Share your AIDA 64 cache and memory benchmark here

[FilipM]

I run

VDDP 900
VDDG CCD 940
VDDG IOD 1060

 

[freeagent]

Thanks guys! No not stable.. things are a little too tight, but its benchable! 900 is a bit too low for occt, 950 is ok most of the time I think.. might have had a hiccup, I gave it a little more "just in case" 1 more tick and the indicator go's from white to yellow

The reality is I am still learning my system, and I don't know all the ins and outs as intrinsically as some others, but I am getting there slowly but surely. There is a lot of shit to know and remember Very different from Z77.. The transition from ddr3 to ddr4 has been smoother than expected. And when I was on Intel I did not even remotely delve as deep as I do here, its crazy shit man.

 

[FilipM]

The CPUs come almost maxed out nowadays, the RAM is the voodoo magic now

 

[oobymach]

Here's where I'm at now tested for stability with p95 for a few mins and a quick run of testmem5. I got bsod and a couple other issues when trying with 670trfc and 4 scl. Also running @ 1.35v
The kit I bought is G.Skill F4-4000C18D-32GVK

 

[tabascosauz]

Now were talking..

View attachment 194135

You win lol. I got nothing. I knew you guys don't play fair by benching in safemode but right now I can't quite get back into my right on the dot 50.0ns profile to see it break 50 in safemode. Hardware probably got sick of my shit lmao I havent reflashed BIOS in a long while so this one is definitely feeling used and abused after all the VTTDDR bullshit

It looks like you've both got at least about 1-2ns advantage on me Zen2 to Zen3 alone. Can't compete. Maybe if I get my hands on a Cezanne APU, but I'd still get nowhere without A2 B-die and I just don't know about the extra CPU and RAM expenditure on this dead platform.

 

[FilipM]

The reason I am benching in Safe mode is that my Windows is a bit bloated and gives inconsistent results without killing services. Like Aura Sync takes 1ns.

Safe mode gives much more consistent results so you can see the little improvements that you make here and there.

I ran Samsung C die at 3733 both on a 3600x and a 5600x. Zen 3 was roughly 6ns faster at those settings which i did a copy-paste of when i installed the new CPU.

 

[freeagent]

You win lol. I got nothing. I knew you guys don't play fair by benching in safemode but right now I can't quite get back into my right on the dot 50.0ns profile to see it break 50 in safemode. Hardware probably got sick of my shit lmao I havent reflashed BIOS in a long while so this one is definitely feeling used and abused after all the VTTDDR bullshit

It looks like you've both got at least about 1-2ns advantage on me Zen2 to Zen3 alone. Can't compete. Maybe if I get my hands on a Cezanne APU, but I'd still get nowhere without A2 B-die and I just don't know about the extra CPU and RAM expenditure on this dead platform.

I didn't even know you could do that stuff in safe mode lol.. Not until the last page where Filip spilled the beans.. and I was like heeyy.. what the heck..

I'm sorry man! I was just playing.. if it helps its not stable!

 

[FilipM]

Sorrryyyy... Well not really, it got things going

 

[Taraquin]

Seems frequency affects latency the most. I got 54.8ns with 4.8GHz and both 3800cl16-17 1t gdm on tRFC 260 @ 1.45V and 3800cl15-16 gdm off 2t tRFC 245 @ 1.5V. The last setting is gamestable but, spits errors after a few mins in karhu due to overheat.

 

[FilipM]

I had this stable at 3800 - 25 Cycles TM5

Only ran once, 1.5 Vdimn

 

[freeagent]

Well.. I figured out how to get TM5 to run more than a few minutes.. turns out my 2000 14-15-15 wasn't stable even with 1.55. Ahh well.

Edit:

It was but had interconnect errors.

 

[FilipM]

Copy mine, give it a shot

Edit. You edit after me

 

[Taraquin]

What to do when your 5600X is gold and your B-die is crap

I had this stable at 3800 - 25 Cycles TM5

Only ran once, 1.5 Vdimn

View attachment 194204

You have a very good bin. Mine is 4400cl19 and only timings it shines at are tRC and tRFC while primaries suck big time, haha.

 

[FilipM]

If its really golden

PBO 200MHz offset
-30 negative curve
+20mV offset

Set appropriate limits

EDC ~110
TDC ~75
PPT ~110

Enjoy 4.85 with peak voltage around 1.35V, in nearly every scenario, except R20 where depending on cooling, it will sit at 4.7 to 4.75 at 1.25V

But that's for another thread/topic

Disclaimer: May need tweaking

 

[Taraquin]

If its really golden

PBO 200MHz offset
-30 negative curve
+20mV offset

Set appropriate limits

EDC ~110
TDC ~75
PPT ~110

Enjoy 4.85 with peak voltage around 1.35V, in nearly every scenario, except R20 where depending on cooling, it will sit at 4.7 to 4.75 at 1.25V

But that's for another thread/topic

Disclaimer: May need tweaking

It does all core 4.8 1.34 cb20 stable, but get to hot for cooler, 1.32V is non avx stable. 4.7@1.23V is cb20 stable. I know there are even better samples out there, but I'm satisfied. Gigabyte removed the 1.2.0.1 bios before I had the chance to download it so I'm stuck at 1.1.0.0 without curve optimizer for now :/

 

[freeagent]

My FIT is around 1.25v, that's as high as I will go for hardcore avx stuff and is good for 4600. After that I use 1.3875 to bench @ 4800 and is good for 3D and some crunching. I haven't messed with curves or anything like that. Though I should.. it looks interesting. Maybe later.

 

[Taraquin]

My FIT is around 1.25v, that's as high as I will go for hardcore avx stuff and is good for 4600. After that I use 1.3875 to bench @ 4800 and is good for 3D and some crunching. I haven't messed with curves or anything like that. Though I should.. it looks interesting. Maybe later.

Seems the binningprosess is quite loose at AMD. I read about 5600X struggling with 4.6@1.35V and some who run 4.6@1.1V. The first should have been a 5600 vanilla, while the latter is sort og XT-territory.

 

[freeagent]

One thing I found over the years is that chips that react with voltage, and scale well run hotter than the chips that seem to suck up juice..

 

[Taraquin]

One thing I found over the years is that chips that react with voltage, and scale well run hotter than the chips that seem to suck up juice..

Yeah, you are probably right. My 3600 was a terrible bin which maxed out at 4.1@1.24V, and was unable to run 4.2 stock. Allcore it ran 3.9 stock using 90W, while in most reviews 4.2@stock was no problem. But it was really cool at same consumption as my good binned 5600X.

 

[Arctucas]

You win lol. I got nothing. I knew you guys don't play fair by benching in safemode <SNIP>

I also was not aware one could run the benchmark in Safemode, so, of course, I had to try it.

But, not a lot of difference.

 

[Taraquin]

Yes, betabios with 1.2.0.1 out. 3866/1933 stable, 53.8ns aida. Testing 3933/1966 now.

 

[freeagent]

I run

VDDP 900
VDDG CCD 940
VDDG IOD 1060

Wasn't there a rule or something where you should run VDDG IOD +-.05v of VDDP or something?

Or was the something to do with VTTDDR?

 

[tabascosauz]

Wasn't there a rule or something where you should run VDDG IOD +-.05v of VDDP or something?

Or was the something to do with VTTDDR?

No, it applies to VDDP and both VDDGs, but only in relation to VSOC. iirc the minor voltages are split off of VSOC, so taking into account droop and whatnot they cannot exceed a certain value in relation to VSOC. -0.04V to 0.05V depending on who you ask, but it depends on VSOC LLC on your system. Otherwise there's a risk that the board will ignore what you enter and do whatever the F it wants, because there's simply no way the board will allow you to get more on a CLDO voltage than is physically possible out of VSOC.

VTTDDR is different, don't mess with it unless you're benching some seriously high freq or VDIMM

 

[freeagent]

Thank you sir, and duly noted :salute:

 

[Taraquin]

4000\2000 stable, aida ran in safemode with 4.8GHz. Terrible binning on my B-die

Thank you sir, and duly noted :salute:

View attachment 194250

Wow, that IO-die voltage. I need 1.07V for 2000 IF. You won the IO-die lottery! You should try 16 tFAW, tRDWR is also a bit loose, tRFC should do 280-290, but tCWL is very good and scl's are good.