-
Posts
633 -
Joined
-
Last visited
-
Days Won
10
Content Type
Profiles
Forums
Events
Blogs
Posts posted by ground
-
-
Wow! Nice run ;)
-
That Optane doing work! Getting into DDR5 now, great to see you back :D
- 1
-
love that dice spoon :D
-
holy shit, that SP though...
-
Found 11MHz more, but I'm sure there is more to dig out. Will continue trying in the next couple days, and once I found the limit take it under chiller
Who needs 1 DIMM per channel?
-
12 hours ago, George_o/c said:
Nah I asked because he mentioned it in the video that 6-8-6 works better for him than 7-6-5
he talks about 6-7-5, which is the usual hyper timingset.
-
I never got why you run 6-8-6, I get significantly better times at 6-7-5, around 4s between the two timing sets for 32M, assuming same clocks.
-
Have you tried CPUcool instead of setfsb? Gained me around 20 MHz FSB for valid on my Supermicro.
-
the 7€ sock keeps on going :D
- 1
-
Binning paid off :D best two ambient chips had coldbug at -40 and -50. Third one finally had a good cb…
-
4GHz 45nm; not maxed yet. Refusing over 4GHz uncore since I don't think that plays within the spirit of this challenge.
- 4
-
-
-
-
P6T just isn't a good board for BCLK. You can improve the behavior a bit by flashing it to P6X58D, but it still won't do amazingly well. Most boards top out around 8200 QPI, strongest boards I've had were doing 8800ish. There is little reason to increase QPI anyways, it only improves throughput to the chipset; not relevant to any 2D benches. As a matter of fact, I usually bench with QPI Slowmode enabled (which stock P6T doesn't support), since that allows running higher BCLKs.
CPU matters a lot for BCLK capability; 32nm chips vary from 220-260ish, with outliers into the high 270s (though 270+ is <1% of chips). 45nm is usually worse, with chips ranging from 210-240ish. early 30xx chips usually don't do particularly well on that either.
I never managed to kill any x58 boards with voltage, though scaling is often nowhere near worth it to bother increasing them far enough.
Good chips typically do 4500+ Cinebench R15 at 1.3V; going up to 1.4V my best chips end up in the 4760+ range.
CMOS battery holder I wouldn't even bother, I never bother with a bios battery anyways, makes booting up the system after a session on cold far less dangerous not running one in the first place, and improves your habit of documenting settings.
-
-
14 hours ago, TASOS said:
Here is how an Elpida Hyper looks like
Look for the dot's at the edge of the chip.
IMG
IC shape is also quite unique, only other DDR3 ICs with non rectangular edge are 1Gbit BASE (non hyper), which, to my knowledge, doesn't show up in bins higher then 1600c7 and 512Mbit BASE, though that one is almost square and isn't gonna show up in 2GB Dimms anyways.
1600 6-6-6, 1866 7-8-7, 2000 8-8-8, 2133 8-8-8, 2250 9-9-9, 2200 8-8-8, 2000 7-8-7 and 2250 8-8-8 are almost safe bets to get hyper (some gskill 2000 8-8-8 are BBSE).
In my experience, corsair bins are the best bet; as unobtainable as GTX2 and 2000c7 are, they are much stronger bins then pretty much anything else around. Best findable bins are triple kits of 1866c7 and 2000c8 corsair, but much beyond 2000 6-7-5-20 2.0V is not common on those bins on ambient cooling, though I've found some unicorns...
- 1
-
So glad to see this concept working out :D
- 1
-
Congrats!
First one managing Fullpot on Cezanne?
-
Who needs firewire, Esata or Audio? Congrats!
-
550 fsb on a 45nm quad on asus p45 is really impressive!
-
Crazy chip and great work getting the max out of it! :D
-
impressive! Board seems like magic for quads :D
-
crystalright - GA-EP45-UD3 @ 733.8MHz - 733.81 MHz Reference Frequency
in Result Discussions
Posted