Silicon Lottery Ryzen 3000 numbers

[nicalandia]

https://siliconlottery.com/pages/statistics
3700X 4.05GHz 1.237V 100%
3700X 4.10GHz 1.250V Top 74%
3700X 4.15GHz 1.262V Top 21%

3800X 4.20GHz 1.275V 100%
3800X 4.25GHz 1.287V Top 53%
3800X 4.30GHz 1.300V Top 20%

3900X 4.00GHz 1.200V 100%
3900X 4.05GHz 1.212V Top 87%
3900X 4.10GHz 1.225V Top 68%
3900X 4.15GHz 1.237V Top 35%
3900X 4.20GHz 1.250V Top 6%



All core OC isn't viable it looks like, better try per CCX, also the v core is rather low

 

[Mopetar]

I think they mentioned that the lack of an AVX offset holds the AMD chips back a little bit.

I don’t know if it changes the bins at all, but it’s probably possible to get at least another 100 MHz out of these on average if you exclude AVX from the workload.

 

[0ldman79]

I wish they had 3600x numbers as well.

I know it's a lower tier chip, likely limited, but 3 cores per CCX may offset the heat generated a bit, so it may overclock decently.

The FX 6300 would clock quite well compared the 8 core FX. Not sure if that will repeat itself or not, but I'm curious.

 

[moinmoin]

I think they mentioned that the lack of an AVX offset holds the AMD chips back a little bit.

I don’t know if it changes the bins at all, but it’s probably possible to get at least another 100 MHz out of these on average if you exclude AVX from the workload.

For static all core OCs the lack of offsets hurts, but then again static all core OCs on Matisse are pretty useless to begin with. I'd love to see statistics whether (or rather how much) a high static all core OC ensures stock PB2 using higher frequencies.

 

[EXCellR8]

Abandoned my static OC and my chip is boosting up to 4.3ghz on the regular with new X370 BIOS...

[EDIT] Will occasionally touch just under the advertised max ~4.4ghz boost under certain workloads; this is with no set BIOS options except for XMP profile on the FlareX. Not bad; ASRock looks to have overhauled the BIOS for the board--it even looks different.

Is anyone running 3xxx on the X370 Taichi?

 

[Fanatical Meat]

Beside the e-peen aspect I have no idea why someone would pay extra. I sort of understand the $30 premium but you are effectively opting out of the potential to get better doing that.
Why pay the extra money to get what is likely an a performance boost in benchmarks and no real world gain

 

[IEC]

3800X is the best option if you want the best chance of highest all-core clocks out of the box and with OC.

 

[SPBHM]

I think your time is better spent on ram timings than this kind of overclock,

 

[nicalandia]

Retailers Begin to Sell Pre-Binned AMD Ryzen 3000 CPUs
https://www.anandtech.com/show/14737/retailers-begin-to-sell-prebinned-amd-ryzen-3000-cpus

 

[DrMrLordX]

Hmm. My 3900x will do 4.25 GHz easily in the most horrendous of AVX2 workloads (Prime95 Small ffts). Requires 1.25v . It gets warmer than I'd like, but it'll do it.

For CBR20 (also AVX2) it can do 4.4 GHz. I think Silicon Lottery is sandbagging a bit here. Either that or i'm just enjoying the effects of using an overkill custom WC rig.

 

[nicalandia]

I think Silicon Lottery is sandbagging a bit here. Either that or i'm just enjoying the effects of using an overkill custom WC rig.

They are, I mean CPU Vcore: 1.250V is listed for all of the OC numbers they have, so more can be expected because the max safe all core voltage for Zen 2 is 1.325v

For example the 4.2 Ghz OC on 3900X
https://siliconlottery.com/collections/all/products/3900x42g?variant=29472818823254

 

[Markfw]

According to Ryzen Master, my cpu @100% load is using 1.3-1.32 vcore @ 4025 mhz

All stock

 

[lopri]

Their numbers make sense to me because my 3700X fails in AVX2 stress test at 4.3 GHz even with 1.45V. It's due to uneven quality among cores. Some are quite a bit better/worse than the others.

 

[DrMrLordX]

They are, I mean CPU Vcore: 1.250V is listed for all of the OC numbers they have, so more can be expected because the max safe all core voltage for Zen 2 is 1.325v

For example the 4.2 Ghz OC on 3900X
https://siliconlottery.com/collections/all/products/3900x42g?variant=29472818823254

The thing that jumps out at me is that SL isn't telling people what clockspeeds people can realistically expect from PB2/XFR during an AVX2 workload. Even with good cooling, you typically don't get "real" clocks in the 4.2 GHz range using default boost behavior running something like Prime95 or y-cruncher. It does not happen. 4.2 GHz is a really high clockspeed for a workload like that - far above what can be realistically sustained from boost.

Their numbers make sense to me because my 3700X fails in AVX2 stress test at 4.3 GHz even with 1.45V. It's due to uneven quality among cores. Some are quite a bit better/worse than the others.

Despite the difference in quality among cores, mine all seem to do about the same. The good ones aren't that good and the bad ones aren't that bad.