Asrock Z97 OC Formula-PSC

[l0ud_sil3nc3]

What happened to them mate?

 

 

Sold maybe?

[Noxinite]

Playing with my XDZ G.Skill Trident 2000 9-9-9 some more (1.97v):

 

[Guest Bullant]

Waiting on the cold results

[Noxinite]

Waiting on the cold results

 

Me too

 

Got a 4770K the other day, but I cannot get it to run over 2400 on PSC. Samsung works fine at similar volts to the G3258 I used for all previous PSC testing, but I cannot get PSC to clock at all on this 4770K.

 

Getting some passes now. It needs completely different SA/IOA/IOD volts to my G3258 and 4670K, but also needs some other settings tweaking.

Edited February 18, 2017 by Noxinite

[GtiJason]

Got a new Z97 OCF, these boards never fail to amaze me. Not sure why this is but these 4 dimmers, for the mostpart overclock better than any other boards I've had / have. I can cold boot psc at 2666-2750MHz+ and Samuel "D" at 2800-2933MHz, and on all the other top boards (mostly 2 dimmers) just training 2666/2800 can be a real headache. The only thing that has kept me holding on to these other boards is the slightly better rtl's, but have noticed a reaccuring cycle going on where I get OS's prepped for 2dimm boards, get boards prepped for water, phase or lin, get excited to run 32m, enter the bios and am angrily cleaning up an hour later with little to no progress and we're already multiple generations past Haswells heyday. This is my fourth Z97 OCF that just works, and does it well. Here's a few pics of first session, OS mostly stock since I'm binning 4770k's for... imc / 5g-5g volts / all out 32m (ln2)

 

 

 

 

 

Album- ASRoĉk Z97 OCF - 2017 4770k binning - Album on Imgur

Edited April 28, 2017 by GtiJason

[Noxinite]

Me too

 

Got a 4770K the other day, but I cannot get it to run over 2400 on PSC. Samsung works fine at similar volts to the G3258 I used for all previous PSC testing, but I cannot get PSC to clock at all on this 4770K.

 

Getting some passes now. It needs completely different SA/IOA/IOD volts to my G3258 and 4670K, but also needs some other settings tweaking.

 

I never updated with I found, so here it is:

 

Using normal 0.3/0.25/0.3 IMC volts (and low/low/high IMC volts) it wouldn't pass tight 2600C8 or 2666C8 with my (2666C8 @1.84v) PSC. But Samsung seemed to clock just as well as on my 4670K, this was rather confusing so I spent ages trying different volts and bios settings to get it working.

 

Eventually I found that I had to use 0.45/0.25/0.45 and manually set the 'training error' setting to 18 to get it to pass @1.84v. Then I found I couldn't get 2736C8 to work at those settings and I didn't dare increase the IMC volts further also setting the 'training error' setting to 19+20 didn't work either.

 

 

I ran some 4GHz 32M with Samsung and got my first sub 20s run:

 

 

I even managed to get a 5GHz 32M run with Samsung (only 4.8GHz cache though):

 

 

And here's some 2900 4/4/42/43 Samsung 32M:

 

 

4770K OCing

Batch = L315B325

 

Non-Delid:

R15 = 4.5GHz @1.14v and 4.7GHz @1.35v. 83C Highest temp.

32M (2/2) = 4.5GHz @1.11v and 4.8GHz @1.28v. 54C Highest temp.

 

Delid (Kryonaut):

R15 = 4.5GHz @1.14v and 4.7GHz @1.33v. 73C Highest temp.

32M (2/2) = 4.5GHz @1.11v and 4.9GHz @1.35v. 52C highest temp.

[Noxinite]

Update with 4770K on phase.

 

Managed to get into the 56s at 5GHz/5GHz with Samsung on air and 4770K on phase. The chip can do 5GHz/5GHz @ 1.2v core/1.3v cache/1.9v input on phase and compared to barely managing to do 5GHz/4.8GHz at 1.4v/1.35v/1.9v for 32M on AIO is some pretty good scaling

 

I managed to get 2918 42/43/4/4 working at 2.36v on my sticks. They just kept eating those volts and taking it. Funnily at lower bclock and higher ram divider it wouldn't pass the initial loop compared to passing with wazza at the shown settings.

 

Here's the screenshot:

 

And the score:

Noxinite`s SuperPi - 32M score: 5min 56sec 969ms with a Core i7 4770K

 

Also, had a play around with max 32M clocks on phase and managed to get 5.6GHz/5.2GHz working at 1.53v/1.44v/2.4v. 5.5GHz would work with 1.9v input, but 5.6GHz required the massive step up.

 

Here's the screenshot:

 

And score:

Noxinite`s SuperPi - 32M score: 5min 23sec 500ms with a Core i7 4770K

[Matt26LFC]

Just bought a Asrock Z97 OCF so hopefully soon I'll get to do some playing too!

[Matt26LFC]

Done a little playing with 4770K 5G/5G and PSC, nothing great as still learning

 

CPU is on Phase and mems are on air

 

Edited September 7, 2017 by Matt26LFC

[Noxinite]

Waiting on the cold results

 

Done.

 

These sticks seem to have no CB and no CBB. This was done on G3258 which is one of the reasons it is soo slow and it was only on phase.

 

One of two wazza passes in the whole session:

 

Noxinite`s SuperPi - 32M score: 7min 19sec 313ms with a Pentium G3258

 

 

I managed to get 2800C6 (quite loose) to pass with non-wazza, but I couldn't get it to pass with wazza at all - even with +0.03V. Which is a bit odd as on air wazza barely needs more voltage than a basic 32M pass.

 

 

So my questions are:

1.) Does wazza capability decline past a certain point with temps even if basic 32M will still pass?

2.) Does wazza capability decline past a certain point with voltage even if basic 32M will still pass?

3.) Would colder IMC help with wazza capability?

 

@l0ud_sil3nc3 @Splave

 

Any help is appreciated guys.

Edited September 12, 2017 by Noxinite

[l0ud_sil3nc3]

Done.

 

These sticks seem to have no CB and no CBB. This was done on G3258 which is one of the reasons it is soo slow and it was only on phase.

 

One of two wazza passes in the whole session:

 

Noxinite`s SuperPi - 32M score: 7min 19sec 313ms with a Pentium G3258

 

 

I managed to get 2800C6 (quite loose) to pass with non-wazza, but I couldn't get it to pass with wazza at all - even with +0.03V. Which is a bit odd as on air wazza barely needs more voltage than a basic 32M pass.

 

 

So my questions are:

1.) Does wazza capability decline past a certain point with temps even if basic 32M will still pass?

2.) Does wazza capability decline past a certain point with voltage even if basic 32M will still pass?

3.) Would colder IMC help with wazza capability?

 

@l0ud_sil3nc3 @Splave

 

Any help is appreciated guys.

 

To answer your questions:

 

Just because you can pass without wazza at said frequency/timings doesn't mean you will be able to pass with wazza with more cold or voltage.

 

It's best to find the max voltage you can run and pass with (usually 2.15v to 2.25v for me depending on the kit) and also the the exact temperature you need for your kit. This can be -125c to full pot depending on the kit.

 

Colder imc usually helps to a certain point, at least -60c to -125c should be a good range to try. Sometimes cold isn't necessarily better but in my experience 90% of the time it's better. Don't be afraid to blast cache voltage and core as well if your cpu needs it.

 

Pm me if you have any more questions and most important have fun

[Noxinite]

Thanks L0ud!

 

I tried again yesterday with cold 4770K and first boot 2800C6 tight passed at 2.3V, but for some reason it was single channel. Then it randomly started coming back and I was getting crashes at stock, so I had to stop the session.

 

CPU was -85 ish and ram was full pot.

[Guest k.pisti]

Hey guys!

 

I know this topic is mostly about PSC, but when binning 2Gbit CFR, what formula should I follow? Either of the bios profiles is good for a start? TBH I dislike them, most are loose at some settings that makes binning with them useless, but I'm not too good with CFR yet.

I acquired 50 sticks of generic single sided 2Gbit CFR sticks, would like to bin them properly. I ran a few in a M5E with a good IMC 3770k, but only tried 2400C10-12-11. Best pair of 8 sticks I tried did spi32m at 1.62V at above speeds.

[Don_Dan]

I think the highest spec for CFR was 1400MHz 11-14-14-35 @ 1.65V, so you could use this as a starting point for your binning.

Maybe @TaPaKaH can also chime in.

[TaPaKaH]

Binning CFR is a waste of time. I had a special 2x4GB kit from G.Skill but even that could not compete with PSC or Samsung D-die. So you have no chance to find something actually useful on OEM.

[yosarianilives]

CFR may be decent for a frequency run, but since there's also such a thing as mfr it's not nearly as useful anymore.

[Noxinite]

Yo, yo. It's that time of year again. Lol.

Cold PSC ftw.

I'm on the Impact VII this time to see how it compares to OCF last year (and because it has working USB ports).

I managed 2800 6-10-6-22 1T, with all sub-timings tight (including WRRD 13 and REFI 65535) except RFC which was at 72. This passed 32M wazza at 2.24V.

35/36/4/4 would boot and run, but 32M exited halfway through each time. And anything over 2800 would fail to train and give me 55/FE codes until I warmed the ram up (either lots or just a bit). Even loosening everything excet CAS it still wouldn't train. The one time it did train was only single channel.

I still have ~10L (I think) so was going to try cold CPU as well? I'm assuming that the IMC is holding me back at this point?

Edit: I was running 4GHz on G3258 and got a 7m 18.1s run.

Edited October 10, 2018 by Noxinite

[TaPaKaH]

Cold on the CPU usually doesn't make a difference. I ran 1450 6-9-6, 1466 6-10-6 and 1500 6-11-6 all with CPU on AIO.

Try booting into OS at 2.20V and then increasing voltage from there, so that Vttddr stays at 1.10V. Most kits don't like above 1.10.

[Noxinite]

Anyone else got any tips or advice?

@l0ud_sil3nc3 @Splave @Bullant

Edited October 10, 2018 by Noxinite

[yosarianilives]

4 hours ago, TaPaKaH said:

Try booting into OS at 2.20V and then increasing voltage from there, so that Vttddr stays at 1.10V. Most kits don't like above 1.10.

Should be able to lock vttddr to 1.1 in the tweakers paradise section anyways?

[Splave]

I thought i remeber that the impact like synced rtl timings like 35/35 or 36/36 for pi

[Bullant]

Hey Nox, you tried putting a USB header on the bottom of the ocf board and running USB from thier ? Or the pcie USB thing in the gpu slot we use on z270 onwards for w7-xp? 

The impact is little different board, I've not really had enough time on it to suggest something.Your best bet is Sam or Alex

 

I've been looking locally for a cheap impact to try myself

[Noxinite]

Yeah, I tried a header but it died eventually. USB card works, but I wanted to see if the Impact was excetionally better.

Tbh, they both did 2800C6 around 2.3V real.

Sub with LN2 PSC: https://hwbot.org/submission/3955332_

Cold CPU didn't allow me to magically boot any higher, think there is a setting wrong somewhere as I tried all loose except Cas and it still got stuck trying to train; on code '15'? Never had that on ambient.

Tried to do high freq 32M with 2666C6, but it failed straight away. More time needed on that (finding good IMC volts at high VCCIN) might focus on maxing out CPU first though.

Edited October 12, 2018 by Noxinite

[yosarianilives]

I'm sure this is already known, but on the m7i if you select additional training and always retrain then the reset button works the same as "retry" on newer boards. I found that the one time I pushed PSC that this made a huge difference as sometimes you have to retrain several times to get it post when you're on the edge of stability. Although I'm not sure how much of that is related to my mem as it wasn't binned and definitely aren't my best sticks based on later binning. http://hwbot.org/submission/3805983_yosarianilives_3dmark03_hd_graphics_(haswell)_26018_marks 

Edited October 12, 2018 by yosarianilives

[Noxinite]

The post code '15' issue I had was with tight and loose subs though.

I might have a try on the OCF and see if it's the same behavior. But I think one of the sticks isn't as good as the other and/or doesn't like the lower temps as much.