Question V-ray 6 / Corona benchmarks

[Abwx]

EDIT: Also, they seem to be doing a 10 min stress test.

This change nothing since they also do 10mn for the full 16 cores and get a score that dont match the 4C one.

I you look at their alledged 4C result and multiply by 1.25 for a 5GHz frequency, 4x for the 16 cores count and 1.32x for SMT this would hardly hit 43k points, yet their 16C result is more or less accurate, if you downscale it by the three factors above you get 47773/6.6 = 7238 pts for 4C/4T/4GHz.

Such a big discrepancy can not be hasardous, Zen 5 4C/4T actual result being 12.3% higher than their made up score.

 

[MarkPost]

Can you try with DDR5-5600 CL36 (assuming they chose sedate RAM timings)?

EDIT: Also, they seem to be doing a 10 min stress test.

Time test doesnt matter because heat isnt an issue @4.0

anyways:




BTW, I'm pretty sure Zen 4 result is also too low, but I can't check it.

I will test zen 3 and 2

 

[Hitman928]

wtf?

whats up with these pcgameshardware little boys?

This is the pefomance of a real zen5 (4 cores, 4.0GHz):

View attachment 117397

PCGH is known for having results that are favorable to Intel.

 

[Abwx]

BTW, I'm pretty sure Zen 4 result is also too low, but I can't check it.

I will test zen 3 and 2

Zen 4 should do about 6200 pts for 4C/4T/4GHz, as pointed above PCGH are known Intel + Nvidia shills with often tricked numbers.

V-Ray 6 Benchmark

benchmark.chaos.com

 

[igor_kavinski]

I will test zen 3 and 2

Forgot?

Here's mine:



We both beat PCGH

@Det0x , feel like claiming the No.1 spot in this IPC test? (4 cores at 4 GHz for 10 min)

 

[Det0x]

Forgot?

Here's mine:

View attachment 118051

We both beat PCGH

@Det0x , feel like claiming the No.1 spot in this IPC test? (4 cores at 4 GHz for 10 min)

I'm running a 9800X3D atm (rebenching 5090), but can try to put up some 4 core scores if i find the time.

Do you want with or without SMT ?
If without, why gimp the Zen5 core just to match Arrow Lake's handicap ?

 

[igor_kavinski]

Do you want with or without SMT ?

How about both?

4C4T SMT off and 4C8T SMT on.

 

[igor_kavinski]

@lightmanek , any chance you could do two cores per CCD enabled with and without SMT? Just as bonus data points out of curiosity.

 

[Det0x]

Sample size of one run

*edit*
See that my thrown together memory settings for the 4-core runs were not fully optimized, so maybe a few more points left in the tank
*edit*

4cores / 4 threads: SMT disabled


4cores / 8 threads: SMT enabled


And just for fun, a vray 5090 run

 

[igor_kavinski]

OMG!

It doesn't need SMT to get the top score and with SMT, it's basically INVINCIBLE!

 

[igor_kavinski]

And just for fun, a vray 5090 run

Thanks but it seems it is under-performing due to bad drivers: https://benchmark.chaos.com/vray/v6/vray-gpu/12?my-scores-only=false&index=1&order=desc

 

[Det0x]

Thanks but it seems it is under-performing due to bad drivers: https://benchmark.chaos.com/vray/v6/vray-gpu/12?my-scores-only=false&index=1&order=desc

Spoiler: Doing alright with the 9800X3D+5090 combo on 3dmark leaderboard atleast :P

No space for more screenshots

 

[Det0x]

OMG!

It doesn't need SMT to get the top score and with SMT, it's basically INVINCIBLE!

Yeah i found the SMT scaling really impressive.. pretty much 50%

The Zen5 core can really spread its wings when not bandwidth limited

 

[LightningZ71]

Test the scaling with all the physical cores enabled. SMT tends to thrash the L3 quite a bit when all cores are fully enabled.

 

[igor_kavinski]

Test the scaling with all the physical cores enabled. SMT tends to thrash the L3 quite a bit when all cores are fully enabled.

That's why we need 500 MB cache, 250MB on each CCD!

 

[LightningZ71]

You get a cache die, and YOU get a cache die, and YOU! Get a cache die. In fact, slap them on the IOD as well! Heck, put them on the memory sticks and SSD controllers too!

 

[Det0x]

Test the scaling with all the physical cores enabled. SMT tends to thrash the L3 quite a bit when all cores are fully enabled.

Scaling drops to ~46% but still good i would say... Per core performance actually went up

8 cores / 8 threads: SMT disabled


8 cores / 16 threads: SMT enabled


Enough of these boring 10mins runs now

 

[igor_kavinski]

Enough of these boring 10mins runs now

Yes, you have been too cooperative. Please have a cookie on our account

 

[Det0x]

Maybe i will update my score when i got the time

This was done on older lower performing agesa
View attachment 116257

V-Ray 6 Benchmark

benchmark.chaos.com

And a newer Corona 10 benchmark-run as a bonus
(this benchmark seems to scale with memory amount)
View attachment 116258

Corona 10 Benchmark

benchmark.chaos.com

Updated with some 9950X3D (ES) numbers

Corona = 18429514 points

Corona 10 Benchmark

benchmark.chaos.com

Vray = 61379 points

V-Ray 6 Benchmark

benchmark.chaos.com

Also did a run in passmark = 20463.6 points

PassMark Software - Display Baseline ID# 2479376

www.passmark.com

Saw i forgot to screenshot the memory + cpu tab, so had to rerun one more time.

 

[igor_kavinski]

Stock (no PBO anything), DDR5-7200C34 FCLK 2133 UCLK 1800 Windows Ultimate Power Plan

More benchmarks here: https://forums.anandtech.com/threads/zen-5-builders-thread.2619706/post-41432630

I have a plebian mATX mobo which doesn't have overclock failure recovery and getting it to boot after failure is unpredictable so I'm gonna use it for a while at these non-optimal settings until I am ready to mess with it more.

 

[MarkPost]

stock? 3.37ghz isnt stock, its downclocked heavily

 

[igor_kavinski]

stock? 3.37ghz isnt stock, its downclocked heavily

Maybe it's throttling badly

UAE summers are bad and it's not even the worst ambient temps yet.

 

[Sgraffite]

Ryzen 9 AI HX 370


Wasn't sure which process to lasso here

 

[MarkPost]

Maybe it's throttling badly

thats for sure. Check AIO installation, it seems to me it doesnt contact correctly with the IHS

 

[igor_kavinski]

Going to try something stupid and desperate tomorrow, without touching the insides or fiddling with BIOS options.

OK wait. Maybe I should just say what I'm going to try in case someone can successfully dissuade me from trying something really stupid.

I'm gonna put the case directly in the flow of the airconditioner's cool air. The back of the case where the backside of the mobo is, was unusually hot even after shutting down the PC.

If I put it there and keep it there a while before turning it on, what are the chances of moisture buildup on the inside?