• 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

Anyone know why there's such a dramatic reduction it write speed between X470 + R7 2700 vs X570 + R7 3700X using the same G Skill TridentZ RGB F4-3200C16-8GTZR x2

X470 + R7 2700 = 44.66GBps
cachemem R7 2700.png


X570 + R7 3700X = 25.61GBps
X570 and R7 3700X cachemem.png


it's nearly half the read speed vs only a couple of Gigabytes off on X470 anybody have any idea as to why or how to fix it or is it not worth worrying about
 
Athlonite said:
Anyone know why there's such a dramatic reduction it write speed between X470 + R7 2700 vs X570 + R7 3700X using the same G Skill TridentZ RGB F4-3200C16-8GTZR x2

X470 + R7 2700 = 44.66GBps
View attachment 181821

X570 + R7 3700X = 25.61GBps
View attachment 181822

it's nearly half the read speed vs only a couple of Gigabytes off on X470 anybody have any idea as to why or how to fix it or is it not worth worrying about
Click to expand...

4, 6 and 8 core Matisse and Vermeer CPUs have halved write speed. It's part of the design, something to do with the width of the IF bus from chiplet to I/O die. Nothing to do with the board or chipset you choose.

Ryzen 2000 and 4000 are monolithic dies, so have full speed write.

Copy speed is what's reflective of your real world performance. R/W matter little outside of benchmarking numbers.
 
Ah right o then still seems a stupid idea on AMD behalf but doesn't seem to affect perf in everyday usage but seems funny that the Vermeer based CPU above my post gets almost the same read/write/copy speeds no big drop off in write speeds there did AMD fix something so that blows your "Vermeer CPUs have halved write speed" speed theory out of the water doesn't it
 
Athlonite said:
Ah right o then still seems a stupid idea on AMD behalf but doesn't seem to affect perf in everyday usage but seems funny that the Vermeer based CPU above my post gets almost the same read/write/copy speeds no big drop off in write speeds there did AMD fix something so that blows your "Vermeer CPUs have halved write speed" speed theory out of the water doesn't it
Click to expand...

Like I said, Write is more theoretical so it was a calculated risk by AMD.

No it doesn't, because if you'd actually read my post, I said it applies to 4/6/8 core SKUs, so single CCD chips. The 5950X is not any of those things. It's not a theory if it's literally a well-documented design choice :confused:
 
I've recently got into manually tweaking my RAM. I have the Corsair CMK16GX4M2B3000C15 Vengeance LPX 16 GB RAM modules. I've currently set it to run at 3200MHz and adjusted timings according to what the Ryzen DRAM calculator suggests on the FAST preset. So far it seems OK in memory tests. I managed to run it to 1600% with 0 errors (HCI Memtest). However, I am concerned about the memory latency readings I am getting. I'm not sure if it's expected for what I have.

Here's a recent result:

ZwefQFx.png


Here's my current timing settings:

RjaglZK.png


Just want to make sure I'm not doing anything that's going to negatively affect how my RAM performs.
 
Just messing around..

4500-1900cl14.PNG
 
Kind of weird ocing memory on this platform. I added my other two sticks for 4x8 total, loosened my timings and I get stronger reads and copies. L3 bandwidth took a hit because I'm guessing it doesn't like the looser timings. Too bad my Black and Whites wont tighten up as well as my Royals.

Capture.PNG
 
Got my memory to be stable at 3200MHz after an overnight MemTest86+ run (and didn't corrupt my BIOS this time).

IZeS1g6.png
 
New CPU
54B8218B-AF35-42FC-AC8B-C8A09163F796.jpeg
 
Playing with my memory a little.. just 2 sticks

Dammit cant get passed 62ns lol :laugh:

1900CL14-14.PNG
 
Yeah, so ... , 1.4V Dram , 1.2V VSoc .
My heatsinkless ddr4's do not care much for voltage when it comes to heat output , having 16 512 MB chips per dimm.
 

Attachments

  • cachemem.png
    cachemem.png
    68.1 KB · Views: 477
2000MHz InfinityFabric Clock? .....ck me silly, I had speedier ddr4's at hand some time ago. Attached bench run was run with D.O.C.P. timings, mostly, apart from tRC reduced by one to 67 and the tRFC from auto to 448, everything else timings wise: auto(1.35V Dram voltage). All due to the new microcodes and improvement on previous ones I guess.
 

Attachments

  • Screenshot 2021-01-18 145051.png
    Screenshot 2021-01-18 145051.png
    386.1 KB · Views: 418
@Dark_ice

Different, different , so , and so forth , many differences.
 
Dark_ice said:
Old 3930K faster than Ryzen :D in DDR memory test
Click to expand...
What's with the high latency?

My X5690 and 3770K can do better in latency, but that's it.. they get creamed in everything else :laugh:
 
My nostalgic PC with its shitty 266MHz RAM.

P2IZA4Y.png


Currently looking for 4x1GB DDR-333 or better, but I don't want to spend too much on it.
 
XGP D41 3200mhz @3800mhz
Ryzen 5 3600 @4.2ghz
 

Attachments

  • Mem3800fastCPU4200.png
    Mem3800fastCPU4200.png
    165.1 KB · Views: 282
1611542236354.png

Stabilizing an Intel cache/memory OC might be one of the most tilting things in PC building...

Seems like I have this one stable after days of stress testing, final result.
 
You must log in or register to reply here.
Back
Top