Share your AIDA 64 cache and memory benchmark here

[Deleted member 178884]

Will they post faster than XMP without changing anything?

They won't even post XMP (hynix d-die) so, nope.

 

[Deleted member 185158]

They won't even post XMP (hynix d-die) so, nope.

I have some Hynix Corsair here. Know all about that issue haha!!! Sig rig won't post XMP either.

Keep working on them. You'll find the sweet spot in there somewhere.

could try a lower frequency at 16-16-16 and push them as fast as they'll go. Test the stability and see how the latency fairs. Might get 3600 at C16??

 

[Voltaj .45 ACP]

These are XMP stock timings - I tweak off of these even at lower frequency.
@ 1801 MHz 16-16-16-36 (CL-RCD-RP-RAS) / 52-631-469-289-10-7-44 (RC-RFC1-RFC2-RFC4-RRDL-RRDS-FAW)

Change RFC1 to 374.

Don't run silly 14-15-15 timings.
Get them straight at 14-14-14-28-40 (or less) and change command rate to 2T

Post at 3200mhz, test latency. Then work up to 3733mhz, each time run the latency test.
Don't loosen the timings if it doesn't like it past 3600mhz. Raise DRAM voltage to 1.60v. (might need this much because of the tighter timings)

Anything beyond this I will not post in open forums.



Not sure how good the IMC is on that 7980X.
Even with that D-Die, that memory will squeeze up.
Would set 1.40v at 17-17-17-38-58 is good. 2T
You could probably cut that tRFC in half at 250, but would try 350 first.
Hope this helps!!

3600 mhz 1800 fclk 14-14-14-28-40 1T 1.45v
my trfc already 350

nope.

without cpu oc this effort is meaningless.

 

[Calmmo]

Here's mine, no all core OC or pbo (as it actually degrades performance for me) and no 3800mts as my CPU's fabric doesn't appear to like that number.

 

[biffzinker]

Here's mine, no all core OC or pbo (as it actually degrades performance for me) and no 3800mts as my CPU's fabric doesn't appear to like that number.

View attachment 138694

Impressive for only being high clocked dual channel DDR4.

 

[heky]

@Calmmo

Care to share the whole primary and secondary timings? Just received my 4x8 b-die kit, so would like to have some values for orientation. Thanks.

 

[Final_Fighter]

latest beta bios. i can adjust the memory voltage up to 1.5v now where as every bios before would only go up to 1.4v. im going to try 3533 and see if i can get that stable. not really worried about lower latency so long as it doesnt increase above 65ns.

 

[biffzinker]

 

[naxneri]

 

[x58haze]

Hello to this wonderful, beautiful, amazing, outstanding community
Was wondering if you guys can give me some tips or hints, in order to drop this latency from 75 to at least 58-60 :O?

PC SPECS:
MOBO: Asrock Fatality k4 Gaming <-- this motherboard don't have Spread-spectrum, missing bunch of options so my cpu bus is lock at 99.75
Ryzen 5 1600 (summit ridge)
Ram: Patriots Viper 2x4 (8gb) 3200 (after 2 years of suffering in know what was this chipset) i had to remove the heatsink and, i found this code (K4A4G085WE-BCR ) And according to this data samsung:
And seems that my rams are E-Die, cause my ram are single rank each dim is 4gb

Data-sheet of E-die: here what amazed me is that they put as 1x4GB = K4a4g045, and mine is 085 interesting (probably a mistake)

https://static6.arrow.com/aropdfconversion/91830571053e347bfa18ac975423ff520fe5ee6d/174g_e_ddr4_samsung_spec_rev1_6_jan_17-0.pdf

Picture of my Ram Chipset.

Also tried the 1usmus, software but there is not FAST-PRESET for single Rank.
Even if i pick (calculate safe) throw error, same with the fast-preset

Also, my mobo have 4 slots rams A1 A2 B1-B2, so best settings are A2 B2 for dual channel
The thing is by using only one RAM of 4gb single rank on the A1 , i get 100% percentile score in user benchmark, but when using in dual channel percentile drop top 90%

Here is my userbenchmark run: with one Dimm

Please click the green human to continue - UserBenchmark

TDLR: I would love to drop the latency from 75 to 58-60 if that possible, also even Single Channel gave me 75ns, dual channel same 75ns, thanks for your help, xoxoxo

 

[Calmmo]

@Calmmo

Care to share the whole primary and secondary timings? Just received my 4x8 b-die kit, so would like to have some values for orientation. Thanks.

View attachment 138796View attachment 138797View attachment 138798

 

[heky]

Thanks man. Just tried a quick run with 3600-14-15-14-28-1T


Will try some higher speeds later this week...

 

[glnn_23]

I have a Ryzen 3950x in an Asus Impact and here's a run with settings 4533 17 18 17 17 34 1T

 

[Deleted member 185158]

Haven't put any Intel subs here yet so....
A quicky. Nothing special, loose at 2T.
Load temp accurate OCCT AVX2
Temp before de-lid same clock 89c
50x multi Auto voltage.

 

[E-curbi]

LOL.

 

[Deleted member 185158]

 

[Voltaj .45 ACP]

View attachment 139340

did you saw 3950x's 69k copy?how?4500 mhz ram that's it?i've made 4200 no it does not it.could be cache.copy is what we need in games not read or write.

 

[Deleted member 185158]

did you saw 3950x's 69k copy?how?4500 mhz ram that's it?i've made 4200 no it does not it.could be cache.copy is what we need in games not read or write.

The benchmark may be inconsistant to version being used. And on top of that, the benchmark isnt super consistant either.

Take these scores with grains of salt.

But yes, memory speed brings better performance.

 

[Non_NPC]

Greetings All!

My first post here, just had to join to say thank you to 1usmus and the entire community for being such a complete source of Ryzen info and experiences.

I have recently jumped onto the AMD bandwagon, and am pretty damn pleased with what I have been able to accomplish on memory clocks thus far. This is completely due to the Ryzen DRAM Calc. What an amazing program!

I hope to refine the timings even further, but this setup was Memtest 453% stable after an 8+ hour run overnight. I am very happy!

Edit: Just tried the Manual Fast timings from the DRAM Calc. Stable thus far, but will run Memtest in BIOS again overnight to check true stability. But impressive gains in AIDA thus far!

 

[E-curbi]

My first memory overclock... it is a corsair dominator platinum 3466 mhz cl16... and now it is a 4040mhz cl14.... voltage at 1.66... is that too high? can i use it like for 6 years to go???
View attachment 137032

That's a sweet 1st score...NOW you've caught the ddr4 overclocking bug, lol and it feels so nice.

Under 33ns latency.

View attachment 137976

32.8ns, that's impossible! OH, you're using a Z390 Dark board - not impossible.

Are you running that 9900K at 5.4Ghz daily? or only for benching? Is that 5.4Ghz all 8cores 16threads? OR did you find a Per Core option somewhere in the bios? I can't find a Per Core even in bios v1.07.

Come on EVGA, I like running a single core boost really high like 5.7Ghz

Have a feeling for the Z490 Dark all EVGA is going to update is the chipset and the socket. lol

5.7GHz single core per core boost on my older Maximus X Apex board (below). Just messin' around. No, it's not Prime 95 stable lol.

 

[Deleted member 185158]

5.7GHz single core per core boost on my older Maximus X Apex board (below). Just messin' around. No, it's not Prime 95 stable lol.

Is this a sustained single core boost (IE: PiMod 32m) or one of those idle speed flickers? Very curious.
edit: (really only interested in sustained for benching reasons)
edit #2: I am running Maximus X Hero btw. Not quite as great a board as the Apex....
Teach me (us)

 

[glnn_23]

Here's another 3950x Aida64 run and stability testing as well.

Also an older run trying to get the lowest latency with a 7740x in an evga dark I had.

 

[E-curbi]

Is this a sustained single core boost (IE: PiMod 32m) or one of those idle speed flickers? Very curious.
edit: (really only interested in sustained for benching reasons)
edit #2: I am running Maximus X Hero btw. Not quite as great a board as the Apex....
Teach me (us)

I set it up in the ROG bios as a Per Core single-core boost with clock speeds of 57-56-55-52-52-52, never crashed used it to get some work done for a few hours, yet no heavy CPU loads.

The 5.7Ghz boost I only played with for a few hours, and 5.8Ghz single core boost would not boot at all.

What I did end up using as a daily work profile was 5.6-5.5-5.4-5.2-5.2-5.2 (bios screenshot below)

It feels about the same responsiveness as 5.4Ghz all core all thread oc, and surprisingly uses about the same voltage. 1.360v I believe for both. Not an extremely high value since only one core at a time is boosting to 5.6Ghz.

Even if your CPU cannot hit 5.4Ghz all core/thread, you "might" still be able to run a 54-53-52-51-51-51 Per Core single core boost oc profile or some similar clock speed combination.

Just set up your ROG bios like this below.

Give it a try and have fun.

 

[Deleted member 185158]

Outstanding!!

I am very much going to fool around with this. You've given me an edge there!

Generally I do not run HT enabled past 5.3ghz all core, this helps bring TDP down a bit.
For most legacy 3D stuff I'll run 5.4ghz and as a quad, only need 2 cores essentially there.

Thanks for the pointers!
Nice rig btw, I saw it yesterday while I visited the thread linked in your sig! Impressive stuff man!!

 

[E-curbi]

Outstanding!!

I am very much going to fool around with this. You've given me an edge there!

Generally I do not run HT enabled past 5.3ghz all core, this helps bring TDP down a bit.
For most legacy 3D stuff I'll run 5.4ghz and as a quad, only need 2 cores essentially there.

Thanks for the pointers!
Nice rig btw, I saw it yesterday while I visited the thread linked in your sig! Impressive stuff man!!

Thanks, it's just an unaesthetic to borderline-ugly computer I use to get my work completed quickly. Although it is on wheels so that's kinda cool. lol

Sorry, I didn't have a level brilliant response on the overclocks, it's only a simple "Per Core" set-up in the ROG bios that works really well, extremely efficient yes less TDP less voltage and less heat output, yet nothing above and beyond what can normally be found in the bios.

If the EVGA bios had a Per Core option, this same 8086K CPU could probably boot into a fairly stable 5.8Ghz single core boost (58-57-56-53-53-53) maybe a 5.9Ghz single core boost but that would most likely be boot and screenshot ONLY.

The Z390 Dark allows 100Mhz stable improvement for all oc profiles over the Apex X, same CPU. And with ddr4 the Z390 Dark really shines far and above the Apex X.

Wish I had an Apex XI to compare and throw some values into the mix but I've never owned one. Asus only released the Apex XI to North America in October 2019 and by that time I already had (2) Z390 Dark boards and hanging on to a very sweet Maximus X Apex board that carries way too many amazing memories to let go. lol