AMD boost (beta) fix - tested by Toms

Nobu said:
otoh, the intel fans (shills) would also say it's nothing. Maybe it actually IS nothing, regardless of whether it happens on an AMD or Intel processor? Feel free to quote me when it happens and I don't throw shit at the fan.

Edit: honestly, I don't think it is exactly nothing. Clearly, there was an issue, and AMD is working to resolve it. 100MHz is not huge, especially when one thread is only hitting it 1% of the time, but depending on how it is handled (in the os and microcode) it can make a big difference in certain workloads. Not in time to completion, necessarily, but in other less predictable ways. The problem is that there isn't really any way to show how this actually affects people in real workloads, and benchmarks + monitoring programs only show part of the problem... sometimes.
Click to expand...
What is the exact number of Hz when lying in marketing becomes a problem? Or is it percentage based? Is 1000 Hz ok? 278 Hz? 1%? 5.43%? The great performance is still there, either way, right? Why not sell "up to 6 GHz" CPUs? It would look great on the box. They could charge extra for "up to 7 GHz" SKUs. Sure, it may in reality hit lower than the up to 6 one, but money is money, amirite?
 
Meeho said:
What is the exact number of Hz when lying in marketing becomes a problem? Or is it percentage based? Is 1000 Hz ok? 278 Hz? 1%? 5.43%? The great performance is still there, either way, right? Why not sell "up to 6 GHz" CPUs? It would look great on the box. They could charge extra for "up to 7 GHz" SKUs. Sure, it may in reality hit lower than the up to 6 one, but money is money, amirite?
Click to expand...
I never said anything about marketing. It's all bs in my opinion. You're preaching to the choir. Etc, ad nauseum.
 
Meeho calm down, there's indeed in general an expectation of 5% possible deviation that's considered acceptable for most things in engineering, not everything but most things, if you are such a tech head you should have seen this through the years as differences at that level being considered "within the margin of error", there are many different factors that influence every single result.

Also it has been explained multiple times that mhz alone don't tell the whole story. In any case the fix is being tested and the performance seems to be staying within that error margin.
 
Revdarian said:
Meeho calm down, there's indeed in general an expectation of 5% possible deviation that's considered acceptable for most things in engineering, not everything but most things, if you are such a tech head you should have seen this through the years as differences at that level being considered "within the margin of error", there are many different factors that influence every single result.

Also it has been explained multiple times that mhz alone don't tell the whole story. In any case the fix is being tested and the performance seems to be staying within that error margin.
Click to expand...

Probably one of the bigger factors is in the base clock of the mainboard. I have yet to own one that the base clock is exactly 100 Mhz and usually, it is closer to 99Mhz, most of the time that I have looked.
 
I have ABBA and my single thread in cb r15 went up 2 points from 202 to 204 and my cores still go to 4550. But I'm using PBO enabled. I might disable PBO and see what pb2 does instead

Hope I can hit 4600 just to bring closure to it. Other than that I could give a rat's ass less about this controversial issue.
 
Revdarian said:
Meeho calm down, there's indeed in general an expectation of 5% possible deviation that's considered acceptable for most things in engineering, not everything but most things, if you are such a tech head you should have seen this through the years as differences at that level being considered "within the margin of error", there are many different factors that influence every single result.
Click to expand...
This is...special.
 
Oh, I'm not complaining. I'm enjoying my Ryzen 3 processor, and I'm getting the real-world performance I expected (even better actually) and I don't mind if the MHz number is not exactly the same on the box.

As much as it's a non-issue in terms of performance, AMD still should not advertise false numbers. I think the problem has been overblown, so we agree there, but I do think it is on AMD to fix it (and all signs are that they are).

Case closed.
 
You may not see the + part if they keep the max fixed to the advertised max, that way at the top you would only get the speed marketed minus whatever could be lost in the way from inefficiencies.

There are simple answers beyond conspiracies, and again the issues started with the post review bios fix that increased the actual effective performance even though the GHz went slightly down, a simple explanation would be that they tightened something beyond the thermals that got the processor cost to a sweet spot. The thermals were a bug apparently from what has been shown and they are being addressed, and yet the performance for most users has seen little to no variance.

It's funny that I'm treated as a shill when I've been laughing and saying that proving actual wrongdoing if they wouldn't have fixed would have been hard due to the actual effective performance increment, if your processor is actually wrong then you should enforce the warranty.
 
Revdarian said:
You may not see the + part if they keep the max fixed to the advertised max, that way at the top you would only get the speed marketed minus whatever could be lost in the way from inefficiencies.

There are simple answers beyond conspiracies, and again the issues started with the post review bios fix that increased the actual effective performance even though the GHz went slightly down, a simple explanation would be that they tightened something beyond the thermals that got the processor cost to a sweet spot. The thermals were a bug apparently from what has been shown and they are being addressed, and yet the performance for most users has seen little to no variance.

It's funny that I'm treated as a shill when I've been laughing and saying that proving actual wrongdoing if they wouldn't have fixed would have been hard due to the actual effective performance increment, if your processor is actually wrong then you should enforce the warranty.
Click to expand...

Yeah they're stuck in a "damned if you do, damned if you don't" type of scenario... I'm sure Intel PR is trying its best too to sway things in one direction and don't get me wrong... you wouldn't want anything else from your PR department given the current situation.
This clash is a good thing for us in some way, it means there's competition and choices at the end of the day !!
 
SilverSliver said:
My 3900x went from 4.3 max single core to 4.55 max with ABB and now with ABBA I'm seeing 4.65 max single core. Benchmark scores are up between 1 and 3% versus ABB and over 5% from the first one.
Click to expand...

Benchmark on single thread or multi? I’d imagine single thread boost would help minimum FPS a bunch, which is where AMD is lacking (at high Hz.)

5% isn’t anything to scoff at either considering a lot of us pay decent money on coolers, ect, to get that.

Good to see AMD making some progress.
 
Dayaks said:
Benchmark on single thread or multi? I’d imagine single thread boost would help minimum FPS a bunch, which is where AMD is lacking (at high Hz.)

5% isn’t anything to scoff at either considering a lot of us pay decent money on coolers, ect, to get that.

Good to see AMD making some progress.
Click to expand...

Multicore is seeing the smallest gains, single core is about 5% from various benchmarks of that sort.

I'm routinely seeing 4500 to 4600 boost in games when before I didn't see over 4400
 
Last edited:
SilverSliver said:
Multicore is seeing the smallest gains, single core is about 5% from various benchmarks of that sort.

I'm routinely seeing 4500 to 4600 boost in games when before I didn't see over 4400
Click to expand...

Please share before and after 1% low fps and average fps with us if possible. Toms results were lackluster but they were boosting a little higher than you did with the original bios.
1080P is fine I guess but in 2019 I would look at 1440P given the price point for gaming benchmarks.
 
MMitch said:
Please share before and after 1% low fps and average fps with us if possible. Toms results were lackluster but they were boosting a little higher than you did with the original bios.
1080P is fine I guess but in 2019 I would look at 1440P given the price point for gaming benchmarks.
Click to expand...

GN posted a video showing some quick game comparisons.. on average it's about 1-2% with some odd games showing bigger benefits. but ultimately it's still a silicon lottery and windows scheduling on whether or not it'll hold 4.6Ghz. but the fix is definitely working and they didn't fix it by throwing more voltage at the chip to make it happen which is nice.

going to try the bios tomorrow with my 3600 and see if i can get the auto overclock to actually do anything.
 
chameleoneel said:
https://www.techpowerup.com/review/amd-agesa-abba-boost-clock-fix-performance-ryzen-3900x/
Click to expand...
"AMD has succeeded in delivering on the advertised maximum boost frequencies with elevated clock speeds across all cores, which results in tiny performance gains at negligible increases in power draw."

In conclusion, a net small gain and respecting the stated spec at the same time. Just what poeple were hoping and asking for and were attacked or ridiculed in threads on this subject. I expect all of those to keep true and not bother installing the latest AGESA to not turn out hypocrites. I won't hold my breath.
 
I was one that thought it was a non-issue, and even with the fix we are talking about negligible gains, but I'll definitely install it.

It might not be much, but I'll take a free 1% gain, sure.
 
Meeho said:
"AMD has succeeded in delivering on the advertised maximum boost frequencies with elevated clock speeds across all cores, which results in tiny performance gains at negligible increases in power draw."

In conclusion, a net small gain and respecting the stated spec at the same time. Just what poeple were hoping and asking for and were attacked or ridiculed in threads on this subject. I expect all of those to keep true and not bother installing the latest AGESA to not turn out hypocrites. I won't hold my breath.
Click to expand...

*Slow Golf Clap* It was a non issue before, it was a non issue now. Your "expectations" of what others do with their own machines are none of you business.
 


Hardware Unboxed full updated video testing multiple boards includes a small explanation in case your board doesn't get the 100mhz fsb you can try and set it to 101 by hand given the chance in the bios.

Watch the full video, tests single and multi core hz and actual performance.

Edit : they did see that you could boost a bit higher than advertised but in their test it didn't seem to last long enough to matter.
 
Really nice video from Hardware Unboxed.

Still want to see what the BIOS does for my chip (not released yet) but it looks like this chapter is done.
 
Revdarian said:


Hardware Unboxed full updated video testing multiple boards includes a small explanation in case your board doesn't get the 100mhz fsb you can try and set it to 101 by hand given the chance in the bios.

Watch the full video, tests single and multi core hz and actual performance.

Edit : they did see that you could boost a bit higher than advertised but in their test it didn't seem to last long enough to matter.
Click to expand...


These guys are the best. The put out data a week before this with the right conclusion but instead fan boys jumped on der8auer's click bait a week later. Really tell's you where the community is at
 
Bios is out for my gigabyte board but unfortunately I can't boot with it, it's causing my m.2 drive to stop being detected. I'm not the only one having that issue so I'd expect gigabyte will fix that up quickly. They've been on the ball with pumping every new update straight out to us.

Glad the results are looking good otherwise.

Now, where's my PBO? HEHE XD :p

attachment.gif


blackmomba said:
These guys are the best. The put out data a week before this with the right conclusion but instead fan boys jumped on der8auer's click bait a week later. Really tell's you where the community is at
Click to expand...

Don't take this the wrong way but I still think the vast majority of ridicule and drama is coming from people who either don't own ryzen 3xxx chips or those who only had very small deficits under advertised boost. I tried my best to remain completely calm on the subject just stating over and over again that I just have an issue and am looking for a fix, thing's I've tried, progress (or lack thereof) i've had, etc. I did and do however want AMD to address the claims in PBO video that drove me and many others to spend upwards of $300 (and way more) on new x570 boards. It would have been easy to be completely calm if it wasn't for the ridicule. I was posting about this here and elsewhere since July, but suddenly after the derbauer video, the assault comes in and we're all being accused of whining. Nobody was whining or mad prior to the video making this mainstream and non-ryzen owners jumping into the threads to fight us.

I have a right to ask for what's on the box to be what I get, and to point out misleading advertisements. I'm very happy AMD responded and worked out a fix. I never read anything about lawsuits or pitchforks until people accused us of such things. What's not right is all of the people jumping in with zero context to accuse us of being babies. What this is telling me is that "where the community is at" is really that we accept corporations lying to us and we accept marketing tricks so much that we actually defend them and call people who want to get what's on the box stupid. That's what I've gathered after 2 months heavily involved in the launch of these new CPUs. Realistically it's probably just people jumping around forums, not reading, and looking for people to insult.
 
Well I stand corrected, they did manage to squeeze a little more out without impacting other metrics which is nice although gain are somewhat in the noise level.
That's one of those scenario where I'm ok with been wrong :)
 
If anyone's interested, I just updated the BIOS on my Asus WS X570-Ace, and ran a Cinebench single-core CPU test with HWInfo64 running. Before that, pretty much the max speed I'd see on any of the cores was 42.5x, although rarely I'd see one bin (25MHz) faster. Now, the two best cores in CCX0 at least hit 44.0, with an average speed of 42.3x. Eventually, 3 cores reached 44.0x, one got 43.75, and the last two (very) briefly hit 43.5. During the test, however, those last two only ever hit 42.5 and 42.0.

So now the CPU (a 3600X) at least is capable of hitting the rated single-core speed on 3 cores, and came very close on the remaining 3, although with varying levels of success. Not great, as my 1600X could do 40 on all cores all day long, but it's definitely better, if not probably noticeably so.

Finally, running the all-core Cinebench, two cores would spend almost all their time at 42.5-43.5x, a third would spend a noticeable amount of time at that speed, and the last three pretty much spent all their time at 37.75, although one at a time would sometimes hit 42.5-43. IIIRC with ABB nothing would ever go past 42.5 previously. [edit: whoops, that was single-core too, hit the wrong button. All-core is the same as before, 40.75-41.0x on all cores; no improvement with the new BIOS.]
 
Last edited:
Aside: I'd love to see profiling software that, given a benchmark like CInebench CPU, could tell me what % of time each core hit any given speed, to see overall how fast it goes.
 
Meanwhile seeing all these BIOS fixes and patches makes me almost glad my 3900x still hasn’t shipped.... I mean I still want it and all but I don’t have the time to fiddle with this stuff anymore. I will give Intel all the crap they deserve for sitting idle for the last few years but at least there is a high degree of consistency with their products once you have them in hand.
 
Revdarian said:


Hardware Unboxed full updated video testing multiple boards includes a small explanation in case your board doesn't get the 100mhz fsb you can try and set it to 101 by hand given the chance in the bios.

Watch the full video, tests single and multi core hz and actual performance.

Edit : they did see that you could boost a bit higher than advertised but in their test it didn't seem to last long enough to matter.
Click to expand...

Thanks, non-issue solved.
Poor dead horse sure got kicked a lot.
 
Lakados said:
Meanwhile seeing all these BIOS fixes and patches makes me almost glad my 3900x still hasn’t shipped.... I mean I still want it and all but I don’t have the time to fiddle with this stuff anymore. I will give Intel all the crap they deserve for sitting idle for the last few years but at least there is a high degree of consistency with their products once you have them in hand.
Click to expand...

You must be ignoring the million mitigations that end users have to have to patch?
 
Lakados said:
Meanwhile seeing all these BIOS fixes and patches makes me almost glad my 3900x still hasn’t shipped.... I mean I still want it and all but I don’t have the time to fiddle with this stuff anymore. I will give Intel all the crap they deserve for sitting idle for the last few years but at least there is a high degree of consistency with their products once you have them in hand.
Click to expand...

Why did you even post this then?

3900x is a monster CPU! But in your eyes is a 50mhz lower boost clock make it a sloppy junk cpu?

Why not cancel your order and pass on the chip to someone who actually wants the CPU with no second guesses?
 
tangoseal said:
Why did you even post this then?

3900x is a monster CPU! But in your eyes is a 50mhz lower boost clock make it a sloppy junk cpu?

Why not cancel your order and pass on the chip to someone who actually wants the CPU with no second guesses?
Click to expand...
Not at all, but when it first came out damned thing was so buggy it couldn’t launch Destiny 2. One of the few games I currently play, most forums from Gigabyte, MSI, and Asus are pages and pages of no POST followed by “Bios update coming” that fixes it. It is an awesome CPU but seems to struggle with Ram compatibility and I am happy that by the time it arrives on my door tens of thousands of other people got to be the beta testers.
 
Lakados said:
Not at all, but when it first came out damned thing was so buggy it couldn’t launch Destiny 2. One of the few games I currently play, most forums from Gigabyte, MSI, and Asus are pages and pages of no POST followed by “Bios update coming” that fixes it. It is an awesome CPU but seems to struggle with Ram compatibility and I am happy that by the time it arrives on my door tens of thousands of other people got to be the beta testers.
Click to expand...

Well it had a bumpy start I agree, but dest 2 is an awful game anyways so I cant relate. However, you're truly going to enjoy its power. On the contrary, at least AMD only had a few misses on clock speed unlike intel with a new violation of critical security protocols weekly. Any more performance killing patches to fix security flaws and you'll have more fun using a 486 dx100 to play that Destiny 2.
 
You must log in or register to reply here.
Back
Top