Winbond BH-5, The Final

[MaSell]

Thank you guys. Brainpower PCB, double sided, 4 chips per side. Corsair 3200LL and Mushkin 3200 LVL2. I think I'm able to break 350MHz pretty easy with one stick, but I will leave this possibility as a backup

As I tested more than 50pcs of BH memory, I can share few interesting facts:
- high binned sticks are mostly crap (<290MHz), nice clocks on low voltages (3.3v) but poor scaling and can't take much voltage.
- avoid single sided, double sided 4 chips per side are usually better.
- BH5 = BH6 in terms of overcloking. BH6 are generally weaker because of cheaper PCB.
- poor scaling with cold, don't expect to gain 50MHz just because of subzero.
- the rule of "good batch" or "good week" doesn't exist, I have 300MHz+ "crappy week" stick and <270MHz with "legendary magic numbers".
- there is extremely difficult to find a stick able to scale up to 4.0v or higher. I found only 3 of them.

Thank you for that, MaSell (btw, where did you go?).

I was just round the corner

[dinos22]

Welcome


http://valid.canardpc.com/show_oc.php?id=1890576



World record for Winbond BH-5 dual channel. I've got also WR for Samsung TCCD 417MHz DDR834.

Regards

hahahah i'm still laughing about the TCCD record, you beat Andre there too, he was not happy! lol

AWESOME overclocking man, well done

[Vassili]

Amazing, xtreme respect for this!

[Eldonko]

Man I remember I used to be happy with 250Mhz haha, this is awesome. WTG m8

[Lastviking]

GratZ to the new WR!! , that speed is sick!

[Aerou]

Amazing.

Hats off also to the CPU IMC

[danielkza]

equivalent DDR3 latency: DDR-2752 @ 6-6-6-15



Those are some amazing clocks and timings. BH5. Wow. Now that was back in the day.

(nitpicking)

Not true, DDR3-2752 has a command clock of 1376MHz. 6 cycles would yield an effective latency of 4.36ns (6 cycles * 10^9 nanoseconds / 1376 * 10^6 cycles per second = 6 * 1000 / 1376 = 4.36) while 2 cycles of a DDR module at a command clock of 344MHz would result in 5.8ns. If you think about it is quite easy to figure out: since you're fixing triple the latency you need to triple the clocks too. 3 * 344MHz = 1032MHz = DDR3-2064.

(/nitpicking)

[sauria]

Nice, I knew I kept my Corsairs for a reason, one set is new.

[I.nfraR.ed]

I think this is the right thread.
Here's the new record, I believe...

512MB Team Xtreem DDR-600 TCCD @ 440.5MHz (DDR-881) 3-6-6-15 2T
DFI LP Utlra-D
Athlon 64 3400+ Venice E3

http://valid.canardpc.com/show_oc.php?id=2482084





Competition ain't over yet, but I'll be surprised if someone manage to beat this. Max on air was ~427MHz same timings.

PS: Btw, does anyone know what is the record for s.A FSB? I'm first in the bot with my NF7-S, but not sure if seen something better in the past.

[Reefa_Madness]

That's very impressive, for sure. Congrats.

[MaSell]

That's impressive for sure, you've got great mems and even better cpu!

[I.nfraR.ed]

Christian Ney already beat me, lol. And I thought it would be hard.
Have to try again.

[sin0822]

very impressive.

[Don_Dan]

Well done I.nfra.Red!

What was the Vdimm needed for these clocks?

[OLDcomer]

Nice job, Ivan! Keep up the good work.

[I.nfraR.ed]

Thanks

What was the Vdimm needed for these clocks?

~3.33V