Jump to content
HWBOT Community Forums

Recommended Posts

Posted
Couple failed 32m 5Ghz runs I recorded,one on 4790k and other was 4770k,running all tight and both running as fast as my best time or very close to it

 

Down clock noneed

 

 

Awesome job Bully, I'm sure no one doubts any of your submission's and respect for getting the Giga sorted, I couldn't get 2750+ c8 on air to ever work and it might be time to revisit as I don't think any other board can run as tight of RTL at those frequencies.

Guest Bullant
Posted
Awesome job Bully, I'm sure no one doubts any of your submission's and respect for getting the Giga sorted, I couldn't get 2750+ c8 on air to ever work and it might be time to revisit as I don't think any other board can run as tight of RTL at those frequencies.

 

Thanks,yeah only problem with gigga ln2 board it's very fragile,I've had 2-3 of them now while testing on air get stuck on code 20,when this happens board just doesn't recognise memory .It starts of being able to boot with one dim then code 20 goes but aventulally code 20 is there to stay lol.The latest ln2 board it happen to me while trying to boot 5ghz air ,now if board does a shut down it'll be stuck on code 20.It also seems once code 20 is there to stay then the gigga ln2 board will only work with patriot psc memory I believe because the patriot psc dims are thicker going into the memory slot.

  • 4 weeks later...
Posted
Thanks,yeah only problem with gigga ln2 board it's very fragile,I've had 2-3 of them now while testing on air get stuck on code 20,when this happens board just doesn't recognise memory .It starts of being able to boot with one dim then code 20 goes but aventulally code 20 is there to stay lol.The latest ln2 board it happen to me while trying to boot 5ghz air ,now if board does a shut down it'll be stuck on code 20.It also seems once code 20 is there to stay then the gigga ln2 board will only work with patriot psc memory I believe because the patriot psc dims are thicker going into the memory slot.

 

I was lucky to find a new SOC Force LN2 board on Ebay for retail price but $35 shipping (outside of US) and the seller was TopDog. Also having stupid code 20 issues at times with certain dimms, was extremely frustrating trying to figure out for a noob like me but still happy being able to buy an unused and unabused board like this. Still working on getting used to Giga bios...classic obviously. Will post some 32m results soon in the other thread

Posted

My 4770K has a strange IMC. 1.250 SA is not enough to stabilise PSC at 1200 or Samsung at 1400, but 1.300 SA (works out at +450mV on ASRock) can easily run 1466 6-10-6 all day long with CPU on water.

 

I binned some G3258s on IMC earlier today, best ones can drop 1333 8-12-8 volts to 1.80:

 

snaphsot0001gmslx.png

  • Like 1
Posted
My 4770K has a strange IMC. 1.250 SA is not enough to stabilise PSC at 1200 or Samsung at 1400, but 1.300 SA (works out at +450mV on ASRock) can easily run 1466 6-10-6 all day long with CPU on water.

 

I binned some G3258s on IMC earlier today, best ones can drop 1333 8-12-8 volts to 1.80:

 

 

Oooo show some 2933c6, 2666c8 at 1.8v is pretty darn good, think you can drop tRFC to 88. That would be impressive!

Guest Bullant
Posted
I was lucky to find a new SOC Force LN2 board on Ebay for retail price but $35 shipping (outside of US) and the seller was TopDog. Also having stupid code 20 issues at times with certain dimms, was extremely frustrating trying to figure out for a noob like me but still happy being able to buy an unused and unabused board like this. Still working on getting used to Giga bios...classic obviously. Will post some 32m results soon in the other thread

 

Good luck man,post some results when you get it sorted

Posted

Hi Folks!

 

Just got a ram pot and hoping to give it a try sometime this week. Since I've never done it before, would appreciate any guides, tips, thoughts, etc to make sure everything's insulated properly, get the cold ram dialed in quickly, etc, etc.

 

From pics, I see that folks wrap the main power connector with shamwow or thick shop towels. Wondering...should I put electrical tape around the power connection and then wrap it in sham?

 

Figure for the first attempt, I'll put the g3258 under water. Have some decent psc sticks that do 2666 tight with 1.80 (stick 2)-1.82v (stick 1). Looking at what others have done, I should target 6-10-6? No idea what subtimings I should go for...Mr Bullant's posted some magnificent stuff but not sure it'd be reasonable for me to expect the same as what he's done.

 

Appreciate any guidance, links, tips, etc! :)

tia!

-Stefan

 

Update: Found these vids from Dinos22 but would love to find content for the z97ocf/psc setup:

  • 3 weeks later...
Posted

Playing around on my OCF with a couple new to me 3258's and two slices of "Pi". Water/Air cooled with my window open while it's snowing. This is about as tight as these will go and still be stable, I've run tRFC lower but not in this combination so maybe a couple ticks left there but otherwise that's it. Standard timings at 1.83v and locked down timings at 1.85v

 

JtQSqEO.png

 

QrkPVE4.png

 

Really thought I was running the speeds below, but sometimes the board boots this way and I have to use Rapid OC just one click up or down, any clock or voltage works. (So like Alt+2 then Alt+A, adds 1 to cpu multi)

 

vFuI2JE.png

 

Never had this one before, The Pi god's obviously weren't happy though

 

p6fRVvO.png

 

gUm6av6.png

  • Like 1
Posted

Maybe someone can explain this to me, since this thread is all about psc mems on OCF and we use super pi 32m to compare results on efficiency. Is it correct in saying you can take the seconds it takes to complete 32m and multiply them with cpu frequency to get a so called efficiency rating. Lets take my above run for example. It took 9m 13s 484ms or 553.484 seconds and multiply that by 3200mhz you get

 

v5WEAIJ.png

 

If you take the fastest 5g time (Bullants time obv.) of 5min 56sec 328ms and multiply that by speed of cpu or 5001.8MHz you get

 

7RdgtgM.png

 

Which led me to trying to figure out if my timings/run was any good, so I took my final number and divided it by 5000(mhz) and ended up with

 

RuXfKNS.png

 

That would be 5m 54s 230ms or almost exactly 2 seconds faster than the most efficient 5g time, so there must be a variable I'm missing because I doubt my run done with ice cold AIR at 2666c8 stands a chance against a run with ln2 on mems at 2750c6. So obviously the scaling isn't linear, but is there a better method of dissecting efficiency ?

  • Administrators
Posted (edited)

You can use IPC only at around same clocks and cpu to compare, the gain per 100MHz at same memory settings decreases the higher you get at frequency.

 

P.S. There might be a super formula which can calculate this starting from a base value, but I am not math prof^^

Edited by websmile
  • Like 1
Posted

Your efficiency is only valid as long as you keep RAM:CPU clocks at 1333:3200 ratio and don't change the timings. So at 5GHz you'd have to have the RAM running at 2083MHz with the exact same timings to maintain the same efficiency.

  • Like 1
Posted (edited)

So I diced my G3258 today and had a bash at 5GHz SuperPi http://hwbot.org/submission/3127041 - however I forgot to do Wazza it would've shaved 4 seconds off. :(

 

Tried dicing my ram and managed to get it to boot 2600 7-11-7-25-1T with 4/38 IOL/RTL @2.2v, but no stability. So I reverted to my air profile as you can see.

 

I was using the Kingpin Dominance pot with it just sitting on top of the ram modules, but I had 2x problems:

1) The pot was only making good contact with one stick.

2) The OCZ heatsinks are plastic so didn't hold the heat at all causing them to shoot from -20C (lowest I saw) to -6C really quickly.

 

There are pictures on the submission showing the thermometer readings for pot, right ram module and ram pot (in that order) as well as one showing how one of the sticks has lots of frost while the other doesn't.

 

So time to rip those heatsinks off and put on some spare corsair dominator ones I have laying around to try again.

Edited by Noxinite
Guest Bullant
Posted

Love you guys enthusiasm ,Noxinite this only psc and memory pot you have ?

Posted (edited)
Love you guys enthusiasm ,Noxinite this only psc and memory pot you have ?

 

Yeah. Got it for cheap with corsair heatsinks, so I thought why not.

 

Pot was at -50 ish and ram was hovering around -10. So when I get it setup so it actually makes contact it should be good. Hopefully. :S

Edited by Noxinite
Guest Bullant
Posted (edited)
Yeah. Got it for cheap with corsair heatsinks, so I thought why not.

 

Pot was at -50 ish and ram was hovering around -10. So when I get it setup so it actually makes contact it should be good. Hopefully. :S

 

Yeah nothing wrong with cheap ahah,yeah you'll need better contact,is the same pot I started out on with Pi ,I just used lacky bands to pull pot down to memory HS for ok contact.Your memory hs prob not good for this

Edited by Bullant
Guest Bullant
Posted

Ran a couple Pentium runs today 4-5Ghz

 

17.562.png

 

 

5GHz

p5g.png

  • 4 weeks later...
Posted

Any chance you could convince the sticks to do 9-12-12? What volts are you running now? I know Samsungs don't like cold a lot but -50°C or so may help a bit... or at least leave you more confident at running ~2.4V-2.5V or so.

Join the conversation

You can post now and register later. If you have an account, sign in now to post with your account.

Guest
Reply to this topic...

×   Pasted as rich text.   Paste as plain text instead

  Only 75 emoji are allowed.

×   Your link has been automatically embedded.   Display as a link instead

×   Your previous content has been restored.   Clear editor

×   You cannot paste images directly. Upload or insert images from URL.

×
×
  • Create New...