*NEW UPDATED 2* Post Your Cinebench R11.5 Score

Page 14 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

Justinus

Diamond Member
Oct 10, 2005
3,166
1,508
136
Does this mean you want me to lower your score?

Well, I suppose not. If I can only have one clockspeed/score on the board I'll need to up my game and beat my 20.39 at 4.75, huh?

The 19.73 at 4.5 is a stable 24/7 score whereas the 20.39 at 4.75 was barely benchmark stable.
 

crashtech

Lifer
Jan 4, 2013
10,521
2,111
146
Well, I suppose not. If I can only have one clockspeed/score on the board I'll need to up my game and beat my 20.39 at 4.75, huh?

The 19.73 at 4.5 is a stable 24/7 score whereas the 20.39 at 4.75 was barely benchmark stable.
Everyone has their own standard, as long as I have the screenshot, that's all the proof I require. My own personal standard is that it has to be pretty stable in Windows, even if the fans are full blast and it's running hotter than blacktop in July.
 

lopri

Elite Member
Jul 27, 2002
13,209
594
126
Ok something weird. Repeated another run @ 4.0ghz... my chip completes it faster @ 4.0ghz than 4.05ghz?

b8Eg69R.png
Maybe you've got one of the golden chips that perform better than average? XFR at work, of course.

The scores Ryzen produces in this benchmark is absurdly good. What kind of score does it get with SMT disabled?
 

IEC

Elite Member
Super Moderator
Jun 10, 2004
14,323
4,904
136
Ryzen 1700 @ 4GHz on a $79.99 B350 motherboard with $52 2x8GB DDR4-2400 CL15 @ CL14. I could probably go lower than 1.38V but haven't dialed it in yet. If you're on a budget, don't count out the R7 1700 and B350 motherboard!

Ryzen%20B350%204000MHz%201700_zpsfftsuwry.png
 

IEC

Elite Member
Super Moderator
Jun 10, 2004
14,323
4,904
136
Ryzen 1700 @ 4049MHz = 19.80 points

Yes, it really requires +70mV for that extra 49MHz versus my 4000MHz OC. 4000 is the best daily OC for my specimen. I was able to boot at 4150MHz, but not complete a CB run at voltages I would consider safe for 24/7 OC.

Cinebench%20R11.5%20%204050MHz_zpsaoauijvu.png
 

IEC

Elite Member
Super Moderator
Jun 10, 2004
14,323
4,904
136
Ran the benchmark twice to see if there was any variance. First run 27.20 pts, second run 27.09 pts. This is a stock Threadripper 1950X.



Edit: I think there might be something wrong with my setup for R11.5. It consistently slows down on the last few tiles, and I see other results online that are higher... troubleshooting now.
 
Last edited:
  • Like
Reactions: Drazick and Burpo

crashtech

Lifer
Jan 4, 2013
10,521
2,111
146
Nice, I'm actually working late right now but I'll see about getting you in the list at #1 ASAP.
 

IEC

Elite Member
Super Moderator
Jun 10, 2004
14,323
4,904
136
Quick and dirty OC to 4GHz @ 1.325V. I think I can actually use less voltage... or go higher in clocks.

30.03 points:
 
  • Like
Reactions: Drazick

rvborgh

Member
Apr 16, 2014
195
94
101
ThreadRipper seems to run very strangely with CB 11.5 vs R15... on my 48 core quad opty with all K10 cores running at 3.2 GHz... i get a 3229cb on R15... which is right around what a ThreadRipper at 4.1 GHz scores. Running 11.5 i score 39.04. The best i've seen from ThreadRipper has been the 30.03 you just posted. Something must be different about the code as i would think it should score in the high 30s. Anyone have any clue about why it scores lower? Very strange to say the least.

PS: a score of 30 is still fantastic though...
 

IEC

Elite Member
Super Moderator
Jun 10, 2004
14,323
4,904
136
ThreadRipper seems to run very strangely with CB 11.5 vs R15... on my 48 core quad opty with all K10 cores running at 3.2 GHz... i get a 3229cb on R15... which is right around what a ThreadRipper at 4.1 GHz scores. Running 11.5 i score 39.04. The best i've seen from ThreadRipper has been the 30.03 you just posted. Something must be different about the code as i would think it should score in the high 30s. Anyone have any clue about why it scores lower? Very strange to say the least.

PS: a score of 30 is still fantastic though...

I'm not sure either. It seems like once it gets to the last set of tiles it drastically slows down. If it didn't I would think it would score much higher? At the same clocks I get 3444 CB15.
 
  • Like
Reactions: Drazick

crashtech

Lifer
Jan 4, 2013
10,521
2,111
146
Something tells me it's not utilizing all the threads properly. Cinebench usually scales fairly linearly. Could be that it's time for this thread to retire, if it proves to be a limitation of the software. I haven't looked at the CB15 thread lately, but I don't believe the OP is maintaining it. This one lives on as the organized, but obsolete version.
 
  • Like
Reactions: IEC

SPBHM

Diamond Member
Sep 12, 2012
5,056
409
126
ThreadRipper seems to run very strangely with CB 11.5 vs R15... on my 48 core quad opty with all K10 cores running at 3.2 GHz... i get a 3229cb on R15... which is right around what a ThreadRipper at 4.1 GHz scores. Running 11.5 i score 39.04. The best i've seen from ThreadRipper has been the 30.03 you just posted. Something must be different about the code as i would think it should score in the high 30s. Anyone have any clue about why it scores lower? Very strange to say the least.

PS: a score of 30 is still fantastic though...

I think k10 in general is much faster in 11.5, I think it's rather the k10 score that looks low on r15, and not TR looking low on 11.5.

Something tells me it's not utilizing all the threads properly. Cinebench usually scales fairly linearly. Could be that it's time for this thread to retire, if it proves to be a limitation of the software. I haven't looked at the CB15 thread lately, but I don't believe the OP is maintaining it. This one lives on as the organized, but obsolete version.

I think it uses up to 64.
 

Justinus

Diamond Member
Oct 10, 2005
3,166
1,508
136
If it was scaling properly, the 1950X should be scoring closer to 39 than 30, though.

But it seems to scale properly on rvborgh's 48 core opteron system. 48 threads is definitely more threads than 32. It's probably not a software thread scaling problem. Seems more like it's a problem specifically with threadripper or his setup.
 
  • Like
Reactions: rvborgh

crashtech

Lifer
Jan 4, 2013
10,521
2,111
146
Seems like I recall rvborgh having to do something to get it to work properly, but I can't remember what.
 

IEC

Elite Member
Super Moderator
Jun 10, 2004
14,323
4,904
136
Seems like I recall rvborgh having to do something to get it to work properly, but I can't remember what.

I would be happy to find out. I've just about given up on CB R11.5 - too much variance. Seen runs from 13.xx to 33.xx which is absurd. Something is clearly wrong with that particular bench as the last set of tiles slows down dramatically in rendering. CB R15 at least is reproducible +/- 20 points or so.

CB R11.5 @ 4025MHz, 33.06 pts:


CB R15 @ 4025MHz, 3521 cb: