• Welcome to Overclockers Forums! Join us to reply in threads, receive reduced ads, and to customize your site experience!

Any ASUS Prime X370 Pro owners out there ... besides me??

Overclockers is supported by our readers. When you click a link to make a purchase, we may earn a commission. Learn More.
Keith Myers said:
I have LLC in Auto still. Just had my first crash with a K_Mode Exception_Not_Handled error. I remember someone posting about all the various LLC level and what they mean but have not found it again yet. Would LLC1 be advisable to settle down Vcore since I run my cores at full load all the time. I always used LLC on my FX processors, but I remember Elmor saying to just leave it at Auto so that is what I have done so far.
Click to expand...
Yeah, it causes vcore to run a smidge high, but never lower than what you set it to (approximately). LLC1 works through ambient cooling (well the limits of ambient cooling) as far as I can tell.
 
Last edited:
I had finished over an hour of P95 Blend with no errors. So thought it stable at 3200 memory. Crashed after a half hour in my distributed computing workload. Just put VDDCR_CPU back to Auto. We shall see what's what with my workload.
[Edit] Just put it into LLC1. Seems to keep the Vcore more stable as a CPU task finishes and another loads. Just as one would expect and what I have seen its effect on my FX systems.
 
Last edited:
Woomack said:
If you have problems with booting at 3200 memory clock after restart then you have to set higher DDR VTT voltage. I have no problems with 3200 memory when this voltage is higher. On air/water you don't have to touch any other voltages than CPU, DIMM and VTT. Everything else is working good at auto and is not really affecting max OC.
Click to expand...

Thanks Woomack for such a good piece of information. I am now at 3200 Mhz for my Corsair CL16 LPX memory. I bumped VTT DDR voltage from default 0.600 to 0.650V and I can boot or reboot into Windows with no issues now. Been running stable for most of the day.
 
Finally catching up with the new bios, pleased to see there is an option in it now to disable SMT. Will make some testing a LOT easier for me...
 
Corrected the VTT DDR voltage setting

Keith Myers said:
Thanks Woomack for such a good piece of information. I am now at 3200 Mhz for my Corsair CL16 LPX memory. I bumped VTT DDR voltage from default 0.600 to 0.650V and I can boot or reboot into Windows with no issues now. Been running stable for most of the day.
Click to expand...
Discovered the formula for VTT DDR. It is Vdimm / 2. So for a 1.35V DIMM voltage, the VTT DDR voltage should be set to 0.675V. The voltage sets up the DDR DIMMs for proper signal impedance matching.;)

It is supposed to be set up in Auto when you first select a DOCP setting. I don't think it worked correctly on the Prime motherboard. Setting it manually got me to 3200 on first try.
 
I've mentioned about VTT couple of times but no one was listening to me ( I'm not saying it was you ;) ).
VTT max is 1/2 of VDIMM. On my setup I had to set higher VDIMM as VTT had to be higher than 0.675V. 3200 14-14-14 is stable at 0.69V VTT in my case so memory has to work at least at 1.38V. Then I have no issues with booting at 3200 after restart or shutdown like some users are reporting ( and probably are not setting higher VTT ).
I'm not sure why but setting VTT/VDIMM higher was not helping to boot with 4 memory sticks above 2666. It was acting the same at auto or manual voltages.
 
Woomack said:
I've mentioned about VTT couple of times but no one was listening to me ( I'm not saying it was you ;) ).
VTT max is 1/2 of VDIMM. On my setup I had to set higher VDIMM as VTT had to be higher than 0.675V. 3200 14-14-14 is stable at 0.69V VTT in my case so memory has to work at least at 1.38V. Then I have no issues with booting at 3200 after restart or shutdown like some users are reporting ( and probably are not setting higher VTT ).
I'm not sure why but setting VTT/VDIMM higher was not helping to boot with 4 memory sticks above 2666. It was acting the same at auto or manual voltages.
Click to expand...
I've listened, and I haven't had any issues running 2933 on my ram since day 2 :)

The only thing I haven't tried was 1.35V w/ [email protected], but that's on me [emoji14]
 
Woomack said:
I've mentioned about VTT couple of times but no one was listening to me ( I'm not saying it was you ;) ).
VTT max is 1/2 of VDIMM. On my setup I had to set higher VDIMM as VTT had to be higher than 0.675V. 3200 14-14-14 is stable at 0.69V VTT in my case so memory has to work at least at 1.38V. Then I have no issues with booting at 3200 after restart or shutdown like some users are reporting ( and probably are not setting higher VTT ).
I'm not sure why but setting VTT/VDIMM higher was not helping to boot with 4 memory sticks above 2666. It was acting the same at auto or manual voltages.
Click to expand...

Oh, I listened and benefited immediately. But I didn't know why you said to raise VTT DDR and I hadn't a clue what it was for. I only bumped it to 0.650 initially until another overclocker in a different forum explained the technicalities of the parameter to me. I then went back in and set it to the proper 0.675V for my 1.35V Vdimm voltage.
 
Well, I tried 1.35V again with VTT set accordingly. It would seem my sticker-clad gskill ram doesn't like running at 1T with 1.35V(it's rated for 2t). Bumped it back to 1.385V w/ corresponding VTT and seems to be working fine again.
 
Command Rate depends on memory controller, not memory kit. All memory kits have XMP profile at CR2 and all Ryzen motherboards set them to CR1. There is no option to manually change it.

Yesterday I've checked that 2x16GB 2666 Crucial memory is working without issues at 2666 by only enabling XMP profile.
In the latest BIOS is also option to boost performance in some applications. Can pick Cinebench 11.5, 15 and AIDA/Geekbench. I fell like it's not working but maybe you have other experience. On CHVI it's for sure working and difference in Cine 15 is like CPU had 100-200MHz more.
 
Woomack said:
Command Rate depends on memory controller, not memory kit. All memory kits have XMP profile at CR2 and all Ryzen motherboards set them to CR1. There is no option to manually change it.
Click to expand...

Just asking, but I thought it was stick dependent as well, hence being part of the SPD profile. Or is that wrong? I really don't know enough about memory other than SPD is a thing that contains settings.
 
Woomack said:
Command Rate depends on memory controller, not memory kit. All memory kits have XMP profile at CR2 and all Ryzen motherboards set them to CR1. There is no option to manually change it.

Yesterday I've checked that 2x16GB 2666 Crucial memory is working without issues at 2666 by only enabling XMP profile.
In the latest BIOS is also option to boost performance in some applications. Can pick Cinebench 11.5, 15 and AIDA/Geekbench. I fell like it's not working but maybe you have other experience. On CHVI it's for sure working and difference in Cine 15 is like CPU had 100-200MHz more.
Click to expand...

Woomack, can you make a guess as to what those settings are changing in the chip? I saw the settings but have never investigated since I don't use or benchmark with those programs. If you were to guess.... which of those program settings would most benefit pure math calculation speed?
 
When I was comparing results with Johan45 then his results at about the same clock in Cine15 were about 50 points higher. It's the difference like to have 100-200MHz higher CPU clock. At the end when I was testing my CPU on single stage phase cooler then I could make 4.4GHz but it was still not good enough to beat 4.2GHz with that boost options and higher memory clock ( as on low temps memory controller has problems to keep higher memory clock stable ).
I have no idea if it helps in anything besides benchmarks.
 
Woomack said:
When I was comparing results with Johan45 then his results at about the same clock in Cine15 were about 50 points higher. It's the difference like to have 100-200MHz higher CPU clock. At the end when I was testing my CPU on single stage phase cooler then I could make 4.4GHz but it was still not good enough to beat 4.2GHz with that boost options and higher memory clock ( as on low temps memory controller has problems to keep higher memory clock stable ).
I have no idea if it helps in anything besides benchmarks.
Click to expand...

Thanks for the input. I figured I would just have to try it out myself and see if it makes any difference in how long it takes to crunch my CPU tasks. I will give the Cine15 setting a try in the real world. Will post my experience.

[Edit] Well I think I know what the Cine15 setting in the BIOS does. It seems to wrangle most of the work onto the first CCX module. I noticed this from starting a Prime95 run and noticing that the first 4 CPU's were more heavily loaded than the other CPU's. I think this is probably the reason for the higher Cine15 scores. Before with just the Auto setting in this option, all cores are equally loaded. Keeping work on the first CCX would reduce the amount of data that has to cross the Data Fabric to the other CCX.
 
Last edited:
Keith Myers said:
Well I think I know what the Cine15 setting in the BIOS does. It seems to wrangle most of the work onto the first CCX module. I noticed this from starting a Prime95 run and noticing that the first 4 CPU's were more heavily loaded than the other CPU's. I think this is probably the reason for the higher Cine15 scores. Before with just the Auto setting in this option, all cores are equally loaded. Keeping work on the first CCX would reduce the amount of data that has to cross the Data Fabric to the other CCX.
Click to expand...

Very interesting... but the further question is, how does it do it? Does the bios somehow redirect the work to certain cores? Does it report a different CPU structure to the OS to steer it to work differently? The thing is, what sort of improvement - percentage wise - does that setting give? At least on past testing with Intel, it didn't seem to matter much about ram bandwidth, with only a small difference from that. Enough to be of interest for competitive benching perhaps, but not for real world.

I will have to fire up my system again on the weekend and have a look myself.
 
These features are point biasing inside the CPU. Whether they deal directly with the neural network attached to the cores, or it just priority selection of certain instructions.. who knows. Something of note to take though.
 
Dolk said:
These features are point biasing inside the CPU. Whether they deal directly with the neural network attached to the cores, or it just priority selection of certain instructions.. who knows. Something of note to take though.
Click to expand...

I believe you have hit the mark with your comment, Dolk. I bet it has something to do with setting up the neural network in the chip for the expected workload.
 
You must log in or register to reply here.

Similar threads

A
ASUS Prime X370-Pro
Replies
7
Views
1K
trents
trents
D
Overclocking problem on Asus Prime X370-Pro
Replies
1
Views
2K
Mandrake4565
Mandrake4565
JrClocker
Planning New AMD Threadripper Build
2
Replies
36
Views
3K
JrClocker
JrClocker
D
Asus Prime ZA270 doesn't boot with error USB device over current status detected
Replies
2
Views
1K
trents
trents
Lochekey
Vostok V1 Thermal Interface Material
Replies
9
Views
1K
Nebulous
Nebulous
Back