3/31, but I don't know if it includes the AGESA update or not. In fact, I don't know how you'd even tell... but if it did, then I could try testing, too.

If you do test this, let us know! Would definitely like to know if this is true or not.
 
You notice he saw a massive increase in only 2 games though but ironically....
RoTR and Hitman :)
The 2 games that has been singled out has having the issues with Nvidia and maybe specific GPUs and shows in other sites.
But I have to wonder if it is just that he messed something up and created a false positive result.

I am a bit confused on something though, did he change the voltage as part of the BIOS update and those benches?
I find that confusing because he mentions tweaking the voltages;
Specifcially he says "This was the last day I had with this board, and I just decided last minute to start tweaking the voltages I was putting in for the 3.8GHz overclock that I was using for the stock cooler".. This is around 2min50secs to 3min10secs.

Anyway maybe this is part of the much worse RoTR performance we are seeing by some, but then RoTR and Hitman seem highly unpredictable between the various reviews and why some refuse to use RoTR even before we had Ryzen and where it is performing even more unpredictable for Nvidia.
Needs a follow up for sure and by others.

Cheers
 
Last edited:
CSI_PC said:
You notice he saw a massive increase in only 2 games though but ironically....
RoTR and Hitman :)
The 2 games that has been singled out has having the issues with Nvidia and maybe specific GPUs and shows in other sites.

I am a bit confused on something though, did he change the voltage as part of the BIOS update and those benches?
I find that confusing because he mentions tweaking the voltages;
Specifcially he says "This was the last day I had with this board, and I just decided last minute to start tweaking the voltages I was putting in for the 3.8GHz overclock that I was using for the stock cooler".. This is around 2min50secs to 3min10secs.

Anyway maybe this is part of the much worse RoTR performance we are seeing by some, although none of them are perfect with Nvidia on Ryzen with this specific game and also sometimes Hitman for some reviews/analysis done.

Cheers
Click to expand...

Yeah, the whole thing is weird with those games. Obviously there are underlying issues that go way beyond what CPU and GPU combination you're running. Sh*t just seems broken.

I'd love to see a real comprehensive test of games between Ryzen and a couple Intel chips. I mean a boatload of tests. Tons of games, new and old... Average the results against some kind of baseline. Then compare the averages all together. This way outliers like these don't throw us off. We'd see where Ryzen stands in gaming much better this way. But damn, it'd be an awful lot of work.
 
DuronBurgerMan said:
Yeah, the whole thing is weird with those games. Obviously there are underlying issues that go way beyond what CPU and GPU combination you're running. Sh*t just seems broken.

I'd love to see a real comprehensive test of games between Ryzen and a couple Intel chips. I mean a boatload of tests. Tons of games, new and old... Average the results against some kind of baseline. Then compare the averages all together. This way outliers like these don't throw us off. We'd see where Ryzen stands in gaming much better this way. But damn, it'd be an awful lot of work.
Click to expand...

Well it does seem a very narrow and specific set of issues so far and comes back to RoTR and to a lesser extent Hitman from the various professional sites that done some recent investigations with Ryzen and Nvidia; Hardware Unboxed/PCGamesHardware.
PCGamesHardware did 20 games context this issue, and only RoTR and Hitman gave such unusual behaviour on Ryzen with Nvidia.
Although personally I do think there is a bit of a niggle with BF1 but not as extensive or as excessive.

Yeah, more tests done by professional sites more data we get, also need more to look at this recent update although it can only touch some aspects and not crucial CPU ones that comes back to AGESA.
I do think it could come back to the voltages though that he mentions tweaking, really not clear in his vid when he applied those, and maybe part of a false positive but still interesting how it only notably affected RoTR and Hitman in the small list of games he tested.

Cheers
 
Last edited:
Alright so I've looked at the Titanium's BIOS files I have on hand: Shipping build 110, Pulled beta 141, Latest build 130
Ran them through UEFITool NE (alpha 40) and did a Text search for AGESA which came up with some hits. Since you can have three instances of the program running I just loaded each into their own and did simple side-by-side.
Now, while I have no "version" to report on (my skill level lacking no doubt), here is the rundown...

1st result "AmiAgesaDxe":
110: "Full size: 450h (1104)"
141: "Full size: 458h (1112)" Larger
130: "Full size: 450h (1104)"
Compressed section>PE32 image section:
110: "Full size: 7A4h (1956)"
141: "Full size: 7E4h (2020)" Larger
130: "Full size: 7A4h (1956)"
​
2nd result was BIOS Setup Menu related, so non-comparable due to code changes to menu system

3rd result "CpuSetAgesaPcd":
110: "Full size: A2Eh (2606)"
141: "Full size: A56h (2646)" ---Larger
130: "Full size: A2Eh (2606)"
TE image section:
110: "Full size: 9CCh (2508)"
141: "Full size: 9F4h (2548)" ---Larger
130: "Full size: 9CCh (2508)"
​
4th result "CpuSetAgesaPcd":
110: "Full size: 13DAh (5082)"
141: "Full size: 152Ah (5418)" ---MUCH Larger
130: "Full size: 13DAh (5082)"
TE image section:
110: "Full size: 135Ch (4956)"
____"Number of sections: 2"
141: "Full size: 14ACh (5292)" ---STILL Larger
____"Number of sections: 3" ---That's a whole extra section! (I srsly dunno what that means exactly lol)
130: "Full size: 135Ch (4956)"
____"Number of sections: 2"​


So yea, seems pretty clear to me, there's an updated AGESA in the 1.41 Beta, but since MSI had yanked down all their mobo's Betas very shortly after 1.41's release due to an error in the AGESA (to what extent, I don't know), it seems these results may need to be taken with a grain of salt for now?

That isn't to say I'm not staying positive about everything. Now that I've determined that the current Latest lacks the AGESA, it seems to point out that he had used 1.41 which DOES have the new AGESA, and we know that there are sub-timing tweaks that have been done to it. Sadly it seems there may have been a problem with it (be it from AMD, or MSI's adding of it to the BIOSes), but it very well may be a sign of things to come... at least provided the variation in performance wasn't down to that nV issue you guys have pointed out.
 
Formula.350 said:
Alright so I've looked at the Titanium's BIOS files I have on hand: Shipping build 110, Pulled beta 141, Latest build 130
Ran them through UEFITool NE (alpha 40) and did a Text search for AGESA which came up with some hits. Since you can have three instances of the program running I just loaded each into their own and did simple side-by-side.
Now, while I have no "version" to report on (my skill level lacking no doubt), here is the rundown...

1st result "AmiAgesaDxe":
110: "Full size: 450h (1104)"
141: "Full size: 458h (1112)" Larger
130: "Full size: 450h (1104)"
Compressed section>PE32 image section:
110: "Full size: 7A4h (1956)"
141: "Full size: 7E4h (2020)" Larger
130: "Full size: 7A4h (1956)"
​
2nd result was BIOS Setup Menu related, so non-comparable due to code changes to menu system

3rd result "CpuSetAgesaPcd":
110: "Full size: A2Eh (2606)"
141: "Full size: A56h (2646)" ---Larger
130: "Full size: A2Eh (2606)"
TE image section:
110: "Full size: 9CCh (2508)"
141: "Full size: 9F4h (2548)" ---Larger
130: "Full size: 9CCh (2508)"
​
4th result "CpuSetAgesaPcd":
110: "Full size: 13DAh (5082)"
141: "Full size: 152Ah (5418)" ---MUCH Larger
130: "Full size: 13DAh (5082)"
TE image section:
110: "Full size: 135Ch (4956)"
____"Number of sections: 2"
141: "Full size: 14ACh (5292)" ---STILL Larger
____"Number of sections: 3" ---That's a whole extra section! (I srsly dunno what that means exactly lol)
130: "Full size: 135Ch (4956)"
____"Number of sections: 2"​


So yea, seems pretty clear to me, there's an updated AGESA in the 1.41 Beta, but since MSI had yanked down all their mobo's Betas very shortly after 1.41's release due to an error in the AGESA (to what extent, I don't know), it seems these results may need to be taken with a grain of salt for now?

That isn't to say I'm not staying positive about everything. Now that I've determined that the current Latest lacks the AGESA, it seems to point out that he had used 1.41 which DOES have the new AGESA, and we know that there are sub-timing tweaks that have been done to it. Sadly it seems there may have been a problem with it (be it from AMD, or MSI's adding of it to the BIOSes), but it very well may be a sign of things to come... at least provided the variation in performance wasn't down to that nV issue you guys have pointed out.
Click to expand...

Excellent. Thanks for digging into it. But this probably means the new BIOS for my board doesn't have the AGESA update... which stinks. I wanted to test that.
 
DuronBurgerMan said:
There's a new BIOS out for my board, as of 3/31, but I don't know if it includes the AGESA update or not. In fact, I don't know how you'd even tell... but if it did, then I could try testing, too.

If you do test this, let us know! Would definitely like to know if this is true or not.
Click to expand...
This highlights one of the things that pisses me off about the Mobo companies: the lack of info on BIOS updates. Few of them give any real info. What is so damn secretive about it?
 
ir0nw0lf said:
This highlights one of the things that pisses me off about the Mobo companies: the lack of info on BIOS updates. Few of them give any real info. What is so damn secretive about it?
Click to expand...

You and me both, brother. They could just say something like "includes AMD microcode updates" and that'd be real helpful in determining whether or not I should bother with it.
 
Update: Asus X370 Prime BIOS 515 does NOT have the new AGESA code in it:


So I'll wait for it before updating the BIOS.
 
ir0nw0lf said:
This highlights one of the things that pisses me off about the Mobo companies: the lack of info on BIOS updates. Few of them give any real info. What is so damn secretive about it?
Click to expand...
DuronBurgerMan said:
You and me both, brother. They could just say something like "includes AMD microcode updates" and that'd be real helpful in determining whether or not I should bother with it.
Click to expand...
Believe it or not, but they usually DO say they've updated the AGESA (or as some will say instead, Microcode). That is... when it has been updated (added to the BIOS). :)

But in general the reason for limited info in their change logs probably comes down to a couple things. One, language barrier, and as we know Google Translate isn't always the best when it comes to translating Asian into English, so a LOT is lost. Two, it's a cut-throat market! Everyone always wants to be able to add what the next is offering, so being as vague as possible ensures that can't (easily) happen.
 
Formula.350 said:
... at least provided the variation in performance wasn't down to that nV issue you guys have pointed out.
Click to expand...
Agree but indirectly it could possibly as RoTR/Hitman had more performance behaviour issues with Nvidia card, I do not think it is Nvidia issue in general but comes back to these two specific games and maybe the CPU-platform+Nvidia GPU exacerbated.
The guy this thread is about was also using Nvidia GPU.
TBH several of us saw that it highighted something more than it being just an Nvidia driver problem and was beyond them, and this video again confirms there is some crap going on with RoTR/Hitman on Nvidia GPUs (not necessarily their fault although possibly will try to improve-increase utilisation on their more powerful GPUs) but not other games.

Would be interesting to see if the NVMe x4 PCIe SSD drives have their stress test behaviour (higher queue depths) also improved, but only two sites are testing to the level required and not sure they will test any BIOS until AGESA officially available.
Also would be interesting to know what is so unusual about these 2 games compared to others tested.
Cheers and thanks for doing that check.
 
Simply for the halibut, I poked around more. Generally I suspect most things are going to stay similar between 110 and 130, with 141 standing out, which remains the case. So unless I state otherwise just assume that's what it is. 110/130 = Old, 141 = New
CpuIo2Dxe - Which I take as meanin CPU I/O
Old size: 2148
New size: 2212

CpuDxe
Old: 33632
New: 33760

AmdCcxXvDxe - I honestly am curious what Xv means in this instance... eXcavator? I mean, that's what Carrizo is built on, but then why CCX? Plus, Carrizo is referenced Cz anyways.
Old: 16096
New:16224

AmdCczZenZpDxe - And, Zp is Zeppelin, the 4C/8T modules in our chips.
Old: 42752
New: 51332

AmdUnbXvDxe - Xv again...
Old: 14400
New: 14464

AmdFabricZpDxe
Old: 25088
New: 25312

FchKernDxe - aka Fusion Controller Hub aka Promotory aka Chipset
Old: 57572
New: 57924

AmdMemAm4Dxe
Old: 708
New: 768

AodDxe -- No clue what this is but caught my eye, as perhaps AMD OverDrive?
O: 1636 N: 1732

What it all means... that's for someone else to decipher unfortunately, cuz it's all beyond me lol Hell, could mean aboslutely nothing for all I know :p
 
I tried ASUS's new AGESA beta bios (like ver 605 or something for B350+)
My DDR4 can now post at 2666, but rarely makes it very far into Windows.
Not quite far enough to complete a bench. Used to be 2400 12,14,14,14,35
@1.2V was my max post. Now 2666 16,18,18,18,39 @1.4V frustratingly
close to almost working, except it doesn't.

I can get far enough to launch CPU-Z and see the command timing at 2666
is only 1T, while always 2T at 2400, go figure... Also there is new setting for
processor on-die termination impedance. Tried several ohmages, no benefit.

So, I'm not feeling any speed difference from the new BIOS alone. Claims
on Reddit that Ryzen Master no longer needs HPET have not (yet) proven
true. I couldn't get RM to do squat without HPET.
 
You must log in or register to reply here.
Back
Top