FEATURED The 4X4 Summer Showdown!

[Zerileous]

Zerileous / R5 2600X / 31,611

Still can only use 4 cores.

Of course. There are still some differences between the 6c and 8c parts. Not sure if this will remain relevant to the 3000 series CPUs or not. You'll note that Johan45 has a slightly higher score from his 8c part in this test, which is exactly as I anticipated.

 

[UltraTaco]

Perhaps more cache? Or the way it's divided among corrs?
Let's nt forget also, his uncore might be set faster...and ram too

 

[Zerileous]

Perhaps more cache? Or the way it's divided among corrs?
Let's nt forget also, his uncore might be set faster...and ram too

Ding ding ding. For Zen and Zen+, the die consists of two distinct clusters of 4 cores, or core clusters (CCX), which communicate via Infinity Fabric (similar to uncore I believe). IF is also used by PCIe controller and IMC. Interestingly for Ryzens there is not a separate clock for this, it runs at the memory clock speed. Mine is faster than Johan45s. In the bios there are options for which cores are disabled. I chose a 2x2 configuration, but I have always thought a 4x0 configuration would be faster for certain tasks, as IF would be less saturated. For the 6c products, 2 cores are disabled. Here I'm making some assumptions that they are disabled in a 3x3 fashion, rather than a 4x2 fashion. If they are 4x2, then I should try running 4x0 (my motherboard has the option, but I didn't try it). I did a little more research and my initial assumption was correct: https://www.legitreviews.com/amd-ry...-symmetrical-pairs_192827#1oupIUBSKfwwKjZG.99

 

[UltraTaco]

It's time to reap the taco fire! Behold, for the numbers will scorch others!

R15View attachment 206186

R20 *note, I could not run the program in windows [strike]7,[/strike]wista so I had to use windows10.View attachment 206187

3dmark11 View attachment 206188

GeekbenchView attachment 206189

Totals
645
+
1318
+
9245
+
12,798
=
Grand total of 24,006

UltraTaco / x5670 / 24,006

:

- - - Auto-Merged Double Post - - -

Due to the age of Tacos processor, can he run a bonus benchmark of his choice to level the playing field?

 

[Alaric]

Due to the age of Tacos processor, can he run a bonus benchmark of his choice to level the playing field?

Catzilla? Run it twice, divide the score by two, subtract half and multiply times zero for your bonus points.

 

[UltraTaco]

I got confused 2/3 down the instruction set..

 

[mackerel]

mackerel/7800X/32,959

Was going to use 6700k, but then wondering if Skylake-X architecture helps here...

 

[Alaric]

R20 *note, I could not run the program in windows 7, so I had to use windows10.

Is this a thing? R20 won't run on Windows 7? That throws a monkey wrench in the works.

 

[UltraTaco]

I made a mistake, I meant vista. Windows vista64

 

[Alaric]

Whew! That was close. LOL

 

[Johan45]

Johan45/ R7 2700X/ 33,769

 

[UltraTaco]

Bump.

New members and existing unparticipated ones. Please participate! Taco urges you to bench. This harvest season brings very few result crops. We need more!>>>>

 

[devlos]

devlos \ 4690k \ 25,416









8351 + 1434 + 609 + 15022 = 24,416

 

[SwartHack]

How do I play this game?
Hey everyone, been a minute!



Cheers!




Removed Inappropriate Image - Janus

 

[UltraTaco]

Sir, you run benchmarks that Johan posted and you take screenshot of the results along with Cpuz and hwinfo. This ones a handful a little because you have to run multiple benchmarks. Calculator may be required for results addition unless you're good with math.

Once you have the results, you post them in the format mentioned in the first post.(along with screenshots of course) See additional member posts for help. most important is to have included photo as wallpaper, otherwise your score may not be counted.

 

[SwartHack]

thanks!
https://hwbot.org/user/swarthack/
been a minute Taco

- - - Auto-Merged Double Post - - -

Super awesome I have my old title. Missed all you geeks!

- - - Auto-Merged Double Post - - -

I am still the funniest **** I ever knew

 

[mackerel]

mackerel/3700X/36,033

Looks like I'm throwing in the first Zen 2 submission. I tried both 4+0 and 2+2 core configurations across the CCX. It was interesting which was faster... it will be interesting to see if others observe similar.

3DMark11 physics score: 4+0 7.7% faster
Cinebench R15: 2+2 0.8% faster
Cinebench R20: 4+0 0.3% faster
Geekbench 4 multicore: 2+2 0.8% faster

In Cinebench I suspect each thread can work independently of each other so it doesn't matter much each way. As it isn't ram bandwidth sensitive either, that may explain why the core configurations don't make much difference. 3DMark11, if the physics simulation threads need to talk to each other, having them on the same CCX may explain the improvement there. I ran 2+2 twice but it remained lower. GB4 I'm just not familiar with and don't really care about, but it didn't seem to strongly prefer either.

If I'm not mistaken, the recent Windows scheduler change was to use cores on one CCX first. If this CPU was running all cores enabled, it would more likely hit the 4+0 scenario first and may give a bit of a boost in tasks that behave similarly.

 

[Zerileous]

Thanks for testing that out mackerel. I'm going to have to ensure I get the newest windows update, update chipset drivers, and give it another try to see if it makes a difference for me.

 

[Johan45]

Johan45/ R5 3600X/ AIO/ 37,253

1081
2478
12453
21241
37,253
R15 1081




R20 2478



3d 11 12,453



GB4 21,241

 

[UltraTaco]

Man, those scores are inching closer and closer, that's okay though, I love competition. X58 FTW! Cnt beat!
Good to see some action again in this thread