Der8auer: Only Small Percentage of 3rd Gen Ryzen CPUs Hit Their Advertised Speeds

[Midland Dog]

all i see is difference between "realistic" people and "pure optimistic" (to be polite i will call it that way)the former is a "oh, it can go up to" while the later is a "oh, it will, 100% definitely absolutely they better be, go to that max speed".


also i am wondering how much of that survey results are user side issue tied ... as you mentioned ... boost clock depend on a lot of parameter ... and are all but guaranteed...


ohhhh well i am fuming ... after Intel rather than AMD ... what's better? being able to OC as i want and reaching boost clock only sometime? or getting a K processor only to find later that a f!cked up microcode update messed up with OC and boost ... which make my 6600K a 6600 with some mhz in addiction to his base clock ...

get an asrock z170 board and bclock oc it bam 6600k again, your case is a 1 off, my 4690k can do 4.7ghz all core, amd cant even do it on 1 and they rate there chips for it, when you rely on higher ipc and lower clocks then small differences in clock are more sensitive to overall perf. i dont think there is a single 9900k that hasnt done 5ghz, im willing to bet that there are stuff all 3900x that can do 4.7ghz, the 3950x will be a huge meme

 

[GreiverBlade]

get an asrock z170 board and bclock oc it bam 6600k again

ohhhh my, changing mobo would solve my issue ... crap why didn't i... oh wait ... i had a ASRock mobo for testing before the actual one i have ... and the OC loss even occurred during her time ...
just in case my 6600K is a 6600k my mobo is a Z170 the OC did work neatly, on both mobo, until the famous microcode update Intel pushed via WUpdate (by mistake probably .... but for me it's still not corrected )

as for the rest ... still laughing nonetheless ... the issue quoted here for AMD is meager compared to what i am seeing with Intel and the 6600K issue i have ...is the last drop
actually i'd be fuming for all the perf loss due to Spectre/Meltdown/name other mitigation .... in addition to the aforementioned issue

at last i will look at my future R5 3600X or R7 3700/3800X boost clock as a "oh, it can go up to" (specially with a 3600X since i would get it for 40chf/$ less than what the 6600K did cost at the time )

 

[Tomgang]

It is a worrying results. I begin to wunder how many ryzen 9 3950X Will hit the 4.7 ghz boost clock. When so many CPU have a Hard time just hitting lower boost clock.

That really makes me wunder how 3950x Will boost and ever hit 4.7 ghz.

 

[Patriot]

quick reminder that 4550mhz is 4.6ghz and 4550 is an average clock not peak.

Yes its shady and not all mobo boards are hitting the same clocks as gamer nexus discovered.

 

[Vayra86]

id be fuming if i bought a chip expecting it to do 4.7ghz in st tasks (probs faster than 5ghz skylake at those clocks) only to be cucked out of up to 300mhz, at least when intel rates a chip to boost it just does it until it hits 100c

Too bad it does hit 100C at times, though, unless you throw a lot of cooling at it. And if you don't, you're stuck with allcore turbo's that are as 'low' as with AMD.

Its not that much greener on the blue side.

 

[TheGuruStud]

My 3600 doesn't hit anything above 4100mhz. Not even single thread low load. With custom water cooling. The box says 4200...

MB support is broken. Completely stock settings on original bios does it on mine.

 

[tony359]

This is a useless debate.

The box of any AMD cpu states clearly, boost up to Y speeds. Up to. That doesnt mean that any CPU will be capable of doing that.

How can you believe that?
In the UK broadband providers were advertising ‘up to’ speeds but the lawmaker changed the rules as some ISPs were only providing those speeds at night or for a fraction of the day.

‘Up to’ means ‘it will hit that speed under ideal circumstances but it won’t hold it for ever’. It cannot mean ‘it may not reach that speed’ - that’s called false advertisement.

 

[toxic80]

Wtf, I have 200 poins in cinebench R15 ST and 3200 in MT, 500 points in ST and 7362 points in cinebench r20 MT.
All the frustrated people need 4.6GHz.....Go buy 9900k for 5GHz....it's simple!!!!!!!

 

[lexluthermiester]

Der*auer seems to have missed something.

I have yet to see any of the Ryzen 3xxx CPU's fail to hit their specs. This is of course after lowering voltage to the correct settings. Motherboard makers are shipping boards with the default voltage set very high. This dumps WAY too much electricity into the CPU and as a result, too much heat to produced. When the voltage is set properly, the CPU's perform as expected.

Jay(JayzTwocents) already addressed in one of his recent video's and I'm going to echo those conclusions. This is a problem created by the board makers, not AMD.

 

[lewis007]

And what? Their selling like hotcakes so, whats the big deal! Its not the first time someone has exaggerated the performance metrics of a product and it won't be the last.

 

[GoldenX]

Nah, AMD should fix this, by either getting those clocks on all boards, or changing what the box says. Otherwise it's just stupid.

 

[yotano211]

And here I am fine with running 3.8ghz with 8 cores on a laptop. Do people really need that extra 50mhz.

 

[EarthDog]

Der*auer seems to have missed something.

I have yet to see any of the Ryzen 3xxx CPU's fail to hit their specs. This is of course after lowering voltage to the correct settings. Motherboard makers are shipping boards with the default voltage set very high. This dumps WAY too much electricity into the CPU and as a result, too much heat to produced. When the voltage is set properly, the CPU's perform as expected.

Jay(JayzTwocents) already addressed in one of his recent video's and I'm going to echo those conclusions. This is a problem created by the board makers, not AMD.

The motherboard, when on auto/optimized defaults, draws from the VID (aka, we call it Voltage I Demand in reference to the CPU) on the CPU, no? Across all X570 boards I've tested (6+ at this point), I've seen the voltage peak to similar amounts each time from within the BIOS (within windows, it has some control so that shouldn't be used). The BIOS 'is what it is' without power savings, etc...

 

[lexluthermiester]

The motherboard, when on auto/optimized defaults, draws from the VID (aka, we call it Voltage I Demand in reference to the CPU) on the CPU, no? Across all X570 boards I've tested (6+ at this point), I've seen the voltage peak to similar amounts each time from within the BIOS (within windows, it has some control so that shouldn't be used). The BIOS 'is what it is' without power savings, etc...

My experiences have been a bit different. Each of the boards in question have been all over the place with voltage when set at the defaults. Setting things manually seems to be the only way to get the voltage to the range it's supposed to be in(1.2-1.3). After this the CPU's behave as they should..

 

[EarthDog]

Are you checking in Windows or the UEFI?

I know with certainty the Intel chips have a VID for each multiplier. When left on auto, the mobo pulls the VID info from the CPU and runs it.

 

[lexluthermiester]

Are you checking in Windows or the UEFI?

Both, compared to each other.

I know with certainty the Intel chips have a VID for each multiplier. When left on auto, the mobo pulls the VID info from the CPU and runs it.

That's Intel. Ryzen is a bit more complicated and I think the mobo makers are struggling to get it right. I'll say this, running any Ryzen 3xxx above 1.35v(which it shouldn't require) needs better cooling than the stock solution. High-end heatpipe or liquid cooling is required to keep temps down and clocks at or above stock. OCing requires liquid cooling. My theory is that the limiting variable is the CCX die. It's still at 14nm. If AMD were to drop the CCX die to the 7nm node, Ryzen 3xxx would be much more capable. That combination may have been the only thing they had at the time, but it's still less than optimal.

 

[EarthDog]

I can't keep a 3700x under 90C (AIDA64 default stress test) with a H150i at 1.35V. The stock potatos should be done sooner.

Maybe the AIBs are putting in some special sauce.. I don't know. In my yesting so far, after setting optimized defaults and then enabling PBO, there are few gains... some losses even. Not much, maybe 2% here or there, but still, odd.

Anyway, pretty sure the CPU is calling the shots when on auto... @TheLostSwede - any ideas there for these?

 

[TheoneandonlyMrK]

I can't keep a 3700x under 90C (AIDA64 default stress test) with a H150i at 1.35V. The stock potatos should be done sooner.

Maybe the AIBs are putting in some special sauce.. I don't know. In my yesting so far, after setting optimized defaults and then enabling PBO, there are few gains... some losses even. Not much, maybe 2% here or there, but still, odd.

Anyway, pretty sure the CPU is calling the shots when on auto... @TheLostSwede - any ideas there for these?

From what I've noted I would agree with Lex, the heat density on my 2nd gen ryzen makes it actually difficult(like high spec intel chips) to actually get the heat out quick enough, I have exaggerated cooling if applied just to the CPU but I can overegg past it , I have found similar to hardwareCanucks tim application and clamping pressure as well obviously as ambient temps and overall cooling performance matters.

Exactly how many would have actually adequate cooling i wonder?, it isnt cheap?.

I have seen 4.5 clocks on mine,, but it wouldn't be there the second day, in fact, it does quite often settle at a different speed each day and to me seems to vary greatly, that said the vast number of slight configurations I have now ran it with will have also affected that, but now and then i settle on a config for a bit thinking it stable etc, the typical variance is still noted between 4.125 and 4.325, , 4 under sustained load.

Hopefully grabbing my new chip tomorrow, I could be screaming a different tune tomorrow

 

[R-T-B]

Chips run at advertised speeds

The problem here is no, they don't always.

Der*auer seems to have missed something.

I have yet to see any of the Ryzen 3xxx CPU's fail to hit their specs. This is of course after lowering voltage to the correct settings. Motherboard makers are shipping boards with the default voltage set very high. This dumps WAY too much electricity into the CPU and as a result, too much heat to produced. When the voltage is set properly, the CPU's perform as expected.

Jay(JayzTwocents) already addressed in one of his recent video's and I'm going to echo those conclusions. This is a problem created by the board makers, not AMD.

Actually, I'd bet the AGESA package may be doing it, but I could be wrong. Either way, it does sound like a software solution is possible. My issue is it should work from day 1... this is a basic, advertised spec.

 

[Darmok N Jalad]

From what I've noted I would agree with Lex, the heat density on my 2nd gen ryzen makes it actually difficult(like high spec intel chips) to actually get the heat out quick enough, I have exaggerated cooling if applied just to the CPU but I can overegg past it , I have found similar to hardwareCanucks tim application and clamping pressure as well obviously as ambient temps and overall cooling performance matters.

Exactly how many would have actually adequate cooling i wonder?, it isnt cheap?.

I have seen 4.5 clocks on mine,, but it wouldn't be there the second day, in fact, it does quite often settle at a different speed each day and to me seems to vary greatly, that said the vast number of slight configurations I have now ran it with will have also affected that, but now and then i settle on a config for a bit thinking it stable etc, the typical variance is still noted between 4.125 and 4.325, , 4 under sustained load.

Hopefully grabbing my new chip tomorrow, I could be screaming a different tune tomorrow

Makes me wonder how we will be able to keep up with the cooling on these CPUs as dies continue to shrink and thermal density increases. I can certainly see overclocking fading away, as chips will simply not take kindly to any extra thermal load. I know process advances help, but they are packing billions of transistors into a very small area.

 

[EarthDog]

From what I've noted I would agree with Lex, the heat density on my 2nd gen ryzen makes it actually difficult(like high spec intel chips) to actually get the heat out quick enough, I have exaggerated cooling if applied just to the CPU but I can overegg past it , I have found similar to hardwareCanucks tim application and clamping pressure as well obviously as ambient temps and overall cooling performance matters.

Exactly how many would have actually adequate cooling i wonder?, it isnt cheap?.

I have seen 4.5 clocks on mine,, but it wouldn't be there the second day, in fact, it does quite often settle at a different speed each day and to me seems to vary greatly, that said the vast number of slight configurations I have now ran it with will have also affected that, but now and then i settle on a config for a bit thinking it stable etc, the typical variance is still noted between 4.125 and 4.325, , 4 under sustained load.

Hopefully grabbing my new chip tomorrow, I could be screaming a different tune tomorrow

My point is that even with the right motherboard, right cooling, right parameters that AMD set forth, reaching max boost still hasn't happened (for me, and many many others, so far). The 1.35V overclocking is my dud of a sample at 4.25 GHz all c/t.

I'm not sure what ambient solution could keep the CPUs under 60C... whoever mentioned that.

This is also Zen 2 we are talking about, not Zen+. These work different;y.

 

[Flyordie]

Der*auer seems to have missed something.

I have yet to see any of the Ryzen 3xxx CPU's fail to hit their specs. This is of course after lowering voltage to the correct settings. Motherboard makers are shipping boards with the default voltage set very high. This dumps WAY too much electricity into the CPU and as a result, too much heat to produced. When the voltage is set properly, the CPU's perform as expected.

Jay(JayzTwocents) already addressed in one of his recent video's and I'm going to echo those conclusions. This is a problem created by the board makers, not AMD.

I second this.

Even my FX8320 (which is still in service with my neighbors PC).. Gigabyte has default voltage at 1.4125V. Yet, it'll run 1.28V just fine at stock clocks up to about 4.025Ghz. So its entirely possible that this is the issue.

 

[Minus Infinity]

You would think they would sample a 1000 or so production ready CPUs, and from the bell curve distribution of attainable max clock speeds, base their rating on the peak of the curve, but it looks like they've done it on wing of the curve where only a small % can make those speeds. It's false advertising plain and simple. I don't really care becuase 25-100MhZ makes bugger all difference in the real world and the real world benchmarks are showing it to be a fantastic update, but we have laws about what you can claim. You don't see Ford saying Mustang can do up to 200mph but can only only hit 170mph and then say well we said "up to". AMD are just opening themselves up to a lawsuit for no good reason other than maybe bragging rights.

 

[Darmok N Jalad]

You would think they would sample a 1000 or so production ready CPUs, and from the bell curve distribution of attainable max clock speeds, base their rating on the peak of the curve, but it looks like they've done it on wing of the curve where only a small % can make those speeds. It's false advertising plain and simple. I don't really care becuase 25-100MhZ makes bugger all difference in the real world and the real world benchmarks are showing it to be a fantastic update, but we have laws about what you can claim. You don't see Ford saying Mustang can do up to 200mph but can only only hit 170mph and then say well we said "up to". AMD are just opening themselves up to a lawsuit for no good reason other than maybe bragging rights.

No, but you need pretty idea conditions to get your mustang to its top rated speed.

 

[Deleted member 185158]

When I set all defaults, my single core boost hit the rated speeds. just doesn't say on the box only one core goes that speed.
Tested and verified using PiMod 32m. (Just had to reset all defaults and check my boot settings)