ROG Crosshair VI

[BenchBrothers.de]

Elmor, may I ask how accurate the measuring-points are? When Setting 1.325v in bios with llc3 cpu-z is reporting 1.33v to 1.35v loaded while measuring vcore with dmm (vcore and gnd pad) reports a continuous 1.4v.

 

Another Ryzen question (maybe noob-question): Is reference clock = pcie-clock? Or do I misunderstand something?!

[ReDXfiRe]

First of all; won't be replying specific questions about Ryzen 5 performance but its already known some units are available due to mistake of providers. I already OCed this CPU but during stress test (I turned it off after coming back and realized after 2 hours there were those spikes in voltage as it shows in the image).

 

Hopefully I can get some help why that happened and if its safe. I'm using BIOS 1001 and I got my RAM working with the OC up to 3200mhz which is rated (Vengeance LED 3200mhz CL16). I can confirm my RAM worked with 1001 (tried everything with 902 and did not work so I ventured to try 1001 and it worked on first try).

 

LLC is on 3 with Vcore Set on BIOS at 1.43. Even at 1.43 theres some jumps to 1.461v as you see; but those voltage anomalies worries me up a little bit and that CPU temp (temps were 78-80 degrees on load on stress test) but I suppose it went up to 89 degrees during that voltage spike.

 

Hope I don't stir drama which is not my intention but actually finding why the surge. I've seen in other posts some people getting jumping from 1.45 to 1.7 spikes but nothing like this.

 

The stress test I used is X264 stress test infinite loop 16 and normal.

 

If the mods feel like removing the picture I understand. Still I'm at your disposal and hoping I get some insight. Sorry if it causes any trouble and if a mod thinks this post should be removed I understand. Thanks.

[matose]

The best LLC setting for me is Level 1: 1.35v set in BIOS you have 1.35v in IDLE and 1.37v in LOAD. BCLK is linked to PCI-Express clock

 

Elmor, any change in seeing OC Panel 1 firmware soon?

[Masterchief79]

Another Ryzen question (maybe noob-question): Is reference clock = pcie-clock? Or do I misunderstand something?!

 

ASUS ROG Crosshair VI Hero Extreme Overclocking Guide (Updated to v0.5) - Overclocking.Guide

 

PCI-E Clock is only the same as REFCLK for up to 104.85MHz I believe. After that, the REFCLK gets decoupled from the rest of the bus so that you can run 140MHZ+ REF and still have your USB ports, graphics card and ssds working (which get their 100MHz freq from another source then). Hope I explained this correctly.

[elmor]

New BIOS 1002

 

Fixes the TurboV Core voltage issue. Mostly just bugfixes in this one.

 

Elmor, may I ask how accurate the measuring-points are? When Setting 1.325v in bios with llc3 cpu-z is reporting 1.33v to 1.35v loaded while measuring vcore with dmm (vcore and gnd pad) reports a continuous 1.4v.

 

Another Ryzen question (maybe noob-question): Is reference clock = pcie-clock? Or do I misunderstand something?!

 

See here regarding DMM measurement under load: ROG Crosshair VI overclocking thread - Page 401

 

For accurate readings try the latest HWInfo64 which has values directly from the VRM controller.

Edited March 24, 2017 by elmor

[Alex@ro]

ASUS ROG Crosshair VI Hero Extreme Overclocking Guide (Updated to v0.5) - Overclocking.Guide

 

PCI-E Clock is only the same as REFCLK for up to 104.85MHz I believe. After that, the REFCLK gets decoupled from the rest of the bus so that you can run 140MHZ+ REF and still have your USB ports, graphics card and ssds working (which get their 100MHz freq from another source then). Hope I explained this correctly.

 

That is wrong,till 104.85 you have pci-e 3.0,then the pci-e drops to 2.0 but the pci-e clocks is still synced to REFCLK.

 

Finishing my tests with Crosshair VI,limit here is 152 refclock(which is suicidal for some benches) however 150.77 could run realbench no problems.

 

Also for @chew,VDDSOC might help when on on the edge of stability,for example in my particular case i did memtest DDR4-3550 14-15-15 which was attainable only after raising VDDSOC to 1.25,1.15 was instant fail and 1.20 failed like 10% progress. I don't know how safe this is,however imc showed no sign or degradation whatsoever after this session.

 

 

[BenchBrothers.de]

See here regarding DMM measurement under load: ROG Crosshair VI overclocking thread - Page 401

 

For accurate readings try the latest HWInfo64 which has values directly from the VRM controller.

 

Thank you, missed that post.

 

 

 

Hynix ... Dual-rank ... What do you expect when we don't even have subtiming adjustments?

 

Just one last question: Do you have access to subtimings like the way you have for command rate (ie: you can set a fixed value in a specific bios version)? Or don't you have any access?

[Masterchief79]

The subtimings depend on the memory ratio you set. With 2933MHz you get trc 69 and with 3200MHz trc is 74. Other than that, there is no way to control them at the moment.

[BenchBrothers.de]

The subtimings depend on the memory ratio you set. With 2933MHz you get trc 69 and with 3200MHz trc is 74. Other than that, there is no way to control them at the moment.

 

I know that.

 

My question is whether ASUS (or other vendors) can control them with bios updates or not. They already do it with command rate so it seems that there is some sort of control. But every change would need a bios flash.

[cdawall]

That is wrong,till 104.85 you have pci-e 3.0,then the pci-e drops to 2.0 but the pci-e clocks is still synced to REFCLK.

 

Finishing my tests with Crosshair VI,limit here is 152 refclock(which is suicidal for some benches) however 150.77 could run realbench no problems.

 

Also for @chew,VDDSOC might help when on on the edge of stability,for example in my particular case i did memtest DDR4-3550 14-15-15 which was attainable only after raising VDDSOC to 1.25,1.15 was instant fail and 1.20 failed like 10% progress. I don't know how safe this is,however imc showed no sign or degradation whatsoever after this session.

 

https://abload.de/img/snaphsot0040a7szh.png

 

https://abload.de/img/snaphsot0011acs0k.png

 

I have been pushing 1.3v through the VDDSOC for the past week or so to keep >3600 stable on my -b's. I have tested as high as 3800 with 0702, I was unable to keep those speeds with 1001/0038. I have not seen any noticeable degradation of the IMC. The chip itself seems to be good for ram, the other one I was testing with wouldn't push the same ram over 3300 regardless of voltages.

[CCarpenter]

New BIOS 1002

 

Fixes the TurboV Core voltage issue. Mostly just bugfixes in this one.

 

 

 

See here regarding DMM measurement under load: ROG Crosshair VI overclocking thread - Page 401

 

For accurate readings try the latest HWInfo64 which has values directly from the VRM controller.

 

 

Thank you for the new bios 1002.

 

With your Beta-BIOS since 1001 my Memory works fine @2666Mhz:

 

Corsair Vengeance 32GB Kit DDR4-2666 CL16 (CMU32GX4M4A2666C16) LED.

 

I have still Problems with the DisplayPort on my MSI Nvidia GTX1070. The Asus Display wakes up and shows no Signal on Displayport. HDMI works fine.

 

Sometimes with this config @Default the System just does not boot (error 0d)

Next time it boots without Problems ... same config.

 

Here is my whole Hardwareconfig:

System: Host: ccsbox Kernel: 4.10.4-towo.1-siduction-amd64 x86_64 (64 bit gcc - Pastebin.com

 

Some clues?

 

Greets CC

[Johan45]

Not a criticism, just an observation. When the 0902 BIOS was released something between it and 5803 really trashed the wPrime and SPi scores. At least for me it did. I have tried everything I can think of but can't get close to the scores I got before. A full second difference in wPrime 32m is a lot and about a minute in SPi32m

[matose]

Not a criticism, just an observation. When the 0902 BIOS was released something between it and 5803 really trashed the wPrime and SPi scores. At least for me it did. I have tried everything I can think of but can't get close to the scores I got before. A full second difference in wPrime 32m is a lot and about a minute in SPi32m

 

Multiple restarts usually solves the SPI performance in my case

[Johan45]

I'll give that a try and see if it helps at all

[Achill3uS]

hi all,

just a little feedback about the crosshair iv... ive got the board last week, updated to bios 1001 immediately and did a couple days of pretest before ln2... i have 2pcs 1800X the first one was able to do 4ghz cinebench at 1.40v aircooled, the second one 4ghz cinebench 1.28v... the second one has also better imc, ddr4-3200c12 was easy but 3400~3600 no go. cinebench efficiency is great by the way its boosting a lot also with more memspeed and the performance bias option, very cool feature!

ive reached 5.32Ghz and 2392 cinebench R15 score during the first ln2 testing so im really happy about it, but also wants to share some of the most important issues ive been facing with and still looking for solution...

 

i know that elmor's guide is just reference but for me several things not working...

issues on air:

-refclock is just not posting above 110mhz for me, not even if i lower cpu below 4ghz and dram to 2400~2600mhz...

my voltages on air: cpu 1.3~1.4v, soc voltage 1.20, pll 2.0, sb 1.30, dram as much it needs 1.35~1.80

issues on ln2:

-refclock still not booting above 110mhz, dram also not booting above ~3000mhz at any cas latency and its also not easy to reach sometimes ive got stuck at postcode 55 need to retry/reset to get into windows. but im way far about booting that 28x141 good start point with ln2 what is in the guide. is really that tricky to reach high refclock?

-once im trying to test some memory setting which is ending up into postcode 9F or 0d sometimes it takes 10~20min to get back to the bios, even cmos clear(push for 3+ secs), safeboot numerous retrys getting into a postcode loop with 8, 0d or 62 sometimes.

-3d testing is killing me. on air i can run firestrike/fse no problems at 4.2GHz all cores on cpu, 3200c12 on memory without any problems with aircooled 1080ti at stock clocks. on ln2 ive got pcie issue, blackscreen and display driver stops working on firestrike run during combined test at even -50C and 4GHz. if i run combined test separately sometimes it passes, but during full FS run its stopping all the time there after a couple sconds. ive tried pcie x16 and x8 as well, pcie gen3, gen2, gen1 (gen1 loosing a lot in graphics score) and the issue still remains from 100mhz refclock to 110 as well. on fullpot its got worse, sometimes display driver stopping during the first graphics test in FS.

my voltages on ln2: cpu 1.7~1.80vcore, soc voltage: 1.25, pll 2.2v (above 2.5v os even not booting), sb 1.30, dram 1.35~1.80

 

if anyone have useful hints what can i improve or try with my system i appreciate any kind of help. thanks.

cheers

bala

[I.nfraR.ed]

Let me guess - you're benching 2D with high-end current-gen VGA. Try another VGA that can handle increased PCI-e clocks.

[BenchBrothers.de]

Let me guess - you're benching 2D with high-end current-gen VGA. Try another VGA that can handle increased PCI-e clocks.

 

Running two 1080 Ti in SLI with high(er) PCIe-clock is no Problem for me, at least with watercooling. Tested 125 mhz (3.0) and 146 mhz (1.1/2.0). Voltages at "stock" (1.325 cpu, 1.0 soc, 1.8 pll, 1.05 sb, 1.285 dram).

 

With one older 7970 dcuii I could run up to 151 mhz without issues.

 

But all cards have in common that they run at stock speeds.

[I.nfraR.ed]

Then maybe something else is "failing". I see a lot of variance between different chips. Not that I've tested many, but over the net.

If it is some SATA device, then it would just fail and not be recognized rather than preventing the board to POST, am I right?

[Achill3uS]

Let me guess - you're benching 2D with high-end current-gen VGA. Try another VGA that can handle increased PCI-e clocks.

 

so if youre going to buy 1800X + 1080Ti forget about the refclock oc? sorry i must be missed it to be mentioned anywhere... and how does it solve the 3d issues? do not bench with 1080Ti??

 

Then maybe something else is "failing". I see a lot of variance between different chips. Not that I've tested many, but over the net.

 

exactly thats why im collecting any useful ideas.

[Johan45]

Multiple restarts usually solves the SPI performance in my case

 

Thanks, that did help my SPI but wPrime 32m is still slower. I also disabled any performance bias.

[cdawall]

Let me guess - you're benching 2D with high-end current-gen VGA. Try another VGA that can handle increased PCI-e clocks.

 

That really depends on card. I did all of my testing with a 1070. BCLK was fine all the way up to 148, 2D/3D benchmarks.

[elmor]

hi all,

just a little feedback about the crosshair iv... ive got the board last week, updated to bios 1001 immediately and did a couple days of pretest before ln2... i have 2pcs 1800X the first one was able to do 4ghz cinebench at 1.40v aircooled, the second one 4ghz cinebench 1.28v... the second one has also better imc, ddr4-3200c12 was easy but 3400~3600 no go. cinebench efficiency is great by the way its boosting a lot also with more memspeed and the performance bias option, very cool feature!

ive reached 5.32Ghz and 2392 cinebench R15 score during the first ln2 testing so im really happy about it, but also wants to share some of the most important issues ive been facing with and still looking for solution...

 

i know that elmor's guide is just reference but for me several things not working...

issues on air:

-refclock is just not posting above 110mhz for me, not even if i lower cpu below 4ghz and dram to 2400~2600mhz...

my voltages on air: cpu 1.3~1.4v, soc voltage 1.20, pll 2.0, sb 1.30, dram as much it needs 1.35~1.80

issues on ln2:

-refclock still not booting above 110mhz, dram also not booting above ~3000mhz at any cas latency and its also not easy to reach sometimes ive got stuck at postcode 55 need to retry/reset to get into windows. but im way far about booting that 28x141 good start point with ln2 what is in the guide. is really that tricky to reach high refclock?

-once im trying to test some memory setting which is ending up into postcode 9F or 0d sometimes it takes 10~20min to get back to the bios, even cmos clear(push for 3+ secs), safeboot numerous retrys getting into a postcode loop with 8, 0d or 62 sometimes.

-3d testing is killing me. on air i can run firestrike/fse no problems at 4.2GHz all cores on cpu, 3200c12 on memory without any problems with aircooled 1080ti at stock clocks. on ln2 ive got pcie issue, blackscreen and display driver stops working on firestrike run during combined test at even -50C and 4GHz. if i run combined test separately sometimes it passes, but during full FS run its stopping all the time there after a couple sconds. ive tried pcie x16 and x8 as well, pcie gen3, gen2, gen1 (gen1 loosing a lot in graphics score) and the issue still remains from 100mhz refclock to 110 as well. on fullpot its got worse, sometimes display driver stopping during the first graphics test in FS.

my voltages on ln2: cpu 1.7~1.80vcore, soc voltage: 1.25, pll 2.2v (above 2.5v os even not booting), sb 1.30, dram 1.35~1.80

 

if anyone have useful hints what can i improve or try with my system i appreciate any kind of help. thanks.

cheers

bala

 

Clear CMOS, set 111 MHz BCLK Frequency, Save + Exit. Does it work or not? What POST code do you get stuck at?

[Alex@ro]

After binning some dozen X1700X i have to share that about 15% of cpu's did not do 3200 divider and 10% would get stuck at 55 over 110 bclk,each cpu was tested after clear cmos with saved profile,same componentes,everything same.Also,most of x1700x did 4000-4050 cinebench,good pieces 4070-4100,best is 4150@ 1.42 llc2 R15 on ambient water.

 

also , Malay or China it did not make a difference in oc ability.

Edited March 29, 2017 by Alex@ro

[elmor]

The best LLC setting for me is Level 1: 1.35v set in BIOS you have 1.35v in IDLE and 1.37v in LOAD. BCLK is linked to PCI-Express clock

 

Elmor, any change in seeing OC Panel 1 firmware soon?

 

The OC Panel II firmware works on OC Panel I.

 

 

Thank you, missed that post.

 

 

 

 

 

Just one last question: Do you have access to subtimings like the way you have for command rate (ie: you can set a fixed value in a specific bios version)? Or don't you have any access?

 

No access at this time.

[Reltdeats]

Hi there elmor,

 

Thank you for the amazing work updating this mobo bios! :ws:

 

I need some advice to get this GSKILL 3200MHz CL14 working at 3200MHz with this mobo.

 

Im using the 0038 bios (cause GSKILL recomends 2N CR with this ram) with Ryzen 1800x. Max I can get for now is 2666MHz 14-14-14-14-34 without CPU OC.

 

I really need to get 3200MHz for a photogrammetry process (software is not taking into account the CCX architecture from ryzen).

 

Any advance config u recomend me to try?

 

(Sorry about the bad english...)

Edited March 31, 2017 by Reltdeats