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

ROG Crosshair VI overclocking thread - Page 1286

post #12851 of 17462
Quote:
Originally Posted by bluej511 View Post

24 is normal so is 40. I had the same reset issue, your ram is the issue its failing its initial training. Its not a pll issue and changing pll may also mess with your tctl temps.

Try bios 1107 if it still needs a reset to boot then up soc slightly. Its a ram issue for sure.

PLL didn't work as you've predicted. 1107 also didn't fix the issue, will try pushing SOC.

Hopefully it's not because of that one instance where SOC was running at 1.4v after bios flash and now it's busting out on cold boot.
Mark IV
(18 items)
 
  
CPUMotherboardGraphicsGraphics
R7 Ryzen 1800x @ 4.0ghz Asus Crosshair 6 Hero AM4 Gigabyte GeForce GTX 1080 Xtreme Gaming Gigabyte GeForce GTX 1080 Xtreme Gaming Water 
RAMRAMHard DriveHard Drive
G.Skill F4-3200C15D-32GTZ  G. Skill l F4-3200C15D-32GTZ WD Caviar Black 1T Sata III Samsung F3 
Hard DriveHard DriveOptical DriveCooling
SAMSUNG 850 EVO 2.5" 500GB SATA III PNY 128gb SSD boot DVD-RW  MCP655-B + EX360 (push) + Xtreme Nova 1080 (p... 
OSMonitorKeyboardPower
Win 10 Samsung 28" U28D590D Cooler Master Storm Lepa G1600 
CaseMouse
Xigmatek Elysium CM sentiel 
  hide details  
Reply
Mark IV
(18 items)
 
  
CPUMotherboardGraphicsGraphics
R7 Ryzen 1800x @ 4.0ghz Asus Crosshair 6 Hero AM4 Gigabyte GeForce GTX 1080 Xtreme Gaming Gigabyte GeForce GTX 1080 Xtreme Gaming Water 
RAMRAMHard DriveHard Drive
G.Skill F4-3200C15D-32GTZ  G. Skill l F4-3200C15D-32GTZ WD Caviar Black 1T Sata III Samsung F3 
Hard DriveHard DriveOptical DriveCooling
SAMSUNG 850 EVO 2.5" 500GB SATA III PNY 128gb SSD boot DVD-RW  MCP655-B + EX360 (push) + Xtreme Nova 1080 (p... 
OSMonitorKeyboardPower
Win 10 Samsung 28" U28D590D Cooler Master Storm Lepa G1600 
CaseMouse
Xigmatek Elysium CM sentiel 
  hide details  
Reply
post #12852 of 17462
Well, I decided to take a Test and Guinness hiatus to actually see what might happen regarding the whole SOC value and Latency debate. Unfortunate it got ugly in his postings, but whatever, I went for some data. Unless someone actually wants to sift through the 12 screen snips, I will only post the results. Here is what I did:

1. Keep ALL settings identical with the exception of changing SOC values.
2. Reboot, test in AIDA, do a snip of results, close AIDA.
3. Test in SIV64, do a snip of results, close SIV64
4. Reboot, rinse, and repeat for each SOC value.

Here is the result:



So what does that mean? It means I need another Guinness!

OK, the assertion was that I should look at the following and my original response using 1.1 for my SOC volts was:

1. "if you have higher then 3.0 latency on l2 cache now you're SoC voltage is not right"

My L2 Cache using SIV64 reads 3.7 ns, so this may be 0.7 ns high?

2. "if you have higher then 10.0 on l3 cache your latencies are not right"

My L3 Cache using SIV64 reads 7.7 ns, so I think I am looking good there.

3. "also if you l3 cache is not above 400 GB/s you are running a bad SoC voltage and it needs to be tuned"

My L3 Cache in AIDA is 415.31 GB/s for Read, 410.84 GB/s for Write, and 480.16 for Copy, so that seems good.

So based on these results I should change my SOC volts to get L2 Cache latency lower?

Then a debate blossomed regarding his premise that SOC volts should be tuned for Cache latencies - with that challenged by others asserting that CPU frequency changed these latencies, not SOC Volts. That did not make the individual happy.

My conclusions are:

1. His theory that an L2 Cache latency higher than 3.0 ns means I need to tune my SOC volts to reduce it. FAIL - No matter what the SOC volts, my L2 Cache latency never changed, nada, none at all.
(Caveat, I had to test this value using SIV64, while he uses AIDA. The methodology variance could be significant, but I should have seen at least SOME change.)

2. His theory that "if you have higher then 10.0 on l3 cache your latencies are not right". UNDETERMINED - I have no idea since all my results were below 10 ns.

3. His theory that "also if you l3 cache is not above 400 GB/s you are running a bad SoC voltage and it needs to be tuned" PASSED - SOC values really did have an effect. Though four out of six voltages had results above 400 GB/s, 1.11V saw the write and copy fall off significantly. 1.15V on the other hand really had odd results in AIDA, low on Read and completely wacky on Write and Copy.

So I started this with a SOC value of 1.1 volts. Did I see a reason to change it? Happily and coincidentally, no. PHEW!

thumbsupsmiley.png

Lastly, it really was regrettable that the tenor in this debate turned ugly on one side. That guy may have knowledge, but the demeanor is not for this place. This is a great forum, and a superb thread - thanks to the high caliber of the participants.

And a +Rep to orlfman! Both for the good information, and for your considerable restraint in not throwing mud back. A measure of character IMOP.
Edited by CeltPC - 4/30/17 at 11:11pm
post #12853 of 17462
Quote:
Originally Posted by CeltPC View Post

Well, I decided to take a Test and Guiness hiatus to actually see might happen regarding the whole SOC value and Latency debate. Unfortunate it got ugly in his postings, but whatever, I went for some data. Unless someone actually wants to sift through the 12 screen snips, I will only post the results. Here is what I did:

1. Keep ALL settings identical with the exception of changing SOC values.
2. Reboot, test in AIDA, do a snip of results, close AIDA.
3. Test in SIV64, do a snip of results, close SIV64
4. Reboot, rinse, and repeat for each SOC value.

Here is the result:



So what does that mean? It means I need another Guiness!

OK, the assertion was that I should look at the following and my original response using 1.1 for my SOC volts was:

1. "if you have higher then 3.0 latency on l2 cache now you're SoC voltage is not right"

My L2 Cache using SIV64 reads 3.7 ns, so this may be 0.7 ns high?

2. "if you have higher then 10.0 on l3 cache your latencies are not right"

My L3 Cache using SIV64 reads 7.7 ns, so I think I am looking good there.

3. "also if you l3 cache is not above 400 GB/s you are running a bad SoC voltage and it needs to be tuned"

My L3 Cache in AIDA is 415.31 GB/s for Read, 410.84 GB/s for Write, and 480.16 for Copy, so that seems good.

So based on these results I should change my SOC volts to get L2 Cache latency lower?

Then a debate blossomed regarding his premise that SOC volts should be tuned for Cache latencies - with that challenged by others asserting that CPU frequency changed these latencies, not SOC Volts. That did not make the individual happy.

My conclusions are:

1. His theory that an L2 Cache latency higher than 3.0 ns means I need to tune my SOC volts to reduce it. FAIL - No matter what the SOC volts, my L2 Cache latency never changed, nada, none at all.
(Caveat, I had to test this value using SIV64, while he uses AIDA. The methodology variance could be significant, but I should have seen at least SOME change.)

2. His theory that "if you have higher then 10.0 on l3 cache your latencies are not right". UNDETERMINED - I have no idea since all my results were below 10 ns.

3. His theory that "also if you l3 cache is not above 400 GB/s you are running a bad SoC voltage and it needs to be tuned" PASSED - SOC values really did have an effect. Though four out of six voltages had results above 400 GB/s, 1.11V saw the write and copy fall off significantly. 1.15V on the other hand really had odd results in AIDA, low on Read and completely wacky on Write and Copy.

So I started this with a SOC value of 1.1 volts. Did I see a reason to change it? Happily and coincidentally, no. PHEW!

thumbsupsmiley.png

Lastly, it really was regrettable that the tenor in this debate turned ugly on one side. That guy may have knowledge, but the demeanor is not for this place. This is a great forum, and a superb thread - thanks to the high caliber of the participants.

And a +Rep to orlfman! Both for the good information, and for your considerable restraint in not throwing mud back. A measure of character IMOP.

Great stuff. +Rep.

You ever manage to get 4.0GHz out of your beast?
post #12854 of 17462
Quote:
Originally Posted by S1L3N7D3A7H View Post

Great stuff. +Rep.

You ever manage to get 4.0GHz out of your beast?

HAHA, no, I'm afraid I got side tracked, and now I am all tested out for the day. As Scarlett said, “After all, tomorrow is another day!”
post #12855 of 17462
Quote:
Originally Posted by CeltPC View Post

HAHA, no, I'm afraid I got side tracked, and now I am all tested out for the day. As Scarlett said, “After all, tomorrow is another day!”

Whelp, glad you ran your test for us though, you really didn't have to. It became clear, the poster had no evidence for their claims. I was highly skeptical, since I had seen no real difference in any SOC voltage on my latencies when I had been checking for stability. Since every time I tune a frequency on my system, I check scores in Cinebench15 and AIDA64 Cache Benchmark, then proceed to stability test.
post #12856 of 17462
Quote:
Originally Posted by Kriant View Post

PLL didn't work as you've predicted. 1107 also didn't fix the issue, will try pushing SOC.

Hopefully it's not because of that one instance where SOC was running at 1.4v after bios flash and now it's busting out on cold boot.

Yikes that is a pretty good overvolt, hope that is not the problem too. This board seems to tolerate a lot, I have not really heard of actual failures except for that early bios deficiency problem.
post #12857 of 17462
I am having great results with 1007 bios. My question is when going for higher memory, I want to get the fastest infinity fabric for games. So is say 3600 mhz with a lower BCLK and higher strap faster or slower than the same 3600 mhz with a higher bclk and lower strap? Which is more important for infinity fabric the higher BCLK or it doesn't matter, only ram speed primarily then latency?
post #12858 of 17462
Quote:
Originally Posted by hotstocks View Post

I am having great results with 1007 bios. My question is when going for higher memory, I want to get the fastest infinity fabric for games. So is say 3600 mhz with a lower BCLK and higher strap faster or slower than the same 3600 mhz with a higher bclk and lower strap? Which is more important for infinity fabric the higher BCLK or it doesn't matter, only ram speed primarily then latency?

Not Gospel, but in the testing I have read in regard to Ryzen FPS with games, the most important factors were in descending order:

1. CPU Frequency

2. Memory Frequency

3. Memory Timings

I have not seen any comments that it matters how you get there. Get all of these optimized for your rig and you should be Golden. Needless to say the graphics card is the elephant in the room. Regardless of what flavor your card is, tuning / overclocking it will net you as much or more in FPS gains as anything else.
Edited by CeltPC - 4/30/17 at 11:33pm
post #12859 of 17462
Quote:
Originally Posted by CeltPC View Post

Not Gospel, but in the testing I have read in regard to Ryzen FPS with games, the most important factors were in descending order:

1. CPU Frequency

2. Memory Frequency

3. Memory Timings

I have not seen any comments that it matters how you get there.

For games it is actually

1. mem freq
2. cpu freq
3. mem timings

See here:


tl;tw The 1700 at stock freq with faster memory is beating 1800x at higher freq with slower memory.
Main Rig
(16 items)
 
  
CPUMotherboardGraphicsRAM
AMD Ryzen 1700 - 3.810GHz - 1.286v Asus Crosshair VI Hero MSI 1080 Ti Gaming X - 1.000mv 2012/6210 G.Skill Trident Z 16GB 3600Mhz CL16 RGB 
Hard DriveHard DriveOSMonitor
850 evo 120GB (for OS + programs) 850 evo 250GB (for Games) Windows 10 LG 27ea73LM-P 
KeyboardPowerCaseMouse
Razer BlackWidow Ultimate 2013 Corsair RM750x Corsair Obsidian 750D Razer DeathAdder 2013 
Mouse PadAudioAudioAudio
Razer Goliathus Razer Kraken Pro Sennheiser HD 598 Sound Blaster Z  
  hide details  
Reply
Main Rig
(16 items)
 
  
CPUMotherboardGraphicsRAM
AMD Ryzen 1700 - 3.810GHz - 1.286v Asus Crosshair VI Hero MSI 1080 Ti Gaming X - 1.000mv 2012/6210 G.Skill Trident Z 16GB 3600Mhz CL16 RGB 
Hard DriveHard DriveOSMonitor
850 evo 120GB (for OS + programs) 850 evo 250GB (for Games) Windows 10 LG 27ea73LM-P 
KeyboardPowerCaseMouse
Razer BlackWidow Ultimate 2013 Corsair RM750x Corsair Obsidian 750D Razer DeathAdder 2013 
Mouse PadAudioAudioAudio
Razer Goliathus Razer Kraken Pro Sennheiser HD 598 Sound Blaster Z  
  hide details  
Reply
post #12860 of 17462
Update: Woke up and turned my pc on, back to f9 and immediate boot loop. Guessing it has to do with memory since it's the only thing that resets. Going back to 2933 an 1.0v of soc and it also dropped my cpu package power from 29w to 23w (still a couple watts more then before so the issue was down to soc).

Shame it's not sticking to 3200mhz, may try tweaking prodct but not sure it may help with cold boot. For now im gonna stick with what works and see if it happens again. It if does then its down to the BIOS as nothing else has changed.
    
CPUMotherboardGraphicsRAM
R7 1700X asus crosshair 6 Sapphire R9 390 Nitro CMK16GX4M2B3200C16 
Hard DriveHard DriveCoolingCooling
Kingston HyperX 3K Seagate ST2000DM006 EKWB Supremacy Evo EK-XRES 100 Revo D5 - Acetal 
CoolingCoolingCoolingCooling
Alphacool GPX r9 390 m01 alphacool xt45 240 Alphacool xt45 360 Alphacool D5 PWM 
OSMonitorKeyboardPower
Windows 10 LG 29UM68 Corsair Strafe Corsair RM1000 
CaseMouseMouse Pad
Thermaltake X5 Corsair M65 Pro RGB Razer Goliathus Speed 
  hide details  
Reply
    
CPUMotherboardGraphicsRAM
R7 1700X asus crosshair 6 Sapphire R9 390 Nitro CMK16GX4M2B3200C16 
Hard DriveHard DriveCoolingCooling
Kingston HyperX 3K Seagate ST2000DM006 EKWB Supremacy Evo EK-XRES 100 Revo D5 - Acetal 
CoolingCoolingCoolingCooling
Alphacool GPX r9 390 m01 alphacool xt45 240 Alphacool xt45 360 Alphacool D5 PWM 
OSMonitorKeyboardPower
Windows 10 LG 29UM68 Corsair Strafe Corsair RM1000 
CaseMouseMouse Pad
Thermaltake X5 Corsair M65 Pro RGB Razer Goliathus Speed 
  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