3800X build bad performance - what am I doing wrong?

[moproblems99]

It's a 3800X, so that's normal, only the 39xx models have improved write speeds.
He has also done ZERO tuning of the RAM.
That's said, it's hardly impressive with those kind of timings.

I thought the lower write speeds were only for single rank and two slots. I must be confused with dual ccx or something.

 

[TheLostSwede]

I thought the lower write speeds were only for single rank and two slots. I must be confused with dual ccx or something.

Yeah, you need two CCDs to get higher write speeds.
That said, he has a lot of tweaking to do, as this is what I get, admittedly at higher clocks, but with worse timings.

 

[tabascosauz]

@ecopsorn Here's a result for 3600 16-19-21 on a single chiplet CPU like your own. Your 3600 16-16-16-36 kit should have far better latency than the 70ns showing you've posted. Read, write and copy are roughly where they should be, though very slightly on the low side, considering the Royal is B-die with good timings.

3900X takes a slight hit to AIDA latency due to two chiplets, but yours is a 3800X and has no excuse. Go into BIOS and find Power Down Mode for the memory (either under the memory section or some overclocking menu) and make sure it's disabled.

From experience with AIDA64 and CB in particular, I maintain that the cause of the wild score fluctuation is due to the smorgasbord of background processes that are occasionally doing their thing during your benchmarks. Personally, I have a large Dropbox, and having the Dropbox program running in the background is the Achilles heel of CB and AIDA. It will easily add up to 1ns of latency in either membench or AIDA and a penalty of up to 100pts in CB if I have it open. And that's just 1 program.

 

[ecopsorn]

You are relying on an automatic method (DOCP) which itself is a translation of XMP, from a motherboard manufacturer for memory that it may not be able to read correctly, which means it doesn't always equate to better performance.
If all of your other issues aren't taken care of then I really wouldn't worry about scores in cinebench right now.

Sure, well I did the tests now and the plan is to move forward now tomorrow, putting the nvme in the second slot format the whole drive and disable fast boot for win install and select uefi os.

as discussed, will do testing after it and than we can discuss further wether I need a new mobo as well

 

[TheLostSwede]

Sure, well I did the tests now and the plan is to move forward now tomorrow, putting the nvme in the second slot format the whole drive and disable fast boot for win install and select uefi os.

as discussed, will do testing after it and than we can discuss further wether I need a new mobo as well

Keep in mind that if you move the SSD, you've spent that thermal pad, assuming you're using the heatsink that comes with the motherboard.
I seriously doubt that is the issue though.
Just as a silly thing, you did remove the plastic on the thermal pad before putting the heatsink on the SSD?

 

[Waldorf]

@fwiler
what i wanted to state was, that there ARE reasons for changing settings in bios before/after install,
just because you never needed to, doesn't mean there are none.
And in this case, there are some "wrong" settings in use, so changes need to be made prior clean install.


@moproblems99
as others said, anything below 3900 will have about half the write speed compared to read/copy.

@TheLostSwede
didn't have any problems moving nvme's around, no matter if the MSI boards,
or my existing GB, even with the plastic film removed.
As long as you're careful removing it, shouldn't be a problem reusing it.

 

[TheLostSwede]

@TheLostSwede
didn't have any problems moving nvme's around, no matter if the MSI boards,
or my existing GB, even with the plastic film removed.
As long as you're careful removing it, shouldn't be a problem reusing it.

But it's spent thermal interface material, it's not something you're supposed to re-use, as it's been degraded from use...
That's not the same as being able to do something.

 

[Waldorf]

lol, what?
the pad doesn't lose its performance because its been used.
it might not be sticky anymore, but thats about it.
i haven't seen any gaps when i installed drives from 5 different brands, nor did it affect temps.
i even used aftermarket pads (most MB only have beatsink/tp on first slot) which were not even sticky, and still lowered temps (vs no heatsink).
same for any other thermal pad i have seen for purchase, stating to be reusable.

 

[ecopsorn]

Just as a silly thing, you did remove the plastic on the thermal pad before putting the heatsink on the SSD?

I hope so lol, I'm not 100% sure since it is my first nvme install but I'd hit myself if I haven't - won't solve all problems though.
I'm a couple of hours away from wiping it all, waiting for the backup to be copied to my server.

wish me luck guys!

wasn't able to wipe it yet, wife hat other plans for today, but tomorrow is set.

What I did notice now, that I probably never heard the PCH fan, it is the PSU fan. It is rare that it turns on, but that thing is the loudest fan in the case when it is. Is there any way I can see that fans speed and why would that thing turn on while I'm browsing and watching youtube?

and another temp. screenshot before I go to sleep: I was watching youtube and browsing for 3h:

EDIT: sorry my bad: I was copying stuff to my server for the last 4h as well

 

[Waldorf]

it goea by temps and turns on once you reach a certain threshold.
couple people reported lower temps/less noise after removing the cover..

 

[Totally]

lol, what?
the pad doesn't lose its performance because its been used.

Issue lies with deformation not degradation. When the pad is applied it gets squished. When you peel it off, it stays squished. Do you now understand?

 

[Waldorf]

please show which post has that word, or why would i talk about degradation if someone mentioned deformation? right.

all pads i've seen/had worked on are not like paste, and removed from the board i would have guessed they were reusable aftermarket pads,
as they resemble the stuff every pc shop sells.
i have swapped drives on 3 different pads/boards (Gb and MSI), and MULTIPLE times on 1 board, NONE "squished" or unusable.
how many have you reused that were squished past use?

theory doesnt always transfer into reality.

 

[Totally]

please show which post has that word, or why would i talk about degradation if someone mentioned deformation? right.

all pads i've seen/had worked on are not like paste, and removed from the board i would have guessed they were reusable aftermarket pads,
as they resemble the stuff every pc shop sells.
i have swapped drives on 3 different pads/boards (Gb and MSI), and MULTIPLE times on 1 board, NONE "squished" or unusable.
how many have you reused that were squished past use?

theory doesnt always transfer into reality.

degrade: to lower to an inferior or less effective level

lose its performance because its been used.

"the pad doesn't degrade because its been used." i.o.w

Deformation which leads to poor contact is the problem why thermal pads generally are why not reused. I'm assuming that was implicit to what lostswede was stating.

Can't answer that I don't swap out drives unless they die on me and I haven't had one die in 6 years. As for others I almost always replace, If I can see an imprint it gets replaced.

 

[ecopsorn]

So I’m starting reinstalling now and testing.
Left the nvme in the slot 1 so far.

already a great start... windows started , first windows update cycle done, can’t press the windows button anymore - had to restart

anyway - let me report the bootup time before I run another windows update:

CSM always disabled

PBO disabled - DOCCP 3600 pro

• Fast boot disabled / UEFI OS = 23s total (19s until windows sign - 4s later in windows
• Fast boot enabled / UEFI OS = 17s

PBO disabled - No DOCCP

• Fast boot disabled / UEFI OS = 20s
• Fast boot enabled / UEFI OS = 14s

So boot times seem to be fine so far?

what is not fine is a 47degree CPU in bios

gonna disable fast boot for further installation and report benches

 

[thesmokingman]

Standard for Zen 3000 series

Not accurate, but its true for single core chiplet die, ie. 3800x and below. Move up to dual CCD and writes are doubled.

 

[ecopsorn]

Oh nice - got an 8d post code now - f me

 

[Waldorf]

if you need to redo it, install WITH csm on, leave it on.

turn docp off, but set clock/main timings and voltage for ram manually.
check what voltage is applied to SoC as well..

 

[TheLostSwede]

Oh nice - got an 8d post code now - f me

Memory parity error... That's not good news.
Did you set the Voltage to 1.35V when you put in the new modules?

 

[ecopsorn]

Cinebench Score Updates:

• 4891pts - only windows update - default nvidia - default chipset - cpu load 0% /// no PBO - no doccp
• 4928pts - only windows update - default nvidia - default chipset - cpu load 0% /// no PBO - doccp enabled
• 4910pts - only windows update - default nvidia - default chipset - cpu load 0% /// no PBO - manual 16-16-16-36 @ 1.35v
• 4942pts - only windows update - default nvidia - chipset updated - cpu load 0% /// no PBO - manual 16-16-16-36 @ 1.35v
• 4924pts - only windows update - nvidia driver updated - cpu load 0% /// no PBO - manual 16-16-16-36 @ 1.35v
• 4904pts - only windows update - nvme Samsung driver updated & AIDA64 installed - cpu load 0% /// no PBO - manual 16-16-16-36 @ 1.35v
• 4923pts - only windows update - BT+wifi driver update+samsung magician install - cpu load 0% /// no PBO - manual 16-16-16-36 @ 1.35v

Memory parity error... That's not good news.
Did you set the Voltage to 1.35V when you put in the new modules?

when doccp is disabled voltage is on auto, could that be a reason? Gonna run now all benches with doccp on and off

 

[Waldorf]

don't worry about using docp, its most likely the cause.
do manual settings matching xmp profile incl 1.35v

 

[tabascosauz]

Cinebench Score Updates:

• 4891pts - only windows update - default nvidia - default chipset - cpu load 0

when doccp is disabled voltage is on auto, could that be a reason? Gonna run now all benches with doccp on and off

See? That's much more like what you should be seeing.

8D code seems to be poorly documented, but all the instances I've seen relate to RAM stability. Can you run AIDA to see if your latency has changed now on the new installation?

 

[TheLostSwede]

when doccp is disabled voltage is on auto, could that be a reason? Gonna run now all benches with doccp on and off

Yes, very much so, auto Voltage for DDR4 is 1.2V.

 

[ecopsorn]

don't worry about using docp, its most likely the cause.
do manual settings matching xmp profile incl 1.35v

Should I start doing that already? For the sake of testing with and without i thought i do enable disable, but can certainly do
—-> scratch that - will do without doccp - results are above - next up is chipset update

- gonna provide as many data as possible during the next hours - so we can have a conclusion soon - if u guys have time tell me anything I should observe

See? That's much more like what you should be seeing.
8D code seems to be poorly documented, but all the instances I've seen relate to RAM stability. Can you run AIDA to see if your latency has changed now on the new installation?

@tabascosauz yeah the 8d was likely due to the auto 1.2v as @TheLostSwede pointed out.
here are my ram readings at 3600 16-16-16-36 and 1.35v - I still have tRC on auto (currently too high I guess at 84) - not bothering yet with ram tuning

 

[TheLostSwede]

@tabascosauz yeah the 8d was likely due to the auto 1.2v as @TheLostSwede pointed out.
here are my ram readings at 3600 16-16-16-36 and 1.35v - I still have tRC on auto (currently too high I guess at 84) - not bothering yet with ram tuning
View attachment 155460

Looking good, now we're in very sensible latency territory.

 

[Devon68]

Wait if I read this correctly. You have 2x16gb sticks of ram as in 32 gb total. Then it's normal for it to take longer to boot.