Zen SuperPI 32M 5G/4G Challenge

[I.nfraR.ed]

So I have found the last saved screenshot from my tests on the bench ssd. According to my posts in this thread I got it even faster on the CH6, but can't find screenshots.

And that board had been dead for a long time. Still in warranty, so I should get it replaced soon and maybe try again.

It's about 10 seconds faster than the result I have on the bot: https://hwbot.org/submission/3511369_i.nfrar.ed_superpi___32m_ryzen_7_1800x_8min_28sec_857ms

Edited February 15, 2019 by I.nfraR.ed

[speed.fastest]

Wow so now Ryzen APU works on Windows 7?

[Max-Hardware_Numb3rs]

1 hour ago, I.nfraR.ed said:

So I have found the last saved screenshot from my tests on the bench ssd. According to my posts in this thread I got it even faster on the CH6, but can't find screenshots.

And that board had been dead for a long time. Still in warranty, so I should get it replaced soon and maybe try again.

It's about 10 seconds faster than the result I have on the bot: https://hwbot.org/submission/3511369_i.nfrar.ed_superpi___32m_ryzen_7_1800x_8min_28sec_857ms

 

That's very interesting, you have also 15MHz less in CPU frequency and slightly more relaxed timings than my setup.

At that time if I recall correctly the high bus was needed to push the sub timings that were blocked, now we don't have that problem anymore, I'm sure at this point that the OS optimization is the key.

 

1 hour ago, speed.fastest said:

Wow so now Ryzen APU works on Windows 7?

I never had problem with that, it's only matter to patch the installation media with the right drivers or install 7 in another machine and swap the disk

[yosarianilives]

On 2/15/2019 at 5:19 PM, Hardware_Numb3rs said:

At that time if I recall correctly the high bus was needed to push the sub timings that were blocked, now we don't have that problem anymore

There are still many timings that are hidden on both intel and AMD, so perhaps you may still be able to pick up a little extra effi by prioritizing bclk over multi

[Max-Hardware_Numb3rs]

12 hours ago, yosarianilives said:

There are still many timings that are hidden on both intel and AMD, so perhaps you may still be able to pick up a little extra effi by prioritizing bclk over multi

It's worth a try, tonight I'll give it a shot.

Still no luck with Xp I can make it to start the installation but no disks available, I have to try with others pcie cards.

[I.nfraR.ed]

2 hours ago, Hardware_Numb3rs said:

It's worth a try, tonight I'll give it a shot.

Still no luck with Xp I can make it to start the installation but no disks available, I have to try with others pcie cards.

Have you slipstreamed the needed controller drivers? For my external pci-e controllers I have to integrate the drivers in order to get the SSD detected.

[Max-Hardware_Numb3rs]

3 minutes ago, I.nfraR.ed said:

Have you slipstreamed the needed controller drivers? For my external pci-e controllers I have to integrate the drivers in order to get the SSD detected.

Yes, but now I’m working on another iso so I’ll try again with also some other drivers variants, I really hope that the effort to run Xp pays off. 

At least to get waza working properly and gain something there

[I.nfraR.ed]

26 minutes ago, Hardware_Numb3rs said:

Yes, but now I’m working on another iso so I’ll try again with also some other drivers variants, I really hope that the effort to run Xp pays off. 

At least to get waza working properly and gain something there

XP was slow for me, because of no ACPI.

[Max-Hardware_Numb3rs]

1 hour ago, I.nfraR.ed said:

XP was slow for me, because of no ACPI.

So is better to focus on W7 then, didn’t want to spend times on something useless.

thanks for the feedback! ?

[I.nfraR.ed]

23 minutes ago, Hardware_Numb3rs said:

So is better to focus on W7 then, didn’t want to spend times on something useless.

thanks for the feedback! ?

Nice, eliminated the competition, haha.

Try it, it was long time ago and I want to try it myself again.

[Max-Hardware_Numb3rs]

On 3/1/2019 at 4:45 PM, I.nfraR.ed said:

Nice, eliminated the competition, haha.

Try it, it was long time ago and I want to try it myself again.

LoL

I did some test with Intel between my actual Windows 7 build and Barbonenet Xp, I have roughly the same result (without waza), so probably I will keep optimizing 7 instead of wasting time with Xp, lately I have no time for big os experiments.

[cbjaust]

Pretty slow but it's the first time I've run this kit at CAS 12

[flanker]

I think its very good

[cbjaust]

10 hours ago, flanker said:

I think its very good

nowhere near zen and chew, also don't know the efficiency delta between the Gigabyte and Asus boards, so it's difficult to guage the score in relative terms and I don't think I'm about to buy any AM4 Gigabyte boards anytime soon

[flanker]

There is few days old result of @davestarrr Unfortunatelly he has issue to log in to forum (FB login on HWBOT mainpage)

 

 

Very solid I think...

[Max-Hardware_Numb3rs]

10 hours ago, flanker said:

There is few days old result of @davestarrr Unfortunatelly he has issue to log in to forum (FB login on HWBOT mainpage)

 

 

Very solid I think...

Yes, very nice one, if I may, with or without waza? 

[flanker]

without wazza

[KaRtA]

That looks quick Flanker. 

 

For reference, I am at sub 8m10s at 4.2g. 3500c11 memory. On W7 x64 with Waza

 

if i am find a screen, I’ll share. Still too low on time to test any further. 

Edited March 17, 2019 by KaRtA

[Guest]

Will we see a new round of tweaks?

Maybe enough to get around or past chew* and zeneffect?

Edited June 15, 2019 by AutisticChris

[Guest]

@I.nfraR.ed @Hardware_Numb3rs Do either of you have plans on testing Zen 2 and/or Windows XP for SP32M?

[Max-Hardware_Numb3rs]

28 minutes ago, AutisticChris said:

@I.nfraR.ed @Hardware_Numb3rs Do either of you have plans on testing Zen 2 and/or Windows XP for SP32M?

I'm benching SPi32M in this exact moment, with a 3600X on LN2  Win7 (that actually push almost like Xp on intel)

Xp on Zen2, maybe, if I have time

Zen2 and the fabric speed is a big obstacle for Spi, need to figure out a good balance

 

[Max-Hardware_Numb3rs]

Great, my screenshot got corrupted, but it was nothing special, 5.2GHz 2933 ram 1:1 C12.11 the score was like 6m50s or something like that at -125C

I'm with a CH7 with a bios that is horrendous and the CB on the fabric is nasty, now time to go to bed, tomorrow I'll try with more time.

It's going to be challenging, and I like it

[KaRtA]

So you think the old boards are a pain with CB more than new X570 Boards? I just picked up my 3900X, and pre-testing on my C6H. Dont want to waste LN2 if it's going to be that bad on cold.

[Max-Hardware_Numb3rs]

1 hour ago, KaRtA said:

So you think the old boards are a pain with CB more than new X570 Boards? I just picked up my 3900X, and pre-testing on my C6H. Dont want to waste LN2 if it's going to be that bad on cold.

Let's put it this way, Zen2 is a pain with CB and CBB itself, and on top with X370/X470 you have bad bioses who struggle to train memories/fabric as it should (remember to set vmem boot voltage manually)

If you want to be efficient with LN2, prepare on air various profiles starting with low fabric clock, higher the fabric clock, higher the risk of CB/CBB.

I'm trying now with a fast pot, since you have to control temp a lot, and torch a lot, at least to figure out the settings, but this is as well with X570, check the video on youtube of Steponz (bearded hardware), lots of useful info, he uses a MSI MEG, I have tried with CH6 and CH7, is worse.

In 6/7 hours I'll give you more info when I try a second run.

[Guest]

2933 cl9 should be possible, even across dual rank or maybe dual rank + 2dpc with c6h. With good mem 3600+ cl9 should be possible on ambient.

It may be worth seeing if the C7H can do c14, as that would make the fabric clock relatively close to 2933 1:1, but it may only be possible on X570 Godlike or 2 dimm boards.

Edited July 14, 2019 by AutisticChris