Overclock.net › Forums › AMD › AMD CPUs › [Official] RYZEN 7 1800X | 1700X |1700 Owners Club & 4GHz+ Club
New Posts  All Forums:Forum Nav:

[Official] RYZEN 7 1800X | 1700X |1700 Owners Club & 4GHz+ Club - Page 837

post #8361 of 11723
I find IBT is great for finding instabilities nice n' quickly, but I agree one shouldn't hammer your CPU with it for hours on end. Still, personally, I don't qualify an OC as stable unless it has passed at least an hour using ~80% of total system RAM among any other testing done.

On that note, what do people like to use for other tests? I've seen posts about y-cruncher and realbench. Is there a consensus on what settings to use and how long to run those for? I've never used them before but I'd like to have an array of tests for when I think I'm all dialed in and ready to qualify my daily.
Edited by bardacuda - 4/11/17 at 8:01pm
post #8362 of 11723
Quote:
Originally Posted by rv8000 View Post

Not 100% positive, but IBT probably uses different CPU instructions. Different loading/instructions types will affect stability and require more/less voltage.

You may have a great chip, but i find that pretty insane when it takes 1.31v to stabilize my 1700 @ 3.8 for 12 loops of x264 ~ 30mins of testing.

1700x and 1700 are different though. For me its only a 300mhz OC for you its a bit more i believe.

Quote:
Originally Posted by chew* View Post

First of all...prime was updated to work on ryzen...so instead of outdated it is actually updated.

Secondly....i have hammered my system for god knows how many hours since ryzen released. Quality psu? No worries. Crappy psu? Run away and dont post stability in stability threads...

You want it all in one shot?

Run heaven and prime blend custom with 90% ram useage all at once...

RM1000 isn't a crappy PSU compared to some other ridiculously crappy ones, been fine on my 4690k at 1.3 for 24.7 overclock without issues been fine on this one. And i meant prime before it was updated lol.


Quote:
Originally Posted by mus1mus View Post

Nope.

1. Encoding load on Realbench is very quick per pass more often, CPU waits for LuxMark to finish before starting another Encoding pass.

2. Memory test on Realbench is pretty weak. HCI can show failure in a matter of seconds while Realbench passed for hours.

3. CPU lod pegged at 100% with Realbench does not mean equal load as IBT, Prime or OCCT etc.

4. Even GPU load on Realbench is not equal to Heaven or Firestrike. I can pass Realbench's Luxmark for an hour yet Furestrike will already artifact on the GT1.

Just one thing is useful for Realbench IMO. Encoding Benchmark with only Encoding ticked with Infinite loop enabled. That is only helpful when you are trying to pick up the Black screen issue for these chips. Still, can take quite a while.

IBT on the other hand, needs more Vcore to pass than Realbench. You should know what that means.

I am not saying your method is bad. As long as it works well for your need.
Correct.

+1

Correct but if its stable without even running realbench whats the point of running IBT? Not trying to be a smart ass just saying. You guys been using the one with avx or without?
    
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
post #8363 of 11723
Also what are good tests and their settings for finding memory instability? I was going to use Prime95 with 512/1024 FFT size w/ 80% RAM for an hour or two for that...but I've seen people running another program with multiple instances showing coverage % or something. What is that and what settings to use and for how long?
post #8364 of 11723
Y-cruncher isn't as quick as IBT for these chips. I also have an OC dialed in previously that passed 5 runs but Black Screened within 20 minutes of encoding. But for multi-platform testing, it's pretty good. Esp for Intel.

Prime doesn't need as much Voltage as IBT AVX but needs quite a time to show instabilities especially ones induced from Memory OC on this platform. Still, equally as good in my books.

HCI Memtest is purely Memory intensive. Needs hours to complete 400% coverage and is pretty good at finding little instabilities that can be masked with Prime or IBT.

Aida 64 Cache only test is also pretty sensitive. You can't pass 1 hour if IBT AVX doesn't pass 30 Very High runs. It may be a great test for DF induced instabilities. I need to further test this.

In my tests, VCore isn't everything. That's where Realbench Fails or any encoding for that matter.
post #8365 of 11723
Quote:
Originally Posted by bardacuda View Post

Also what are good tests and their settings for finding memory instability? I was going to use Prime95 with 512/1024 FFT size w/ 80% RAM for an hour or two for that...but I've seen people running another program with multiple instances showing coverage % or something. What is that and what settings to use and for how long?

I tick blend then custom then allocate about 90% ram in prime 95. Then i actually use system skype browsing etc (which is why 90% not 100% ).

It covers all fft sizes beyond what normal blend does and covers small/large fft. The coverage is far more extensive than any of the "presets" which can mask instability.

Once i have isolated memory, then cpu, then combined last but not least i toss 3d into the mix...loop a real game benchmark or heaven...

Keep in mind im a reviewer/overclocker.

It has to pass my worst case scenarios or bust so as not to mislead endusers and or potential customers.

If you notice...im one of the few running around toting a sub 4gig oc...pretty sad for a pro overclocker no?

It is not because my chip is a dud....
Edited by chew* - 4/9/17 at 10:49am
post #8366 of 11723
I don't use "very high" in IBT because it's only 4GB...which is not very high these days. I use 75-80% of whatever my total system RAM is. In my case it is 12288-13xxx...but if I had 32GB I would use at least 24576.

I will add those other tests to my...uhh...tests though. Thanks for your input! thumb.gif

EDIT: @chew Yeah I don't use presets either. 512/1024/12288 is a custom setting. I see blend goes as high as 4096 on FFT size. How long does it have to be running before it starts testing in the upper range?

EDIT2: I just realized I have Samsung Magician running and it allocates some RAM to use as a cache for the SSD. I should probably turn that off during my testing or use less RAM in my tests.
Edited by bardacuda - 4/9/17 at 10:17am
post #8367 of 11723
Therein lies the drawback of prime 95 barracuda....to hit all fft sizes repeatedly enough to cause a bsod requires 24 hours....

Thus the 24h prime requirement ive noted in the past...

And that sux...who wants to tie up a system for 24h...
post #8368 of 11723
I don't mind tieing it up for 8 or 12 or so because I can just leave it running overnight. What if you reduce the range? I suppose then the testing won't be as valid.
post #8369 of 11723
Quote:
Originally Posted by bardacuda View Post

I don't mind tieing it up for 8 or 12 or so because I can just leave it running overnight. What if you reduce the range? I suppose then the testing won't be as valid.

Preety much. Ive hit bsods 4-5 hours in in the past. Usually if i can get by 6-8 overnight it will go the full 24.

The full 24 is just for good measure.

This 24h test is almost complete.

5h to go...
post #8370 of 11723
I know when I was overclocking Ol' Bessie I passed 24h of Prime but failed to pass an hour of IBT...but that was years ago and I don't know if I was using good settings. Was probably using one of the presets.
New Posts  All Forums:Forum Nav:
  Return Home
  Back to Forum: AMD CPUs
Overclock.net › Forums › AMD › AMD CPUs › [Official] RYZEN 7 1800X | 1700X |1700 Owners Club & 4GHz+ Club