• Competitor rules

    Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.

*** The Official OCUK Cinebench R20 benchmark Thread ***

Objection!

LTMatts Single thread score of 544 should be in position 17 not 15.

Your rules which you applied to my entry of 544 are that if you dont beat a score already posted you go below previous entries not above!

Consistency when applying rules is important, well unless it was a made up on the spot rule that you have since forgotten!

I have to say that I am very disappointed......




That the Beer above is not in my hand!

:D

On a more serious note it seems 544 is currently the limit for single thread with Ryzens.
We have a 3600 - 3900X and 3950X all achieving the same score.

Well unless the XT's can better it.
 
Objection!

LTMatts Single thread score of 544 should be in position 17 not 15.

Your rules which you applied to my entry of 544 are that if you dont beat a score already posted you go below previous entries not above!

Consistency when applying rules is important, well unless it was a made up on the spot rule that you have since forgotten!

I have to say that I am very disappointed......




That the Beer above is not in my hand!

:D

On a more serious note it seems 544 is currently the limit for single thread with Ryzens.
We have a 3600 - 3900X and 3950X all achieving the same score.

Well unless the XT's can better it.

I did say that didn't i? sorry, i forgot about that...
 
Objection!
On a more serious note it seems 544 is currently the limit for single thread with Ryzens.
We have a 3600 - 3900X and 3950X all achieving the same score.

Well unless the XT's can better it.

With ambient cooling if you go as far as allowing cores being disabled then 555ish is possible but very bad for the silicon. Too risky for me, the degradation from hot spotting the die with a 20watt core is very real :).
 
With ambient cooling if you go as far as allowing cores being disabled then 555ish is possible but very bad for the silicon. Too risky for me, the degradation from hot spotting the die with a 20watt core is very real :).

Absolutely. In fact, i noticed this very early on with my 3950x. MC score has always been much the same, SC can and does go up and and down depending on how the cores are set in the bios. The best SC score i have ever got resulted in a 24watt core that i will never repeat, and is the reason i have never posted it on here.
It's completely unrealistic and not a typical result at all.
 
Objection!

LTMatts Single thread score of 544 should be in position 17 not 15.

Your rules which you applied to my entry of 544 are that if you dont beat a score already posted you go below previous entries not above!

Consistency when applying rules is important, well unless it was a made up on the spot rule that you have since forgotten!

I have to say that I am very disappointed......




That the Beer above is not in my hand!

:D

On a more serious note it seems 544 is currently the limit for single thread with Ryzens.
We have a 3600 - 3900X and 3950X all achieving the same score.

Well unless the XT's can better it.
But i have 4 extra cores, so bow down to me 12 core peasant. :p
 
giphy.gif


SQUAWK !!!

:D
 
@Shac - Switched to Gigabyte X570i due to VRM temp issues on the B450i Gaming AC. Same settings...Cinebench is a little slower (25pts / 2pts) but membench is 2 seconds faster on X570 compared to B450.

uc

uc

Shame there's not a membench thread, I reckon those are some pretty damn good times :D
 
Last edited:
@humbug Could you update my single core score please? Thought I'd say hi to the Intel boys...

uc

Procedure:
1. Set cpu multi and voltages to stock, pbo disabled, SMT disabled, CPPC enabled and CPPC Preferred cores enabled. Run Cinebench R20 single core with hwinfo open - check CPPC is working as expected and make a note of the voltage your speedy core runs at (1.326v in my case) then reboot to bios.
2. Set your cpu voltage to fixed, do some fooling around in windows to make sure the droop under load hits and holds the same voltage you noted above. In my case this is LLC high with 1.375v manual to give 1.326v sustained under load. Once thats dialled in reboot to bios again...
3. Set cpu ratio to per ccd mode, then set your ccx's at or close to your normal 24/7 overclock. Boot to windows.
4. Run Cinebench R20 single core to do a final confirm that your voltage is correct and make sure you're looking at the right one - SVI2 TFN for CPU core in hwinfo.
5. Open Ryzen master, make sure only multiplier adjustment is included and start upping the multiplier on your best ccx only running cinebench each time. Dont run multi by mistake though ;). I went in .25x increments not expecting to get much but it went up 250mhz over my stable all core oc. You'll get a black screen reboot most likely when you get to the limit, so take your screenshot each time it passes.

Ryzen master screenshot below included for my settings on my highest stable run:

uc

I hope that gets some more Ryzen's up there with the Intel boys, its almost a 30 point gain for me. Obviously you can't run this 24/7 as anything more than single core load will crash it, but then I highly suspect the same goes for the rest of the scores in the top 10. This was done with no cores disabled though, so no hotspot risks and due to SMT being off the core doesn't work nearly as hard so current draw is nowhere near as high. I saw a peak wattage of around 18.5w - sure that's high but with PBO enabled I hit 18.9w on a single core running the same single core R20 test so... Still, no special cooling needed, just make use of core parking and cppc (and Ryzens natural programming to avoid sending more work to a ccx under high single thread load) so that you can run the other cores in your CCX beyond stable limits as no load will ever go there.

Good gains in membench with SMT off too...

uc

uc
 
Last edited:
@humbug Could you update my single core score please? Thought I'd say hi to the Intel boys...

uc

Procedure:
1. Set cpu multi and voltages to stock, pbo disabled, SMT disabled, CPPC enabled and CPPC Preferred cores enabled. Run Cinebench R20 single core with hwinfo open - check CPPC is working as expected and make a note of the voltage your speedy core runs at (1.326v in my case) then reboot to bios.
2. Set your cpu voltage to fixed, do some fooling around in windows to make sure the droop under load hits and holds the same voltage you noted above. In my case this is LLC high with 1.375v manual to give 1.326v sustained under load. Once thats dialled in reboot to bios again...
3. Set cpu ratio to per ccd mode, then set your ccx's at or close to your normal 24/7 overclock. Boot to windows.
4. Run Cinebench R20 single core to do a final confirm that your voltage is correct and make sure you're looking at the right one - SVI2 TFN for CPU core in hwinfo.
5. Open Ryzen master, make sure only multiplier adjustment is included and start upping the multiplier on your best ccx only running cinebench each time. Dont run multi by mistake though ;). I went in .25x increments not expecting to get much but it went up 250mhz over my stable all core oc. You'll get a black screen reboot most likely when you get to the limit, so take your screenshot each time it passes.

Ryzen master screenshot below included for my settings on my highest stable run:

uc

I hope that gets some more Ryzen's up there with the Intel boys, its almost a 30 point gain for me. Obviously you can't run this 24/7 as anything more than single core load will crash it, but then I highly suspect the same goes for the rest of the scores in the top 10. This was done with no cores disabled though, so no hotspot risks and due to SMT being off the core doesn't work nearly as hard so current draw is nowhere near as high. I saw a peak wattage of around 18.5w - sure that's high but with PBO enabled I hit 18.9w on a single core running the same single core R20 test so... Still, no special cooling needed, just make use of core parking and cppc (and Ryzens natural programming to avoid sending more work to a ccx under high single thread load) so that you can run the other cores in your CCX beyond stable limits as no load will ever go there.

Good gains in membench with SMT off too...

uc

uc

Good Grief.... nice, that's spit the Intel CPU's up a bit, if you could get that to 5Ghz you would probably end up second or third.

Updated :)
 
Good Grief.... nice, that's spit the Intel CPU's up a bit, if you could get that to 5Ghz you would probably end up second or third.

Updated :)
Thanks dude :)

First boot in the morning I could probably get another 25mhz, the high score was after a whole day benching and P95 testing so everything was hot. I've got liquid metal and a new cpu block on the way, the EK velocity just isnt cutting it and pushing 220watts on a die that small should see some real benefits from liquid metal. After all that a run at 4875mhz may be possible, that should get around 568 to 570 points depending on stability. There's pretty much nothing more left in it from voltage, from 1.272v all the way up to 1.326v only gains 25mhz...these XT's don't seem to scale much above 1.25 volts (at ambient at least). The thing is mine is bad, I only have two cores that will make box boost clocks at desktop one hits 4725 the other hits 4775mhz. I've seen XT cpus with multiple cores that will do 4825mhz out the box on the desktop, so a silicon lottery win will be able to add at least another 50mhz (5 to 6pts) onto anything I can achieve.

Nice view from up here in the top 10 though, I'm enjoying giving all the i7 peasants a royal wave :D :p
 
Back
Top Bottom