The 2nd graph (and text) shows the new version to idle at 3,0 GHz instead of ~2,2 GHz and taking longer to reach max frequencies. Am I missing something or is that just a step backwards? Higher idle frequency means more power draw during idle and slower reaction to load doesn't sound better either.
This sounds somewhat like they try to fix the reported values when in reality the CPU works fine, just the software isn't up to date. And instead of changing the software, now they change the CPU behavior making it somewhat worse but look better? I'd like to at least have some options to adjust this behavior.
"If a processor core is truly running no workload, the processor core will be placed into the “Core C6” (CC6) power-gated sleep state. Presently, AMD Ryzen Master is the only tool that can show this CC6 state (Figure 3 attached). This state may be opportunistically engaged hundreds of times per second. Voltages in this condition might be as low as 0.200V, and clockspeeds down to 0MHz/Sleep are also possible.
They went back to tackle some issue, (propably one that stops linux from using the CPU) I Think they know how to make this better, it just need some more time.
[a] I noticed that you only posted results for 'Windows Balanced'. With the new chipset drivers, is the response time also affected for 'AMD Ryzen High Performance' (and other power plans)?
What if one prefers the 'former' behavior (of sub 1-ms response times)? Is there a way to adjust the settings (with the new chipset drivers) and/or will AMD keep an archive of different chipset driver releases (to suit the requirements of different users)?
I would like answers to this as well. I very much LIKE the way this CPU behaved before this change. I WANT ultra fast ramp up so I get ultra fast responsiveness. I don't give a crap about 40C IDLE or really any of the complaints.
Please AMD, if this has to be like that, please do publish an alternative for the old path.
I dont think race to sleep is affected as it will still change clocks/voltage in under 1ms in response to load alterations once the initial ramp is completed.
the initial step, so probably all tests will see ~50ms difference or last 2 digit in 5000+ result score. same thing happens eveywhere on any software update, so its just whining. nobody complained about meltdown patches that were in 1X%, this will end up sub 1%.
No. You're being willfully ignorant. The tiny change in initial ramp-up time won't affect performance of any actual load. They said as much in the article.
Now, if you WANT it to constantly ramp even when there's no load, due to monitoring programs, just switch from balanced to high performance.
Great, now we just need the motherboard manufacturers to fix their BIOS. I have a MSI B450M Gaming Plus I bought and paired with a Ryzen 3600 and latest R3K compatible BIOS on the 27th. In the last 3-4 days I've managed to boot once in ~50 tries, the rest of the time I get a red CPU debug LED and no boot. I have it bad though, most others with this issue get to boot somewhere between 1/10 and half the tries. The MSI_Gaming subreddit is full of these stories about "compatible" BIOSes not managing to POST. Might put some pressure on MSI if websites like Anandtech decided to cover the issue...
Try switching the memory modules to different slots. I've had a soft reboot problem on MSI B450 Mortar. I've put the sticks to the slots closer to the CPU and it won't reboot properly (needed hard power cycle). After moving the memory to the slots further away from the CPU, there are no more problems.
this update definitely helped (but didn't entirely fix) the hyper-sensitive boosting on my X470 Aorus Ultra Gaming / 3700X combo; I still don't think the "it's supposed to idle at 30W! that's just how it works!" excuse is valid, but I'm sure they'll fix it in a month or two.
zen 1 launch was just as janky but on the RAM side, which as been completely solid on the zen 2 launch for me.
It doesn't idle at 30W--even without the update. As stated in the article, it ramps up with some over-aggressive monitoring software. By taking measurements at a very fast rate, the CPU interprets that as a heavy workload and ramps up accordingly.
Don't use power saver, it artificially throttles the CPU back. I don't know if it's relevant today, but back in the Windows 2000 days, Power Saver didn't even save power, it instead ran a simple process to waste time and slow the effective responsiveness of the machine, but I could be misremembering it horribly.
yeh, RAM wise they have seemingly sorted out all the ills (that I can tell)
I think for Ryzen (Zen, Zen+, Zen2 as well as the APU based on such) there are just far far more "turn this on, turn that off" hundreds of times per second over "lots of space", I would imagine a little "burp" at this level is not a "burp" it becomes a grindy/grinding mess in real quick order (not good)
I know Zen1 and Zen2 share more than Zen2 and Zen+ do (from what I have read) so that likely means Zen 3 or was Zen 4 (final for AM4 according to AMD) will combine all the bits and piece to make the "true" enchilada Ryzen Ultra (something like that)
scheduler from Zen 1 and 2 are similar Zen+ is "it's own" but the turbo and CCX design more like Zen+ than Zen 1, a few new things added to the mix, so, they basically had to "reset" 3 times their base design.
they seem to be "figuring out" quicker but sadly is not like "the good ole days" where say AMD 700 through 900 were "near identical" from the motherboard side of things (YMMV if it acted simple as should have) certainly not a shelf full of buy when you want motherboards 3 generations back to back (sucks balls) but hey, if x570 are being held back by the ones who "might care" to ensure less problems, that might make sense why pretty much impossible to get an x570 motherboard currently (without taking the multi hundred $$$ will it work "trial"
ask Nvidia how "easy" it is to make Motherboards (or should say) easy to screw up than make work "proper"(they still suck arse at drivers for the so called "leader in graphics" leader means "best" not "sometimes better" silly Huang ..
notice Nv has not been in motherboard land for many years, shame they not put their attention to "quality" over parading on stage with leather coat and all (all those who bought RTX 2xxx or 1650/60 you got "super slapped" only cost you a few hundred $$$$ gave Nv a wicked belly laugh as well.
I just built a computer with a Ryzen 5 3600 and a MSI B450 Tomahawk motherboard. After upgrading the BIOS everything boots and performs flawlessly. I think this processor, boost and all, works just fine the way it is and I see no reason to change it.
A blink of an eye (20 miliseconds) is right way to go or a tad less, at least while operator is human. I always had three stage approach; set active idle frequency to one that which OS white noise (common usually appearing background tasks) can compiutionaly sustain on given CPU without crossing utilisation limit to ramp up, set it to rump up almost instantly to ideal leak/performance delta frequency (around 1.2 GHz on FinFET) & make sure it doesn't ramp up without a real need (hi utilised workload) behind some sain set limit (2~2.2 GHz).
"sustained CPU workloads will see largely imperceptible difference, while intermittent workloads such as games also won’t be affected as once the CPU gets over the initial base frequency ramp threshold it maintains the sub-1ms frequency change behaviour."
You know what will be affected? some benchmarks, and this is exactly why they didn't do it at launch.
You would hope anyone running benchmarks would know to set the power plan to Performance, but, yeah, I guess it wouldn't surprise me to learn that some testers don't.
To be clear, I was getting about 7330 in Cinabench 20. After updating bios from 7011 to 0808, and updating to new drivers, I now am gettings scores around 6950.
I can confirm. I address this in the end of the article. You can avoid the perf issue by just installing the new power plan and ignoring the rest of the driver package.
I have same MB and CPU. I can confirm that exactly same thing happened to me was getting scores around 7250 and 521 multi/single before bios upgrade. now its around 7000/500. Thanks AMD
We’ve updated our terms. By continuing to use the site and/or by logging into your account, you agree to the Site’s updated Terms of Use and Privacy Policy.
Facebook
Twitter
RSS
39 Comments
Back to Article
Gothmoth - Tuesday, July 30, 2019 - link
well i don´t see this new "released" driver at amd´s website...Andrei Frumusanu - Tuesday, July 30, 2019 - link
In the link of the source.Dark42 - Tuesday, July 30, 2019 - link
The 2nd graph (and text) shows the new version to idle at 3,0 GHz instead of ~2,2 GHz and taking longer to reach max frequencies.Am I missing something or is that just a step backwards?
Higher idle frequency means more power draw during idle and slower reaction to load doesn't sound better either.
This sounds somewhat like they try to fix the reported values when in reality the CPU works fine, just the software isn't up to date.
And instead of changing the software, now they change the CPU behavior making it somewhat worse but look better?
I'd like to at least have some options to adjust this behavior.
Scabies - Tuesday, July 30, 2019 - link
"If a processor core is truly running no workload, the processor core will be placed into the “CoreC6” (CC6) power-gated sleep state. Presently, AMD Ryzen Master is the only tool that can show
this CC6 state (Figure 3 attached). This state may be opportunistically engaged hundreds of times
per second. Voltages in this condition might be as low as 0.200V, and clockspeeds down to
0MHz/Sleep are also possible.
deil - Wednesday, July 31, 2019 - link
They went back to tackle some issue, (propably one that stops linux from using the CPU) I Think they know how to make this better, it just need some more time.ReeZun - Tuesday, July 30, 2019 - link
[a] I noticed that you only posted results for 'Windows Balanced'. With the new chipset drivers, is the response time also affected for 'AMD Ryzen High Performance' (and other power plans)?What if one prefers the 'former' behavior (of sub 1-ms response times)? Is there a way to adjust the settings (with the new chipset drivers) and/or will AMD keep an archive of different chipset driver releases (to suit the requirements of different users)?
hansmuff - Tuesday, July 30, 2019 - link
I would like answers to this as well. I very much LIKE the way this CPU behaved before this change. I WANT ultra fast ramp up so I get ultra fast responsiveness. I don't give a crap about 40C IDLE or really any of the complaints.Please AMD, if this has to be like that, please do publish an alternative for the old path.
The_Assimilator - Wednesday, July 31, 2019 - link
You're either a moron, illiterate, or a robot if you believe you can feel the difference between 840µs and 17.5ms.ImSpartacus - Wednesday, July 31, 2019 - link
The user probably can't, but their battery might."Race To Sleep" is a thing and I hope AMD doesn't hold back when they implement this stuff in mobile because it will result in battery life savings.
jaju123 - Wednesday, July 31, 2019 - link
I dont think race to sleep is affected as it will still change clocks/voltage in under 1ms in response to load alterations once the initial ramp is completed.Andrei Frumusanu - Wednesday, July 31, 2019 - link
Just use the High Performance plan. I added in a graph comparison between Balanced and High Perf on the new drivers.Phynaz - Tuesday, July 30, 2019 - link
Wait, new drivers to disable one of their performance features. After all the reviews.All those tests are 100% invalid now.
willis936 - Tuesday, July 30, 2019 - link
Yeah? The ramp up transient affects the long term load tests that every CPU test suite is composed of?deil - Wednesday, July 31, 2019 - link
the initial step, so probably all tests will see ~50ms difference or last 2 digit in 5000+ result score.same thing happens eveywhere on any software update, so its just whining.
nobody complained about meltdown patches that were in 1X%, this will end up sub 1%.
Phynaz - Wednesday, July 31, 2019 - link
YesAlexvrb - Wednesday, July 31, 2019 - link
No. You're being willfully ignorant. The tiny change in initial ramp-up time won't affect performance of any actual load. They said as much in the article.Now, if you WANT it to constantly ramp even when there's no load, due to monitoring programs, just switch from balanced to high performance.
Hyper72 - Tuesday, July 30, 2019 - link
Great, now we just need the motherboard manufacturers to fix their BIOS. I have a MSI B450M Gaming Plus I bought and paired with a Ryzen 3600 and latest R3K compatible BIOS on the 27th.In the last 3-4 days I've managed to boot once in ~50 tries, the rest of the time I get a red CPU debug LED and no boot. I have it bad though, most others with this issue get to boot somewhere between 1/10 and half the tries.
The MSI_Gaming subreddit is full of these stories about "compatible" BIOSes not managing to POST. Might put some pressure on MSI if websites like Anandtech decided to cover the issue...
shtldr - Wednesday, July 31, 2019 - link
Try switching the memory modules to different slots. I've had a soft reboot problem on MSI B450 Mortar. I've put the sticks to the slots closer to the CPU and it won't reboot properly (needed hard power cycle). After moving the memory to the slots further away from the CPU, there are no more problems.29a - Wednesday, July 31, 2019 - link
The manual should tell which slots to use based on the type of memory you have.drexnx - Tuesday, July 30, 2019 - link
this update definitely helped (but didn't entirely fix) the hyper-sensitive boosting on my X470 Aorus Ultra Gaming / 3700X combo; I still don't think the "it's supposed to idle at 30W! that's just how it works!" excuse is valid, but I'm sure they'll fix it in a month or two.zen 1 launch was just as janky but on the RAM side, which as been completely solid on the zen 2 launch for me.
Daver73 - Tuesday, July 30, 2019 - link
It doesn't idle at 30W--even without the update. As stated in the article, it ramps up with some over-aggressive monitoring software. By taking measurements at a very fast rate, the CPU interprets that as a heavy workload and ramps up accordingly.drexnx - Tuesday, July 30, 2019 - link
I can tell you that mine did/does, unless I switched windows power plan to "power saver" - I tried the 85/100 trick, it didn't work.Ryzen master said 41% of 88W PPT at idle
whatever, it's not a huge deal as long as they fix it (and I'm sure they will)
yetanotherhuman - Wednesday, July 31, 2019 - link
Don't use power saver, it artificially throttles the CPU back.I don't know if it's relevant today, but back in the Windows 2000 days, Power Saver didn't even save power, it instead ran a simple process to waste time and slow the effective responsiveness of the machine, but I could be misremembering it horribly.
drexnx - Wednesday, July 31, 2019 - link
I only do it when I'm done with it for the night and know it'll be idle until the next nightQasar - Thursday, August 1, 2019 - link
i just turn my comp off when i am done for the night :-) saves even more powerDragonstongue - Tuesday, July 30, 2019 - link
yeh, RAM wise they have seemingly sorted out all the ills (that I can tell)I think for Ryzen (Zen, Zen+, Zen2 as well as the APU based on such) there are just far far more "turn this on, turn that off" hundreds of times per second over "lots of space", I would imagine a little "burp" at this level is not a "burp" it becomes a grindy/grinding mess in real quick order (not good)
I know Zen1 and Zen2 share more than Zen2 and Zen+ do (from what I have read) so that likely means Zen 3 or was Zen 4 (final for AM4 according to AMD) will combine all the bits and piece to make the "true" enchilada Ryzen Ultra (something like that)
scheduler from Zen 1 and 2 are similar Zen+ is "it's own" but the turbo and CCX design more like Zen+ than Zen 1, a few new things added to the mix, so, they basically had to "reset" 3 times their base design.
they seem to be "figuring out" quicker but sadly is not like "the good ole days" where say AMD 700 through 900 were "near identical" from the motherboard side of things (YMMV if it acted simple as should have) certainly not a shelf full of buy when you want motherboards 3 generations back to back (sucks balls) but hey, if x570 are being held back by the ones who "might care" to ensure less problems, that might make sense why pretty much impossible to get an x570 motherboard currently (without taking the multi hundred $$$ will it work "trial"
ask Nvidia how "easy" it is to make Motherboards (or should say) easy to screw up than make work "proper"(they still suck arse at drivers for the so called "leader in graphics" leader means "best" not "sometimes better" silly Huang ..
notice Nv has not been in motherboard land for many years, shame they not put their attention to "quality" over parading on stage with leather coat and all (all those who bought RTX 2xxx or 1650/60 you got "super slapped" only cost you a few hundred $$$$ gave Nv a wicked belly laugh as well.
LordConrad - Wednesday, July 31, 2019 - link
I just built a computer with a Ryzen 5 3600 and a MSI B450 Tomahawk motherboard. After upgrading the BIOS everything boots and performs flawlessly. I think this processor, boost and all, works just fine the way it is and I see no reason to change it.ZolaIII - Wednesday, July 31, 2019 - link
A blink of an eye (20 miliseconds) is right way to go or a tad less, at least while operator is human. I always had three stage approach; set active idle frequency to one that which OS white noise (common usually appearing background tasks) can compiutionaly sustain on given CPU without crossing utilisation limit to ramp up, set it to rump up almost instantly to ideal leak/performance delta frequency (around 1.2 GHz on FinFET) & make sure it doesn't ramp up without a real need (hi utilised workload) behind some sain set limit (2~2.2 GHz).USGroup1 - Wednesday, July 31, 2019 - link
"sustained CPU workloads will see largely imperceptible difference, while intermittent workloads such as games also won’t be affected as once the CPU gets over the initial base frequency ramp threshold it maintains the sub-1ms frequency change behaviour."You know what will be affected? some benchmarks, and this is exactly why they didn't do it at launch.
Daver73 - Wednesday, July 31, 2019 - link
You would hope anyone running benchmarks would know to set the power plan to Performance, but, yeah, I guess it wouldn't surprise me to learn that some testers don't.The_Assimilator - Wednesday, July 31, 2019 - link
I don't see how idling at 3GHz is an improvement compared to idling at 2.2GHz, especially when Intel CPUs idle at a mere 800MHz.fatweeb - Wednesday, July 31, 2019 - link
On idle it goes to 0 GHz (CC6 sleep) which no software except Ryzen Master can read without also waking up the CPU.kliend - Wednesday, July 31, 2019 - link
I actually had a performance hit of about 5% in my testing. Using 3900x. I checked, double checked, and triple checked.Asus x570 Pro
Ryzen 3900x (stock cooler @ 100%)
Corsair Vengeance, 3000mhz, C15, 4x8GB
0808 BIOS version
Please see if you see the same? @Anandtech
kliend - Wednesday, July 31, 2019 - link
To be clear, I was getting about 7330 in Cinabench 20.After updating bios from 7011 to 0808, and updating to new drivers,
I now am gettings scores around 6950.
Andrei Frumusanu - Wednesday, July 31, 2019 - link
I can confirm. I address this in the end of the article. You can avoid the perf issue by just installing the new power plan and ignoring the rest of the driver package.kliend - Wednesday, July 31, 2019 - link
Thank you sir!chalupa13 - Thursday, August 1, 2019 - link
it has nothing to do with new chipset drivers. its the new AGESA 3AB microcode that lowers performance. roll back you bios to 7011 or 7004chalupa13 - Thursday, August 1, 2019 - link
I have same MB and CPU. I can confirm that exactly same thing happened to me was getting scores around 7250 and 521 multi/single before bios upgrade. now its around 7000/500. Thanks AMDMugur - Wednesday, July 31, 2019 - link
Why I still see the old horizontal UI in Ryzen Master, instead of the new vertical one? Is it specific to new Ryzen 3000 only? I have a 1600...