Techpowerup! Hwbot Club

[buildzoid]

Good news every one I just got my hands on an asrock A85X extreme 6 and an A10 6800k so I can join the FM2A10 OC competition. I also have a LN2 pot coming but for now I'll see if I can get the A10 to boot and screenshot at 5.6Ghz using an H100. I will only join for the CPU-z side of the competition because even my best ram stick only does 2240mhz at stock timings and 1.680v and is pretty unstable doing it.

 

[vega22]

what ram is it dude?

only 1.68v isnt lots

 

[buildzoid]

2133mhz 9-11-11-31 G.skill RipjawsX
I tried loosening all the timings by +1 to see if I could get 2333mhz running the results were that it booted I got into the OS tried to run XTU and got a BSOD

 

[d1nky]

I find with my ripjawsx I can get anywhere on 11-12-11 upto 2650mhz on normal cooling.

yours are the better binned ones so should do a little better

 

[buildzoid]

what voltage do you use though I do believe it might be the fault of my CPUs IMC and I haven't tried 11-12-11 yet so I'll see what that will do this weekend.

 

[Durvelle27]

New scores submitted

 

[d1nky]

hey guys Ive been really busy with work and training lately. 9-10hr shifts lately.

I hope everyone is benching hard!

 

[Durvelle27]

hey guys Ive been really busy with work and training lately. 9-10hr shifts lately.

I hope everyone is benching hard!

Nahhh

 

[buildzoid]

So I just tried to run some SuperPi and even at a higher RAM frequency I couldn't get below 7.191 and when I started upping CPU frequency and lowering RAM timings my times got worse and then got stuck at 7.548 any ideas on what the problem might be?

 

[d1nky]

sounds like instability, best try pushing a good ram speed/timings and then cpu speed

 

[vega22]

it is mhz vs latency dude.

slacker timings need more mhz to get back what you lose in latency.

 

[d1nky]

depends what he means by this, lower numbers/tighter or high numbers/looser?

and lowering RAM timings

 

[buildzoid]

Ok here's the full story if you check my last SuperPi score it's 7.191 and that's using 5.25Ghz CPU and ram at 2240 9-11-11-30-1T. Well today I tried Pi again except using 5Ghz and 2333 11-12-11-34-1T and got the same score so I tried to push higher and dropped timings down to 11-11-11-30-1T and upped the CPU to 5.25Ghz and guess what my time went from 7.191 to 7.548 srsly WTF higher CPU speed and tighter memory timings and I get a worse score.
BTW the modules I have aren't exactly the same one pair has a higher default tertiary timing (can't remember which).
In the end I just gave up and turned of my computer because I was too lazy to reset voltages back to my 24/7.
XTU is equally annoying never using 100% of the CPU yet causing higher voltage drops than any other benchmark nor does it scale well. 5.2ghz gets the same results as 5.3ghz and every time I open it up it crashes but still loads up but with misread frequencies reading 5.11 when I'm at 5Ghz.

 

[vega22]

in super pi mhz wins.

you might find the best score is with only 1 stick if one clocks higher on its own.

have you used aida64 to check if you are being throttled once you go past x52?

 

[buildzoid]

unfortunately I don't have aida 64 but I can say that all other benchmarks except for spi1m and XTU scale very well with frequency

 

[vega22]

it's the only thing i can think of that might be causing that to happen to your results dude

i dont know of any other software which shows throttling like aida but i am sure the stress test is in the free version too.

 

[Random Murderer]

Ok here's the full story if you check my last SuperPi score it's 7.191 and that's using 5.25Ghz CPU and ram at 2240 9-11-11-30-1T. Well today I tried Pi again except using 5Ghz and 2333 11-12-11-34-1T and got the same score so I tried to push higher and dropped timings down to 11-11-11-30-1T and upped the CPU to 5.25Ghz and guess what my time went from 7.191 to 7.548 srsly WTF higher CPU speed and tighter memory timings and I get a worse score.
BTW the modules I have aren't exactly the same one pair has a higher default tertiary timing (can't remember which).
In the end I just gave up and turned of my computer because I was too lazy to reset voltages back to my 24/7.
XTU is equally annoying never using 100% of the CPU yet causing higher voltage drops than any other benchmark nor does it scale well. 5.2ghz gets the same results as 5.3ghz and every time I open it up it crashes but still loads up but with misread frequencies reading 5.11 when I'm at 5Ghz.

in super pi mhz wins.

you might find the best score is with only 1 stick if one clocks higher on its own.

In SuperPi 1M, raw speed wins. However, in 32M, memory tweaking is extremely important, and dual-channel mem will see an improvement over a single stick, triple-channel will see an improvement over dual, etc.

unfortunately I don't have aida 64 but I can say that all other benchmarks except for spi1m and XTU scale very well with frequency

If XTU is scaling well, then you aren't throttling. XTU is damn hard on procs. If anything would cause throttling, XTU would.

it's the only thing i can think of that might be causing that to happen to your results dude

i dont know of any other software which shows throttling like aida but i am sure the stress test is in the free version too.

I'm pretty sure it's not throttling. If it were, XTU would have stopped scaling. My two cents: far too unstable to get consistent results. Once your RAM hits 2200, you need to set your CPU clockgen filter to 10uF, once you hit 2400 it needs to be set to 20uF. For speeds above 2000, Rampage Tweak should be set to 2 in your DRAM timing section. What's your VCCSA at for the 2200+ runs? Also, on the newer BIOSes, the RIVE doesn't seem to like flat timings. Instead of going for 11-11-11, try 10-11-11 or 11-12-11.

 

[buildzoid]

10-11-11 doesn't boot my vccsa is 1.05 also I have mode 2 in rampage tweak and 20uF set already

 

[EarthDog]

I guess that clockgen filter thing is only for ASUS boards? I have gone well over 3K without adjusting it...

 

[Random Murderer]

10-11-11 doesn't boot my vccsa is 1.05 also I have mode 2 in rampage tweak and 20uF set already

WHOA, 1.05 VCCSA is LOW, but possibly right for your setup. VCCSA should be within .6V of VDIMM and within .3V of VCore. I would bump it up to 1.1V and see if it helps.
I forgot to mention that around 2200 on the RAM and higher, PLL plays a role in stability as well. What is your PLL at currently?
EDIT: Clockgen filter should not be set to 20uF until you've hit 2400. Lower it to 10uF if you're in the 2200-2399 range.

I guess that clockgen filter thing is only for ASUS boards? I have gone well over 3K without adjusting it...

Yep. Just another option to play with, lol. I'm pretty sure it's for X79 only.

 

[buildzoid]

My PLL is 1.8375 I used to use 1.9 but that turned out be less stable so I dropped back down to 1.8375

 

[Random Murderer]

My PLL is 1.8375 I used to use 1.9 but that turned out be less stable so I dropped back down to 1.8375

Ok, so you have that covered. 1.8375 might be a bit off, but you've already started playing with it, so you're on the right track there.
Also, I'm not sure if you saw the edit I just added before you posted, but CPU clockgen filter should not be set to 20uF until you break 2400. In the 2200-2399 range, it should be set to 10uF.

 

[vega22]

In SuperPi 1M, raw speed wins. However, in 32M, memory tweaking is extremely important, and dual-channel mem will see an improvement over a single stick, triple-channel will see an improvement over dual, etc.

If XTU is scaling well, then you aren't throttling. XTU is damn hard on procs. If anything would cause throttling, XTU would.

i would agree with that whole heartedly dude, but as he is talking about a 7 second result i went along with the idea he must be running 1m and not 32m

he also said xtu isn't scaling past x52 and as it's a pretty hard test throttling would explain it.

 

[Random Murderer]

i would agree with that whole heartedly dude, but as he is talking about a 7 second result i went along with the idea he must be running 1m and not 32m

he also said xtu isn't scaling past x52 and as it's a pretty hard test throttling would explain it.

I missed the part where he said XTU isn't scaling well, lol
In that case:
@Buildzoid: Have you turned off Anti-surge support? Do you have CPU power duty set to thermal probe or extreme? And I can't believe I hadn't asked this before now, but how well are the VRMs cooled? In your system specs, it says "custom loop" but doesn't specify if you have a VRM block or not. If not, make sure you have decent airflow over the VRM heatsink.
If all those are satisfactory, then I'm out of ideas.

 

[buildzoid]

My VRM has a a water block on it and XTU never caused any major heat issues the only benchmark that put's my loop to the test is cinebench at 5.3Ghz and that still doesn't throttle.
BTW I just got a magical 7.176s run but then I got a clock interrupt not recieved just as I was opening the second CPU-z window