-
Ever wanted an RSS feed of all your favorite gaming news sites? Go check out our new Gaming Headlines feed! Read more about it here.
[Der8auer] Good number of Ryzen 3000 chips does not reach advertised boost clock
- Thread starter Serious Sam
- Start date
Well, that's not good.
There's no point changing from my 2700X is there? I feel like the only upgrade I want to make is jumping on the RTX series as I'm currently rocking a 1080.
There's no point changing from my 2700X is there? I feel like the only upgrade I want to make is jumping on the RTX series as I'm currently rocking a 1080.
Not hugely, but if your motherboard supports the new CPUs with a BIOS update, like most will, no harm in setting up some deal alerts to pick up a cheap one down the road.
This is mostly affecting the 3900X. And then those restults seem... questionable.
With AGESA updates this should hopefully be mitigated in the near future tho.
With AGESA updates this should hopefully be mitigated in the near future tho.
I switched from a 2500X to a 3700X and could not be happier.
Without knowing what cooling the users used, what kind of motherboard (and therefore how good the VRMs are) were used, and other influences (see the guru3D news) these numbers say absolutely fucking nothing.
Laughable
Laughable
Is this just the 39xx processors? Cause all that analysis it just denotes that series of processors, not the entire range of Ryzen 3000s. Headline may be intentionally misleading if that is the case. Not like the 39xx processors were ever really the smart call over a 3700x, most in-game testing put the 3700x just a few percentage points behind the 39xx. Basically negligible and margin of error stuff.
Also, yeah, what Mazzle said.
Also, yeah, what Mazzle said.
Yeah I have no issue hitting 4.4Ghz even with the wraith prism cooler. AMD did not skimp on the cooler like Intel who has been skimping for like 3-4 generations now.
Yeah the included cooler is quite decent. I just have the Noctua lying around because I wanted to build an Intel rig and decided to go Ryzen last minute. Kinda overkill to be honest.
Almost everything you said here is false. As for finding the results questionable or not, I think he was very candid about the way he picked them and some models do have a higher sample pool than others but most of them, to varying extent, fail to meet advertised boost speeds.
Some representatives have stated that this is not something they believe can be fixed with some patch. It seems to be due to the silicon quality.
AMD is false advertising the specifications of their products and it is not the first time they have done this. Google "AMD Bulldozer".
Before you think I am biased against AMD and try to use that to discredit me, as you discredited all the other information in the video, I am typing this on a brand new AMD computer with a 3700x Ryzen. I am very happy with its performance, but that doesn't mean that AMD's false advertising is alright or acceptable.
Woah. What are your settings? I still haven't tried to do manual OC, to be honest.
meh it really doesn't matter the performance/price of those 3000 chips are quite exceptional even if some of them don't hit the ''advertised'' boost speed
I mean there some some of these situations back in Sandy bridge actually... Small % to justify especially with 0.5 differences and this is just on the 3900x.
No need to. PBO alone is doing such a good job, that you don't need manual OC.
Again, without knowing the cooling solutions, the Motherboard, the Case, ambient temperature and whatnot there results are pretty useless
He removed any results that had ridiculous cooling solutions or BIOS settings that may have impacted results. Obviously that doesn't mean that some of the users couldn't have just outright lied, and there are outliers among his results, but all of them seem to fall within a pattern that allows us to make sound conclusions from.
Either way, I trust der8auer's ability to parse and judge the numbers more than a random forum poster. Given the potential liability of bringing forth this numbers, and given that he is a businessman who has always been frank and maintained good work ethic, at least as far as I know, I trust his results and findings.
It's important to contextualize what this means - it does not mean that the reviews you saw were invalid, nor does it mean performance is bad. The silicon is as was sent to reviewers, and all conclusions drawn from these reviews are valid. Der8auer emphasizes this in his video on the subject (available in English and in German). It only means that there has been false-advertising with regard to rated clock speeds - AMD should simply have lowered the rated clock speeds on the packaging and marketing materials.
If by this you mean it is mostly not affecting other chips, that is a complete misreading of the data. I linked the video above in this post, and you can see for yourself.
3900X - 94.4% below spec
3800X - 73.3% below spec
3700X - 85.3% below spec
3600X - 90.6% below spec
3600 - 50.2% below spec
The only chip to get close to 50% hitting advertised clocks according to this is the 3600. It got 49.8% matching spec, 50.2% not. Even if we assume these results are trash and the true value for all chips was 10x lower, 5.02% - 9.4% failing to meet spec would still imply that between 1 in 20 and 1 in 10 people got a dud, and that would imply disastrous reliability.
Whether or not they can release an update that results in it hitting the boost clock, it's not likely to improve performance to any statistically detectable degree and may degrade longevity by forcing very high voltages to achieve it. These boost clocks are edge case, only being maintained for very short bursts, and only apply when a single core is being stressed but no others are.
If by this you mean it is mostly not affecting other chips, that is a complete misreading of the data. I linked the video above in this post, and you can see for yourself.
3900X - 94.4% below spec
3800X - 73.3% below spec
3700X - 85.3% below spec
3600X - 90.6% below spec
3600 - 50.2% below spec
The only chip to get close to 50% hitting advertised clocks according to this is the 3600. It got 49.8% matching spec, 50.2% not. Even if we assume these results are trash and the true value for all chips was 10x lower, 5.02% - 9.4% failing to meet spec would still imply that between 1 in 20 and 1 in 10 people got a dud, and that would imply disastrous reliability.
Whether or not they can release an update that results in it hitting the boost clock, it's not likely to improve performance to any statistically detectable degree and may degrade longevity by forcing very high voltages to achieve it. These boost clocks are edge case, only being maintained for very short bursts, and only apply when a single core is being stressed but no others are.
He also points out that the risk that dissapointed users could have faked the actual numbers; it's worth mentioning i guess.
I'm a huge AMD fan and when i can, i support them; but i have to say this launch has not been optimal. I follow the AMD subreddit and all i see is people complaining about systems not POSTing, huge booting times and BSOD. Even RX 5700 is fucked with the current drivers apparently.
I myself have switched back to Intel cause of two broken motherboards on Ryzen 3600 and the difference in polishness is incredible.
I look forward to the next months to see if things improve with bios releases.
I'm a huge AMD fan and when i can, i support them; but i have to say this launch has not been optimal. I follow the AMD subreddit and all i see is people complaining about systems not POSTing, huge booting times and BSOD. Even RX 5700 is fucked with the current drivers apparently.
I myself have switched back to Intel cause of two broken motherboards on Ryzen 3600 and the difference in polishness is incredible.
I look forward to the next months to see if things improve with bios releases.
I was having similar issues and turns out after weeks of trying to figure it out it was just a faulty SATA cable on my SSD. Since then my computer has been working fine.
Admittedly AMD doesn't have the best track record when it comes to reliability but I can't fault them for the issue I was having. I can't fault them for not knowing the impact of certain bios settings I forgot to adjust. I can't imagine I am the only person who had similar issues that are no fault of AMD. While the experience may be easier out of the box with Intel, you can't fault AMD for user errors and without knowing the problems these people were having and parsing them or at least reporting them in some form, it's kind of a vapid critique.
OP
OP
All good.
I think one of the more important things here is that AMD has now gone one step further than Intel already did a few generations ago. In the past, boost clocks were symmetrical - they meant that you could get the rated clock on all cores, and that all chips could be expected to hit this with stock cooling in reasonable circumstances (e.g. not in the middle of a 50 degree desert). A few years ago, Intel started advertising single core turbo clocks - meaning the peak clock speed when only 1 core was stressed. AMD's 2nd generation Ryzen processors also adopted this policy of advertising single core boost rather than all core boost.
Now, with Ryzen 3000, they are advertising single core peak boost (only maintainable for very short periods) that a large portion (whether it's only 10% or maybe up to 90%) can't reach. The advertised turbo speeds for Intel and AMD have become a bit of a joke over the last few years, and this is really just the pinnacle.
My one problem has been the machine not cycling or shutting down or restarting after I click to do so in windows. It will leave the windows blue screen, stop displaying to the monitor, but the board, power, fans etc. will still be on. I asked another ryzen user in NuMiQ and he has the same problem. Anyone else getting this here?
I eventually just have to manually power it down with holding the power button on the case for 10 seconds.
Consistent with what I'm seeing with my 3700x: single core boost never exceeds 4367 MHz. PBO increases multicore boost under heavy load but has no effect on absolute max boost (may even decrease it a little, oddly). Auto OC doesn't seem to do anything other than cause a rise in voltages under load.
I think this is shady from AMD. What potentially makes things worse is that AGESA 1.0.0.3 onwards apparently tightened thermal limits and tweaked the boost algorithm to boost less aggressively in order to preserve chip longevity. That would suggest to me that the silicon doesn't quite perform as well as hoped. Either that or potentially something far more questionable was going on. Whatever the case, it makes it quite unlikely that a future AGESA will improve boost numbers.
Despite being generally happy with the performance, I'm quite annoyed by this. The stated boost speed should be achievable at stock (no PBO, no Auto OC) with stock cooling, not be some theoretical figure that relies upon winning the silicon lottery and having significantly better cooling than is supplied with the processor.
I think this is shady from AMD. What potentially makes things worse is that AGESA 1.0.0.3 onwards apparently tightened thermal limits and tweaked the boost algorithm to boost less aggressively in order to preserve chip longevity. That would suggest to me that the silicon doesn't quite perform as well as hoped. Either that or potentially something far more questionable was going on. Whatever the case, it makes it quite unlikely that a future AGESA will improve boost numbers.
Despite being generally happy with the performance, I'm quite annoyed by this. The stated boost speed should be achievable at stock (no PBO, no Auto OC) with stock cooling, not be some theoretical figure that relies upon winning the silicon lottery and having significantly better cooling than is supplied with the processor.
The motives are fairly obvious methinks - the desire to see numbers increase for marketing, and the change in how precision boost algorithms work.
The precision boost / PBO is a factor. These chips are, effectively, making overclocking borderline irrelevant, because they're basically dynamically overclocking all the time, and this OC is what their top speed is, ergo that's what they rate their boost clock on. They have chosen to extract every inch of performance stock that they can (limited by cooling and eventually voltage) to better compete with Intel, but this means the random chip to chip variance that used to only be a concern for overclockers now means you may not be able to reach your advertised clock speeds.
If these stats were close to accurate, they would have to drop rated boost clocks by up to ~400mhz for the vast majority of 3900X to hit it, and advertising "12 cores @ 4.2GHz!" doesn't sound nearly as exciting as "12 cores @ 4.6GHz!". Looking on store shelves a casual customer might see that it doesn't look that much more exciting than the predecessor boost clocks, and assume not much had changed. Basing clocks on the best silicon rather than the worst silicon lets them seem much more impressive at a glance. The R5 3600 would have to be rated at something like ~4-4.1GHz to get to 95-99% of the chips hitting advertised clocks, which would make it seem like it was the same number of cores and 100mhz slower than the 3600X - basically a disaster for casual marketing even though the overall perf is far superior in real world applications.
I had that problem on a 2700x. I think I disabled fast startup to fix it.
Unless you need ray tracing in games like control, BF5, metro etc, I would wait.
I'd imagine raytracing cards will mature quite quickly once next gen consoles are out.
Completely forgot to tell you that I managed to fix this by disabling CSM as a boot option. Restarts working fine now with only Uefi enabled
Well, it depends on what you value. The new chips are significantly better in pretty much every performance benchmark.
I do not have such an issue. I'm running a 3900x with an ASUS TUF Gaming X570 Plus. I did upgrade the bios as well as the X570 chipset driver since, but I never had such an issue, no matter the bios/driver. So far it has been a problem free experience for me. Memory is overclocked (mhz increase with much tighter timing), so is the cpu (via AutoOC + PBO). Everything is stable af.
---
Anyway, on topic. I did not specifically took a note of single threaded speed on my system, but even if that was the case, that my cpu would not achieve the rated clock speed, it would not really bother me to be honest. I noticed such a huge gain in performance switching from my overclock 3770k to the 3900x that at this point, it would be dishonest of me to be annoyed at this while I had nothing but a great experience since the day I swapped the components and I'm not really using any single thread app that would greatly benefit from a clock boost from 4.475-4.5ghz to 4.6ghz.
I understand that it's still not cool of AMD to advertise a speed that isn't achievable by most. I'd rather have the advertised speed than not. That said. I'm not going to be upset over this.
Interesting.
I have a new Ryzen system for two weeks now. 3700x.
Mine only boosts to 4.3 max, mostly it hovers around 4.25. Water cooled.
Being used to Intel the whole PBO things is somewhat confusing. I changed some settings in the BIOS but may have made things worse.
Question: Anyone running an ASUS Board (I have a X570-P) and is willing to share their PBO/BIOS Settings?
I have a new Ryzen system for two weeks now. 3700x.
Mine only boosts to 4.3 max, mostly it hovers around 4.25. Water cooled.
Being used to Intel the whole PBO things is somewhat confusing. I changed some settings in the BIOS but may have made things worse.
Question: Anyone running an ASUS Board (I have a X570-P) and is willing to share their PBO/BIOS Settings?
I don't have an Asus board but bear in mind what PBO does: PBO removes the power consumption limits that are set by the chip specification. Because of those limits there are scenarios where the CPUs clock speed will become limited not by heat but by not being able to draw enough power. PBO alleviates that by letting the CPU draw as much power as the board can deliver. Where PBO is not terribly useful is for the single core maximum boost speed as that's not going to max out CPU power consumption.
In theory Auto OC is supposed to increase max single core boost speed, but it doesn't help me much and seems to increase core voltage. That worries me enough to stop me using it.
Here's my max boost frequency findings between AGESA 1.0.0.3 and AGESA 1.0.0.3AB
BIOS v17 (AGESA 1.0.0.3):
With BIOS v17, the CPU almost always hit the advertised max boost frequency of 4,400 MHz, on at least 4-5 cores during Cinebench R15 & R20 single-threaded workload with the ambient temp being around 25°C or lower. Ambient temperature plays a role in boost frequency, unsurprisingly. The warmer the room (in my case around 30°C), the more aggressively the CPU appeared to throttle, therefore, the CPU didn't always reach the max advertised boost clock of 4.4 GHz (it would usually be in the 4,250-4,300 MHz range max). I think the obvious solution here is to use aftermarket, air or liquid CPU cooler - instead of the stock Wraith Prism - which would allow the CPU to hit its max boost frequency more often. So, it's a success. The CPU does reach its max boost clock with AGESA 1.0.0.3 as long as the CPU temp is kept lower.
BIOS v18 (AGESA 1.0.0.3AB):
With BIOS v18, the CPU never hit the max 4,400 MHz frequency - not even on one core. It would max out at 4.374.9 MHz or lower. No matter the workload, whether it's Cinebench R15 & R20 single-threaded or outside of CB in lightly-threaded workloads, even the ambient temperatures (of around 20-21°C) didn't matter this time around. So, I concluded that AGESA 1.0.0.3AB to be the culprit for not hitting the advertised boost frequency on any workload. So it's a failure. The CPU does not reach its max boost clock with AGESA 1.0.0.3AB despite being under colder conditions.
TL;DR:
With AGESA 1.0.0.3 - Ryzen 7 3700X hits max 4.4 GHz in single-thread tests on MSI B450M Mortar motherboard.
With AGESA 1.0.0.3AB - Ryzen 7 3700X fails to hit max 4.4 GHz in single-thread tests on MSI B450M Mortar motherboard.
- CPU: Ryzen 7 3700X
- Mobo: MSI B450M Mortar
- Cooler: Wraith Prism (Stock)
- RAM: 2x8GB 3200 MHz CL14
- BIOS versions tested: v17 & v18
- Default motherboard settings (No PBO, Auto OC)
- Windows 10 (1903)
- AMD Chipset v1.8.19
BIOS v17 (AGESA 1.0.0.3):
With BIOS v17, the CPU almost always hit the advertised max boost frequency of 4,400 MHz, on at least 4-5 cores during Cinebench R15 & R20 single-threaded workload with the ambient temp being around 25°C or lower. Ambient temperature plays a role in boost frequency, unsurprisingly. The warmer the room (in my case around 30°C), the more aggressively the CPU appeared to throttle, therefore, the CPU didn't always reach the max advertised boost clock of 4.4 GHz (it would usually be in the 4,250-4,300 MHz range max). I think the obvious solution here is to use aftermarket, air or liquid CPU cooler - instead of the stock Wraith Prism - which would allow the CPU to hit its max boost frequency more often. So, it's a success. The CPU does reach its max boost clock with AGESA 1.0.0.3 as long as the CPU temp is kept lower.
BIOS v18 (AGESA 1.0.0.3AB):
With BIOS v18, the CPU never hit the max 4,400 MHz frequency - not even on one core. It would max out at 4.374.9 MHz or lower. No matter the workload, whether it's Cinebench R15 & R20 single-threaded or outside of CB in lightly-threaded workloads, even the ambient temperatures (of around 20-21°C) didn't matter this time around. So, I concluded that AGESA 1.0.0.3AB to be the culprit for not hitting the advertised boost frequency on any workload. So it's a failure. The CPU does not reach its max boost clock with AGESA 1.0.0.3AB despite being under colder conditions.
TL;DR:
With AGESA 1.0.0.3 - Ryzen 7 3700X hits max 4.4 GHz in single-thread tests on MSI B450M Mortar motherboard.
With AGESA 1.0.0.3AB - Ryzen 7 3700X fails to hit max 4.4 GHz in single-thread tests on MSI B450M Mortar motherboard.
Last edited:
From my knowledge eariler BIOS revisions actually hit the advertised numbers but later one's dont.
Hopefully it's fixed in the latest AGESA update.
I will say the marketing of PBO was really shady. They were advertising numbers that people straight up won't hit like 4.8GHz.
Hopefully it's fixed in the latest AGESA update.
I will say the marketing of PBO was really shady. They were advertising numbers that people straight up won't hit like 4.8GHz.
Yeah. The AMD rep over on reddit has been pretty much silent since the claims for PBO and Auto OC didn't jibe with reality. PBO does work under certain loads, but the claims of Auto OC gaining 200 MHz were bullshit. Maybe on a liquid nitrogen cooled system, but not on anything widely used.
Assuming AMD weren't lying through their teeth, the obvious assumption is that they discovered a worrying number of chips frying in testing. I'm sceptical of this fix they're coming out with but we'll see. If boost speeds do increase, it'll be interesting to see if benchmark scores do the same.
The fix will probably get chips up to actual boost speeds.
PBO will always be bullshit though. No way 3700x/3900x are hitting 4.8 in any actual workload without hitting crazy voltages.
4.8 no, but it does gain me 200 MHz when running Cinebench. Without it the power supply limits kick in hard.
So noob question. I'm about to update the bios of my b350 mortar, which bios update should I get for that and where?
I want to upgrade my ryzen 3 2200g to a r5 3600 and want to see if I can use the old motherboard, as money is tight and I figure why not try that first, but I'm really new to this, and especially that bios stuff is difficult to figure out
EDIT:
Note: there are two B350M Mortar models.
- MSI B350M Mortar
- MSI B350M Mortar Arctic (white color)
ORIGINAL:
There's only one BIOS update for your motherboard, which brings in support for Ryzen 3000 series. And it's in beta.
BIOS files can always be found at your motherboard vendor's website, under your product page's 'Support' section.
Here: https://www.msi.com/Motherboard/support/B350M-MORTAR
Download the "7A37v1MM(Beta version)" BIOS file from there, and update it before installing your Ryzen 5 3600 into the motherboard.
This is a good tutorial on how to update BIOS, in case you didn't know:
How to Update A B350 Bios For Ryzen
Last edited:
Thank you! Yes, I have the regular B350 Mortar.
But now I'm wondering if it's ok to update to that bios while keeping the old cpu for a while or is that counterproductive?
I'll get the new cpu asap if the update goes smooth but yeah I haven't ordered it yet.
September 10th isn't a long wait. Hopefully everyone is satisfied with the results.
This was posted today
So after a couple of weeks I got my 3600 running the way I'm comfortable with it. According to Der8aure's video I got a chip from the better half for once maybe?
My motherboard is a regular MSI B450 Tomahawk with the latest non beta bios (7C02v1A). I disabled PBO, assigned 4200MHz to each ccx in Ryzen Master with 1.33125V cpu voltage and it keeps the clock speed 99.9% of the times under workloads like the Ryzen Master stress test, x264 rendering, cinebench R15 or Battlefield 1.
PBO is a bit scuffed in my case since it didn't push higher frequencies, but it sent 1.45 V to the CPU and the whole thing got like 10°C hotter for no real gain.
I could maybe push for higher clocks but I'm already obscenely gpu limited with my GTX 1060 and I'd have to push further the EDC and PPT limits that I don't really know how exactly I'm suppose to do that with nor I'm sure if I should touch such settings at all.
That AMD marketing video where they explain pbo was definitely not accurate with the numbers.
My motherboard is a regular MSI B450 Tomahawk with the latest non beta bios (7C02v1A). I disabled PBO, assigned 4200MHz to each ccx in Ryzen Master with 1.33125V cpu voltage and it keeps the clock speed 99.9% of the times under workloads like the Ryzen Master stress test, x264 rendering, cinebench R15 or Battlefield 1.
PBO is a bit scuffed in my case since it didn't push higher frequencies, but it sent 1.45 V to the CPU and the whole thing got like 10°C hotter for no real gain.
I could maybe push for higher clocks but I'm already obscenely gpu limited with my GTX 1060 and I'd have to push further the EDC and PPT limits that I don't really know how exactly I'm suppose to do that with nor I'm sure if I should touch such settings at all.
That AMD marketing video where they explain pbo was definitely not accurate with the numbers.
I recently returned my 3900x and this was a part of the reason, although it was mostly for sitting at 60c idle. My issue with this launch is that I just can't be bothered to do all the tinkering and fucking around that I would have enjoyed in college. My new 9900k took ten minutes to overclock to 5.0 and now never goes above ~50. It just worked.
I will also say that AMD deploying AGESA 1.0002 to reviewers and then nerfing the performance in consumer AGESA 1.0003+ is reaaalllly fishy. People should definitely give them the benefit of the doubt until Sept. 10th though.
I will also say that AMD deploying AGESA 1.0002 to reviewers and then nerfing the performance in consumer AGESA 1.0003+ is reaaalllly fishy. People should definitely give them the benefit of the doubt until Sept. 10th though.
Just a bump to say that the new AGESA, 1.0.0.3ABBA has been released. I'm using the beta BIOS for my Aorus Elite and 3700X and I have witnessed it hitting 4.4Ghz now. I don't think they're just reporting higher numbers either as benchmark scores have also seen a slight increase.