Overclock.net › Forums › AMD › AMD Motherboards › ROG Crosshair VI overclocking thread
New Posts  All Forums:Forum Nav:

ROG Crosshair VI overclocking thread - Page 1099

post #10981 of 17443
Quote:
Originally Posted by Frikencio View Post

That happened me with my Phenom II system, HW detected 16Gb but only 12Gb usable (4x4Gb sticks). Turned out that one stick of ram was not seated properly.



Well i tryed all slots and always the same thing... I'll try to run memtest and see if the stick is faulty.
Interesting thing is, i have better results for ram in Aida with 1 stick (single channel) than with 2 sticks (dual channel)...

Quote:
Originally Posted by Praz View Post

Hello

The accessible memory timings have no impact on an incorrect ProODT setting. Memory voltage will impact ODT only if the reflections are marginally over the limit of stability.

Thx for the info thumb.gif
post #10982 of 17443
Quote:
Originally Posted by Reikoji View Post


~4149 Wrecking it... 98.8 bclk


~4156 at 102 bclk... also throws out errors because that tiny bit over 4150 too much for 1.5v! whats happening here?!

I guess I might not want to worry too much about my memory speed...

system hang if I try to go too low, figures. Stopping at 98.8 for now ! gonna do some tests.
Quote:
Originally Posted by Reikoji View Post

is it really HPET off cheating the scores? or is HPET on cheating the scores redface.gif I think the cenebench run should let on that its an actual performance gain, or does cinebench go by some weird calculation and not how fast the processor completes the task? I certainly don't know for sure, but what I do know is HPET off + bclk change + sleep = doom.

Whats needed is some gaming benchmarks done in this scenario do come to better conclusion.

Changing the BCLK with HPET off instantly skews the timer, and after that (until the original BCLK is restored), your internal timing isn't going to reflect reality. Your realtime clock is going to run fast/slow, your FPS measurements will be off and all benchmarks will be invalid.

Yes, this is cheating, CPU-Z doesn't care, but I think 3D Mark doesn't accepts results from Ryzen with HPET off for those reasons. I'm not aware of any way to fix the TSC if the BCLK is changed after boot.

In short: you can cheat the benchmarks, but you won't see any real gains in any task (the numbers will look good, but that's because what the computer thinks is 1 min is actually 1.02 mins)
    
CPUMotherboardGraphicsGraphics
AMD Ryzen 1800X Asus Crosshair IV Hero NVidia GTX 980 Ti NVidia GTX 660 
RAMCoolingCoolingCooling
G.SKILL 64GB (4 x 16GB) Ripjaws V Series DDR4 3... EK-Supremacy MX EK-FC980 GTX Ti WF3 EK-CoolStream XE 360 
KeyboardPowerCaseMouse
Corsair K90 EVGA Supernova 850 T2 Cooler Master - K280 Corsair M60 
  hide details  
Reply
    
CPUMotherboardGraphicsGraphics
AMD Ryzen 1800X Asus Crosshair IV Hero NVidia GTX 980 Ti NVidia GTX 660 
RAMCoolingCoolingCooling
G.SKILL 64GB (4 x 16GB) Ripjaws V Series DDR4 3... EK-Supremacy MX EK-FC980 GTX Ti WF3 EK-CoolStream XE 360 
KeyboardPowerCaseMouse
Corsair K90 EVGA Supernova 850 T2 Cooler Master - K280 Corsair M60 
  hide details  
Reply
post #10983 of 17443
Ok, so I have problems getting correct temperature readings on my 1700.
Sensemi skew disabled gives me way too high temperatures and auto offset (272) gives me way too low temperatures (sub ambient on idle).
What is the correct sensemi offset for the 1700? I am on bios 1002 btw.
Tried sensemi offset 072 (200 extra on 272 is for the 1700x and 1800x 20°C offset?) and got some reasonable readings.

Here are my readings after 5 mins of realbench stress testing at stock using the 072 offset (H80 cooler):

post #10984 of 17443
Quote:
Originally Posted by r4m0n View Post


Changing the BCLK with HPET off instantly skews the timer, and after that (until the original BCLK is restored), your internal timing isn't going to reflect reality. Your realtime clock is going to run fast/slow, your FPS measurements will be off and all benchmarks will be invalid.

Yes, this is cheating, CPU-Z doesn't care, but I think 3D Mark doesn't accepts results from Ryzen with HPET off for those reasons. I'm not aware of any way to fix the TSC if the BCLK is changed after boot.

In short: you can cheat the benchmarks, but you won't see any real gains in any task (the numbers will look good, but that's because what the computer thinks is 1 min is actually 1.02 mins)

Well it was interesting nonetheless. Idk how timer skews can alter ones FPS in a game tho, sounds far fetched but guess it all goes by system time in the end. It is what it is I suppose.

I hope no ones looking at me like i'm trying to cheat the system redface.gif I only turned HPET off to test Elmors app and then ran into this interesting affair ! Now that i know whats up, its HPET back on for me.

Is it just Ryzen that has this effect of bclk changes with HPET off? no other processor?
Edited by Reikoji - 4/23/17 at 8:23am
post #10985 of 17443
Quote:
Originally Posted by Reikoji View Post

With the 3600mhz kit, you have a high ability to hit the 3200mhz kits speed and timings, 3600mhz with relaxed timings and bclk increasing (if youre lucky). With the 3200mhz kit you'll have to relax the timings are run at 2666 or 2933 or ratios for the timings.

thank you! thumb.gif
post #10986 of 17443
http://valid.x86.fr/2nmnh3

This one is real and uncheated, i swear! HPET is back on.

Its not too much lower than the cheated one to be honest. I wish those numbers were real though redface.gif.

I'd like to know a bit more about HPET tho. People have reported performance gains in gaming from the simple act of turning HPET off. I didn't see any mention of changing the bclk after the fact. Is the true nature of this actually that they saw the same fake increase due to the system clock being tricked, and it doesnt really matter if its on or off? I'd immagine they would have been increasing their bclk to eek out some more performance tho, not decreasing it. increasing it with HPET off would yield fake performance losses.

Also. Windows 10 clean installs supposedly have HPET off by default as I have heard (I think). Wouldn't this be a bad thing? I upgraded from windows 7 and it is supposedly on by default there, thus it was already on for me in windows 10. I suppose its because they had no expectation of anyone altering the bclk within windows 10? Seems retarded to have such a sensitive item turn off by default.

Is there an HPET option hidden somewhere in the bios or is it forced enabled/disabled?
Edited by Reikoji - 4/23/17 at 9:01am
post #10987 of 17443
Quote:
Originally Posted by poisson21 View Post

I didn't reinstalled windows for now, and like you i 've had a fx before, and no problem for now. I'll do a fresh install when i'll get a m2 and my new gpu later this year.
And i'm stable at 4.07Ghz with 64G ram kit at 3200 Mhz.

What CPU do you have and ram timings do you use?

I'm stable on 1700 @ 3.937GHz with 1.38V
64Gig Ram running @ 3173 with 14,14,14,14,34 T1 settings and 1.37V
Edited by buttmen - 4/23/17 at 9:17am
post #10988 of 17443
Quote:
Originally Posted by SpecChum View Post

Woohoo!

I'm not going to lie, I've had my 960 EVO for a couple of months now and can't say I've noticed any difference between that and my Crucial MX SSD's.

Yes, I'm also quite dissapointed, even thought of returning it. It is the same speed or slower than my 850 PRO in everyday windows 4k snappiness, the only
place it is faster is in large sequential read/writes (which I almost never do). I guess if you are copying huge files or working with them it makes sense. But
I just have it be be future proof over buying a regular ssd and hope AMD drivers can speed it up a bit later.
post #10989 of 17443
Quote:
Originally Posted by hotstocks View Post

Yes, I'm also quite dissapointed, even thought of returning it. It is the same speed or slower than my 850 PRO in everyday windows 4k snappiness, the only
place it is faster is in large sequential read/writes (which I almost never do). I guess if you are copying huge files or working with them it makes sense. But
I just have it be be future proof over buying a regular ssd and hope AMD drivers can speed it up a bit later.

I was thinking about getting one after seeing what speeds they claim they operate at redface.gif. I suppose if its not really like 400k iflops and 3400 sequential read and such i'll hold off for now. It may be driver related, or even immature platform related. thats goin kinda far for false advertisment :3
post #10990 of 17443
Quote:
Originally Posted by Reikoji View Post

Well it was interesting nonetheless. Idk how timer skews can alter ones FPS in a game tho, sounds far fetched but guess it all goes by system time in the end. It is what it is I suppose.

I hope no ones looking at me like i'm trying to cheat the system redface.gif I only turned HPET off to test Elmors app and then ran into this interesting affair ! Now that i know whats up, its HPET back on for me.

Is it just Ryzen that has this effect of bclk changes with HPET off? no other processor?

FPS = number of frames / duration of measurement. The time of measurement is derived from the internal timers, and if they don't match the real time going outside the computer, the number isn't frames per second anymore. If you have a monitor with Freesync that's capable of showing the true FPS on itself, you could compare with the software FPS and see the skew in action.

As for other platforms, I don't know. It all depends on if the processor can detect a changing BCLK and keep the timers in sync. HPET is an extra module with an independent clock, so it can't be skewed without a lot of extra work.

Quote:
Originally Posted by Reikoji View Post

http://valid.x86.fr/2nmnh3

This one is real and uncheated, i swear! HPET is back on.

Its not too much lower than the cheated one to be honest. I wish those numbers were real though redface.gif.

I'd like to know a bit more about HPET tho. People have reported performance gains in gaming from the simple act of turning HPET off. I didn't see any mention of changing the bclk after the fact. Is the true nature of this actually that they saw the same fake increase due to the system clock being tricked, and it doesnt really matter if its on or off? I'd immagine they would have been increasing their bclk to eek out some more performance tho, not decreasing it. increasing it with HPET off would yield fake performance losses.

Also. Windows 10 clean installs supposedly have HPET off by default as I have heard (I think). Wouldn't this be a bad thing? I upgraded from windows 7 and it is supposedly on by default there, thus it was already on for me in windows 10. I suppose its because they had no expectation of anyone altering the bclk within windows 10? Seems retarded to have such a sensitive item turn off by default.

Is there an HPET option hidden somewhere in the bios or is it forced enabled/disabled?

I'm not sure exactly where the overhead of checking the HPET comes in, but I know the main timer on Ryzen (TSC) is kept at a register, so its access is pretty much instantaneous.

As for disabling the HPET in the BIOS, I haven't found it anywhere either.
    
CPUMotherboardGraphicsGraphics
AMD Ryzen 1800X Asus Crosshair IV Hero NVidia GTX 980 Ti NVidia GTX 660 
RAMCoolingCoolingCooling
G.SKILL 64GB (4 x 16GB) Ripjaws V Series DDR4 3... EK-Supremacy MX EK-FC980 GTX Ti WF3 EK-CoolStream XE 360 
KeyboardPowerCaseMouse
Corsair K90 EVGA Supernova 850 T2 Cooler Master - K280 Corsair M60 
  hide details  
Reply
    
CPUMotherboardGraphicsGraphics
AMD Ryzen 1800X Asus Crosshair IV Hero NVidia GTX 980 Ti NVidia GTX 660 
RAMCoolingCoolingCooling
G.SKILL 64GB (4 x 16GB) Ripjaws V Series DDR4 3... EK-Supremacy MX EK-FC980 GTX Ti WF3 EK-CoolStream XE 360 
KeyboardPowerCaseMouse
Corsair K90 EVGA Supernova 850 T2 Cooler Master - K280 Corsair M60 
  hide details  
Reply
New Posts  All Forums:Forum Nav:
  Return Home
  Back to Forum: AMD Motherboards
Overclock.net › Forums › AMD › AMD Motherboards › ROG Crosshair VI overclocking thread