Some Qs for the Memory timing monkeys

A place to talk about all things forever
Post Reply
User avatar
enderzero
Site Admin
Posts: 3442
Joined: Mon Mar 17, 2003 2:40 am
Location: Highland Park, Los Angeles, CA
Contact:

Some Qs for the Memory timing monkeys

Post by enderzero »

Gents, as I recently wrote in the specs thread, the new PC is all up and running. I'm re-using the same RAM that I had in the old PC, a set of 2x G.Skill Ripjaws V 16GB DDR4-3200 DIMMS running at 15-15-15-36. The AIDA64 memory benchmarks came back with some interesting results:

Image

Obviously the AMD chipset has different memory bus handling but is it pretty weird that Read speeds are identical, copy speeds are a bit better, but both write and latency are considerably worse?

Memory setting are all default in the bios. I've never been much of a memory tweaker so I'm hoping you guys might have some idea on where to start or if it's even worth it.

TIA

User avatar
R3C
Star Post
Posts: 6220
Joined: Mon Mar 24, 2003 4:59 pm
Location: Boot Sector

Re: Some Qs for the Memory timing monkeys

Post by R3C »

I'm not 100% sure about this, but I know that Intel still beats AMD in the instructions per clock on a single core. Maybe the Skylake was just able to write out more under this benchmark or something. I would guess it's the memory controller more than the memory settings, but then I also haven't touched a memory setting in years. Every motherboard I've had in the last say 5-10 years or so just detected things and ran with no issues. I also haven't overclocked anything in probably ten years.

[H]ard|OCP forums might be a good place to look for general info on this. The people that hang out in those forums are probably still the defacto tweakers. :D

User avatar
McNevin
Post Apocalyptic
Posts: 2802
Joined: Mon Mar 17, 2003 12:39 pm
Location: Lat: 47.6062095, long: -122.3320708
Contact:

Re: Some Qs for the Memory timing monkeys

Post by McNevin »

Very interesting. Unfortunately it appears it's by design.

https://community.amd.com/thread/242205

"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."

"In short, the pathway from the chiplet to the memory controller for the write data has been cut in half. This explains why it wasn’t noticeable when testing the Ryzen 9 3900X since it has two pathways to the controller, one from each chiplet so the results appeared normal. It was apparent from all the testing that this decision on the part of AMD had no noticeable effect on expected performance."
Image

User avatar
enderzero
Site Admin
Posts: 3442
Joined: Mon Mar 17, 2003 2:40 am
Location: Highland Park, Los Angeles, CA
Contact:

Re: Some Qs for the Memory timing monkeys

Post by enderzero »

Ahh that explains it. Thanks for the insight both of you. No need for tweeeeeeaking in this day and age.

User avatar
R3C
Star Post
Posts: 6220
Joined: Mon Mar 24, 2003 4:59 pm
Location: Boot Sector

Re: Some Qs for the Memory timing monkeys

Post by R3C »

enderzero wrote:
Wed Jun 17, 2020 5:16 pm
Ahh that explains it. Thanks for the insight both of you. No need for tweeeeeeaking in this day and age.
Sometimes I miss it. :lol:

But then I realize that these days, I can just click a button, my game is installed, and I can play it.

As opposed to:

Putting in a card. Realizing the jumpers are set wrong. Fight with jumpers for two hours.
Finally get the jumpers set.
Realize that something is wrong with network settings between two machines. Oh shit, the end of the 10 base 2 network isn't terminated correctly.
Cool, the machines are talking.
Oh no! Sound settings aren't working. This game says it supports General MIDI, but even though the AWE32 says it's General MIDI compliant, it isn't detected. Crap, I guess FM it is. (actually these days I prefer FM :shifty: )
Ok, sound's working. NO!!! Not enough XMS. Reboot, load a different config.sys and autoexec.bat.

***Intermission***

SHIT! Mr. Drake is coming! No!!!

"What are you two knuckleheads doing to my autoexec.bat?!?!?!?! I paid McNeal and Associates $500 to set those up!!!"

Hehehe...

***/Intermission***

Ok, I can finally load the game, hear the game, see the game. Yay!!! Set up the game options.

Oh... Great. It took so long we don't have time to play the game. See you guys later.

Ok, that's the ancient version of tweaking. The Interim Tweakage Era as it will hereby be known, was a little easier. Mostly BIOS settings, and drivers. A touch of overclocking. Tidings of good will Pentium 3 processors! That was fun, but it's still nicer to click a button and play I think.

:D

User avatar
McNevin
Post Apocalyptic
Posts: 2802
Joined: Mon Mar 17, 2003 12:39 pm
Location: Lat: 47.6062095, long: -122.3320708
Contact:

Re: Some Qs for the Memory timing monkeys

Post by McNevin »

That is an epic post to immortalize the early days of PC gaming. I especially loved the intermission! Little did Drake know he was in the presence of two autoexec Zen masters that had reached a new plane of himem enlightenment.

We sure learned a lot about the inter-workings of the x86 architecture by just wanting to play Wing Commander 2 with speech.

But yeah, kinda nice to just click and play. Sometimes I even except automatic graphics settings...
Image

User avatar
enderzero
Site Admin
Posts: 3442
Joined: Mon Mar 17, 2003 2:40 am
Location: Highland Park, Los Angeles, CA
Contact:

Re: Some Qs for the Memory timing monkeys

Post by enderzero »

So good. Just the idea of sound/sound drivers being something you ever had to think about is such a hilarious idea in this day and age. But oh my hours of time trying to get serial cable null modem LANing was like brain surgery. And don't forget life when reboots took like 2 or 3 minutes made changes considerably more trying.

Post Reply