SuperPi32m 5Ghz Ivy All Out Challenge!

[Splave]

For sure,a good wazza will make your run start fast and help your last itineration greatly.Also great bandwith and tight latencies helps the middle run.killer run with great wazza would mean great initial value as well...at least that's my findings so far

Alex check this one
http://i.imgur.com/jI6Uwu9.jpg

I have a run at 6m16.210 that started 2.266 and was my fastest run till last loop got beat by .016 at then end

[angoholic]

You can check your waza with a polynomial closeness function eg with Matlab. I did this to do a forecast when testing.
Even Excel diagrams with loop differences are sufficient to check efficiency.

[bullant]

I haven't tried 7-9-6 yet but I've tried 6-9-6 2600Mhz and it fails quick on 32M,temps only -95 tho,will try 7-9-6 next time .Even booted into window,just @ 5-10-5 2600Mhz tWCL6,Did a run same as my above sub with tWCL8,so slow

[bullant]

This probably won't drive since i don't have CPU-Z (motherboard) and to cheap to pay for Aida64 but i just want to be a part of this epic thread

xBrytanx - 6:18:562 - 49x102.1 - 1327MHz - 7-11 -7-28 ASRock OC Formula - Gskill PI 2200 CL7 2x2GB - 2.1v - Cpu@H20/RAM -150c LN2 (first freeze)

https://www.dropbox.com/s/v7kvlzyd26...0run%205.0.JPG

Welcome mate,well done on your first freeze,nice time,be sure to look deep into everyone's subs,that is the key.Well done

[hiwa]

Just some trick for u guys ....
cl6 wl 6 is not always faster than cl7 wcl6, sometimes cl7 wcl6 or 7 at higher freq (2800)is fasterthan them all, best tets on memories and ivy 32m is
Try high cl and high wcl ( just be sure it can pass ) but tighter possible on other timings ( test them one by one ) when you find best combination just set 2.15 go to lowest temp your sticks handle or your pot can go, then change only cl and twcl, you cant go wrong on them, there are some ruls on ivy performance,
Cl 6 twcl 6 , Cl 7 twcl 6 for 2600 to 2700 mhz
Cl 7 twcl 6 , cl 7 twl 7 gpfor over2700 to 2800 mhz
for sure 8gb is faster than 4gb about .500/.700 so calculate how u can run ur sticks at what freq and cl twcl and others.
Cw is OS depending, my Os and cw cant be same as bullant if we use 2 different ISOs
670 maxmem is safe maxmem ( no more ram error ) and its cpu imc depending, some cpu wont run low maxmem at all.

[bullant]

Great advice Hiwa,I found 4Gb v 8Gb not that much of difference because I can push the Mhz on 4Gb to catch up to 8Gb,I've only tested 4gb a few times in last few months compared to 8Gb.So much stress on 8gb and can't tighten subs as much as 4Gb.But on a side note what you have said about the cl and tWCL and maxmem is the key,I haven't even messed with maxmem for long time,thanks for the advice

Edit maybe I miss understood you and you meant what I said,can push 4gb harder on subs to make up difference

[XBrytanX]

Thanks Hiwa this was exactly what was happening @ 2666 not 2700+ tho ,great advice

[hiwa]

Bullant did u tried higher CW when u run 8gb? i can see on ur ARM (should be around 57XXXX with ur maxmem)seems u are using low cw, or u are run Pi when Available and system cache are not balanced well.
tra higher CW 3Dmark01 files and only D to C and C to D once then change the time u wait for available and system cache to have higher ARM and u can do 15s with those timings and freq and 8Gb .
or try 640 maxmem and same CW u are trying for 670 maxmem.

a lot of times when u have even low ARM the run is fast, but for sure the run with higher ARM and right CW if the combination is good it's the fastest run.
i found Ivy 32m very interesting but the big issue is there is no a lot of ruls to respect, sometimes the setup its the winner sometimes same setting/setup its loss, need to try and try changing CW and maxmem.

[bullant]

Bullant did u tried higher CW when u run 8gb? i can see on ur ARM (should be around 57XXXX with ur maxmem)seems u are using low cw, or u are run Pi when Available and system cache are not balanced well.
tra higher CW 3Dmark01 files and only D to C and C to D once then change the time u wait for available and system cache to have higher ARM and u can do 15s with those timings and freq and 8Gb .
or try 640 maxmem and same CW u are trying for 670 maxmem.

a lot of times when u have even low ARM the run is fast, but for sure the run with higher ARM and right CW if the combination is good it's the fastest run.

I haven't tried the ARM higher before,I always been hitting it @51XXXX-53XXXX,so I should be hitting it higher with 8Gb and with my high mazmem =57XXXX.So if I was to bring maxmem down should I then bring ARM down too.I'll try the 57XXXX next time for sure,also think Ive still not quite sorted out wazza and everthing that goes with it system cache and available.Thanks for the advice Hiwa much appreciated

Ok i can try 640 maxmem with same setting

[l0ud_sil3nc3]

revisited the second set of 2300 c8's, but this time with a different cpu, and similar results:p









I ended up going to a higher maxmem (640) and started seeing some better results, going to give 670 a shot again after a refill.

Frosty pics in a bit. . .

[Alex-Ro]

George can you run them at higher speeds rather than tight timings?Say 7-11-7,or 7-11-8 at 2800+ ? Thinking a strong imc can cope with that,and i think higher freq can give nive boost too

[l0ud_sil3nc3]

George can you run them at higher speeds rather than tight timings?Say 7-11-7,or 7-11-8 at 2800+ ? Thinking a strong imc can cope with that,and i think higher freq can give nive boost too

2700 7-11-7-28 tWCL 7 seemed a bit slower, but I am about to refill so I can try another kit which is good for 2700 c7.

[l0ud_sil3nc3]

Probably one of the frostiest sessions









slowly working it's way down



and a new low











SF3D from last session, awesome pot just didn't seem to get as cold.

[Splave]

Epic

[jimba86]

ice,ice baby..show me more..

nice pics and results. what pots are they? SF3D triple point and KPC venom?

[l0ud_sil3nc3]

ice,ice baby..show me more..

nice pics and results. what pots are they? SF3D triple point and KPC venom?

Bingo, the last two pics are the Triple Point, basically the non gold one.

[XBrytanX]

xBrytanx - 6:18:031 - 50x100.0 - 1300MHz - 7-10 -6-22 ASRock OC Formula -Adata 2000x (BBSE) CL9 2x2GB - 2.1v - Cpu@H20/RAM -80c LN2 (not enough LN2 tho ..2650 was reachable or maybe more )

[Alex-Ro]

that is one hell of a isolation,nice job man,keep pushing!

[XBrytanX]

thanks bro ,you still don't want this sticks ?

[Alex-Ro]

Don't know man,i have quite a stash to test under cold

[l0ud_sil3nc3]

Sam's BBSE under cold







brix shat:p



This kit is good for 2600 on air, but it seems that it does not like voltage, all the above was done with 1.9v as more would cause it to fail quicker. Also dimm temps needed to be at -85c to -100c as much more would cause it to fail. I am starting to think some dimms do not like the extreme cold, and once again, bad on air = good on cold, and good on air = bad on cold. This kit would not do 2600 c7 or c6 at any volts.

I corrupted Windows more times in this session as setting anything => 2.0v was insta corrupt and reinstall.

Still have the black PCB to test

[TaPaKaH]

does it stop at 1300 even if you loosen up the other timings?

shouldn't have sold you the 2200C9s, damn it

[l0ud_sil3nc3]

They were booting 2700+ 7-10/9-7-22 and finish 1-2 loops before failing so the are pretty good. I will test them in the backup board as I have never had any issues setting vdimm over 2.0v so hoping it is my board and not your kit.

[Alex-Ro]

nice job george,i think the magic combo would be a kit that does 2600 on air and eats voltage and cold,too many parameters to meet it seems those kits would be extremely rare....

[TaPaKaH]

If the second 2133C8 kit is also owned by Allen's Avantium that fail 1200 8-10-7 on air (which is very likely) then BBSE binning becomes a clear issue. If you start looking for bad air kits then you'll end up with stuff that fails 1066 9-9-9 and I bet is equally bad cold. And testing each kit on LN2 is not very practical from time/money/resale viewpoint.

All hopes now on PSC at being more consistent...