• Welcome to TechPowerUp Forums, Guest! Please check out our forum guidelines for info related to our community.

Share your AIDA 64 cache and memory benchmark here

Can someone answer me a question here I have a R7 2700 and it's write speed for memory seems to almost double what I'm seeing with R9 3xxx CPU's take the above screen shots of Zach_01 and Voltaj .45 ACP they both have near half the Write speed (29xxxMB/s) compared to their Read speed (5xxxxMB/s) despite them both running higher speed kits where as I'm seeing 40000+MB/s read and 40000+MB/s write as seen below

cachemem R7 2700.png
 
Last edited:
Can someone answer me a question here I have a R7 2700 and it's write speed for memory seems to almost double what I'm seeing with R9 3xxx CPU's take the above screen shots of Zach_01 and Voltaj .45 ACP they both have near half the Read speed (5xxxxMB/s) compared to their Write speed (29xxxMB/s) despite them both running higher speed kits where as I'm seeing 40000+MB/s read and 40000+MB/s write as seen below
View attachment 130967
From AMD: “This is an expected result. Client workloads do very little pure writing, so the CCD/IOD link is 32B/cycle while reading and 16B/cycle for writing. This allowed us to save power and area inside the package to spend on other, more beneficial areas for tangible performance benefits.”

s_vayner said:
In short, the pathway from the chiplet to the memory controller for the write data has been cut in half.

 
Last edited:
From AMD: “This is an expected result. Client workloads do very little pure writing, so the CCD/IOD link is 32B/cycle while reading and 16B/cycle for writing. This allowed us to save power and area inside the package to spend on other, more beneficial areas for tangible performance benefits.”


Ah I see well then that's alright then I just thought it a bit weird was all but makes sense then since apps and games do sod all writing to use the area for other better perf bits
 
Ah I see well then that's alright then I just thought it a bit weird was all but makes sense then since apps and games do sod all writing to use the area for other better perf bits
copy is important not write.if one chip in 3xxx series half write speed but two chip like 3900x is normal write speed.

my 2700x result:
131008
 
The latest stable...
DRAM 1900MHz, 1.44v with 1:1:1 mode

AMD RYZEN MASTER 12-Sep-19 12_59_21.png
cachemem_12.png
 
@Zach_01 what version of ryzen master is that
 
Can you help me to improve this?? TRIDENTZ 3200 14 to 4000 cl16 8GBx2

At this moment > DRAM voltage@1.45v VCCIO VCSSA @ 1.23
NO HT in CPU as I use PC only for gaming

RAM TEST + cache + FPU not showing errors , cinebench and realbench stable atm.


131807
 
Last edited:
So, turns out manually letting the board assign subtimings etc worked better than my Z270X Gaming 7 / X299 Gaming Carbon did, Asus power?
Well I'm doing 2933mhz @ 14-16-16-31 CR2 with 1.3V and everything appears to be stable! :)
131823
 
Can you help me to improve this?? TRIDENTZ 3200 14 to 4000 cl16 8GBx2

At this moment > DRAM voltage@1.45v VCCIO VCSSA @ 1.23
NO HT in CPU as I use PC only for gaming

RAM TEST + cache + FPU not showing errors , cinebench and realbench stable atm.


Try 1T
Try to max your tREFI
Try fTFAW = 16
lower tRFC

These are my results with a 3600C15 kit

http://71.82.154.27/BinaryLore/data/uploads/hardware/gentoo64/aida64_9900k_50x50x47x-1.25v-4000_16-17-17-36-1T-1.425v_tRFC288_SA1.15v.png
http://71.82.154.27/BinaryLore/data/uploads/hardware/gentoo64/asrock_tc_9900k_50x50x47x-1.25v-4000_16-17-17-36-1t-1.425v_trfc288_sa1.15v.png
 
.
 

Attachments

  • Aida64 @ fast mem timings @ Default CPU Bios 1.20.png
    Aida64 @ fast mem timings @ Default CPU Bios 1.20.png
    87.3 KB · Views: 1,005
9900k.png

This my new setup .
 
That is very nice!
Can you give us some specifics of your setup?
Hey. It's very nice, but I cannot replicate it no matter what I do. :D


This is the setup. What happened is, I installed a new bios 1.20, set the memory timings 16-16-16-32 as per dram calculator @ 3600 and left everything on auto. And then, like magic the windoiws task manager sees my CPU @ 4 GHz as base and 4.52 GHz as turbo speed (check first attachement); the HWiNFO however showed 4.1 GHz max boost. The scores in CB15 incresed significantly around 1720 (see second attachement). I restarted the system, saved the configuration in BIOS and rebooted. And the boost was lost. I tried to replicate it again with no success.

Normally I can run stable 4.25 GHz @ 1.35 ........ but I settled for 4.1 @ 1.225 at least until ABBA stable bios arrives.
 

Attachments

  • Capture1.PNG
    Capture1.PNG
    81.6 KB · Views: 555
  • Capture2.PNG
    Capture2.PNG
    8.1 KB · Views: 539
Last edited:
DRAM - 1.45v
mine 3800cl16.PNG
 
Back
Top