WR? 2x512mb BH-5 @ 282mhz 2-2-2-5 @ AIR!

[afireinside]

Really? Damn... I had a set drop dead at 3.5 Crap just randomly dies.

[STEvil]

dont confuse CH-UTT with BH-UTT.

CH cannot handle volts and will eventually die

[Zac89]

ehm.... a pair of mine is CH 'cos they need 3,1v to do 200 2-2-2 and over 3,5v they are not fully stable....guys you are really making me think i'm lucky! (ehm... my bad english.... dunno if the grammatic is right... )

[bachus_anonym]

Really? Damn... I had a set drop dead at 3.5 Crap just randomly dies.

Your stuff is PC4000 Redline - UTT-CH not UTT-BH5...

EDIT: STEvil beat me to it

Your CPU-Z is right. Combination of multiplier 11 + divider 183 (multi + 1) is bugged, divider 166 (multi + 2) is used instead.

If you had a x11 multi CPU you would know that CPU-Z is wrong in reporting RAM frequency @ x11 div183 I had done a lot of benching with my SD 3700+ and my only conclusion was that there is no "div180" - only div183. Everest shows true RAM speed It's easy to prove - run any benchmark @ x11 "div180" and x12 div200... Results will be identical...
I also PMed Franck long time ago about it and that's what he said: "The 180MHz setting is more doubtful. In fact I assume that the controller does not allow to go at more than 180, and in your case a 12x ratio would mean 183MHz. So the question is : is the limit 180 or 183 ? I don't know, as this setting is kind of "not official". That's why until now, he left it alone

Nice results, but I have two kits that say that isn't a WR
2X512MB Corsair 3500C2 - 285Mhz 3.86V on P4GD1 (for now )
2X512MB G.Skill 1GBGH - 286Mhz 3.92V on DFI Ultra-D
http://img7.imageshack.us/my.php?image=2861yp.jpg

If you guys want to break RAM speed records for 1M, could you at least try to get a decent time Micutzu, your 1M is very slow for such clocks

[-De$troyeR-]

32M.........???

maybe tomorrow

OFF
to LV
did you get my last pms?!

nope, pm's = disable

really good, but why your spi shows "real memory=90755072"??

maxmem = 87mb (smaller windows)

maxmem = 104mb about 0,2sek slower.........

[Micutzu]

You ran UTT at 3.92V!?!? You sir, have king kong sized testicles

As STEvil and Bachus said, it's BH-UTT and can handle volts like mad. I've ran them at 3.9V 24/7 for some time ... even went a little over 4V for this screenshot:

really good, but why your spi shows "real memory=90755072"??

He is probably using Maxmem tweak.

[EvilCloudStrife]

i was lead to beleive that this was original bh-5.

those are really good scores, and HIGH volts. good job

[funkflix]

Everything on that 3.4GHz screen is just perfect: kick a$$ CPU clocks on water, awesome RAM clocks and very fast 1M to top it off
Just art

Second that!

[jmax_oc]

not quite, I think I had 296.5 1:1 the other days for 1m, I ca di like 285mhz for 4m I think

Can you please show us screens ?

I've done 289.9Mhz spi1m with my 2x512mo sticks (3.9V)

[-.-PhanTom-.-]

Did you use active cooling on the RAM for this?
I have same RAM that I haven't really tried to push yet... wonder how they would do in my current rig.... still like the idea of 2 Gb meory for every day stuff

[jmax_oc]

Of course, everyone use fans

[masterofpuppets]

Of course, everyone use fans

My oldskool BH-5 doesn't even heat up while running memtest #5 at 3.4v, I know it's nowhere near 3.9v, but I'd expect them to at least get warm. But it's still best to keep a fan over the RAM just to be safe...

[Spyrus]

As STEvil and Bachus said, it's BH-UTT and can handle volts like mad. I've ran them at 3.9V 24/7 for some time ... even went a little over 4V for this screenshot:





He is probably using Maxmem tweak.

Xm i aggree on that but i cannot explain the fact why OCZ gives warranty for max vdimm for BH-die(GOLD) at 3.1 and for CH-die 3.6(VX).

And so many here run GOLD VX for 3.3-3.4 & above every day and BH for 3.1 @ 240 max for 24/7.

The only reason i am suspect (except the fact that CH need a lot voltage to run at 250 2-2-2 & so OCZ had to support high vdimm in order these sticks to run actually at specs) is to push the market to buy CH more than BH.

My opinion & from old BH experience is that BH can handle up to 4Volts( it is written in Winbond manuals) and CH less (i do not know exactly).

So OCZ should reverse the supported vimm for GOLD VX & GOLD GX.

[Micutzu]

Both are rated for max 3.6V, but BH is on larger fabrication process (175nm) compared to CH (130nm), that's why it can handle more volts.

Winbond BH
Winbond CH

[Lastviking]

Many CH -UTT sticks have died becuse of high vdimm but i stil have not seen so many dead old bh5 or bh5 -utt that have died by high vdimm, i have realy triend to kill 1 kit of bh5 -utt with high vdimm but it didnt work (high is 3.9vdimm+)

The most sticks thas die that have Ch - utt is muskin redline pc4000 sticks.

[Spyrus]

Both are rated for max 3.6V, but BH is on larger fabrication process (175nm) compared to CH (130nm), that's why it can handle more volts.

Winbond BH
Winbond CH

You are right!
By the way i had these manuals a long ago.
Thank u for find them again.

So is there any explanation why GX do not have 3.6Vdimm warranty protection?

[dinos22]

maxmem = 87mb (smaller windows)

maxmem = 104mb about 0,2sek slower.........

do you use different values for 1M/8M/32M runs

nice results

[funkflix]

So is there any explanation why GX do not have 3.6Vdimm quarantee protection?

ICs not equal PCB!

[Spyrus]

I had the opinion that both are Brainpower PCB 808.
Am i missing something?

Other manufacters do not have a vdimm limited warranty so can someone put as much as he wants without voiding!!??

[beta67]

Your stuff is PC4000 Redline - UTT-CH not UTT-BH5...

EDIT: STEvil beat me to it

If you had a x11 multi CPU you would know that CPU-Z is wrong in reporting RAM frequency @ x11 div183 I had done a lot of benching with my SD 3700+ and my only conclusion was that there is no "div180" - only div183. Everest shows true RAM speed It's easy to prove - run any benchmark @ x11 "div180" and x12 div200... Results will be identical...
I also PMed Franck long time ago about it and that's what he said: "The 180MHz setting is more doubtful. In fact I assume that the controller does not allow to go at more than 180, and in your case a 12x ratio would mean 183MHz. So the question is : is the limit 180 or 183 ? I don't know, as this setting is kind of "not official". That's why until now, he left it alone
If you guys want to break RAM speed records for 1M, could you at least try to get a decent time Micutzu, your 1M is very slow for such clocks

What do you mean by "the controller does not allow to go at more than 180.." 200 divider works fine @ 11 multi its true 1:1. Use the A64 mem calculator to calculate the tricky 183 divider. Destroyers mem did "only" run @ 261 Mhz.

[mursaat]

My oldskool BH-5 doesn't even heat up while running memtest #5 at 3.4v, I know it's nowhere near 3.9v, but I'd expect them to at least get warm. But it's still best to keep a fan over the RAM just to be safe...

Hmm mines get really hot at 3.4v in same test. And my case has one lid open, with 4xcase fans + 1 fan over pwmic + 1 fan over ram. Can it be such a difference between our temps?

//EDIT: @beta67: I think that's correct, I run the config in my sig: 289x9, 183 (180 in BIOS) divider and the ram runs at 260mhz. The problem is CPU-Z updates voltages and cpu speeds "on the fly" but to see a divider/mem speed change you have to close and reopen? Not sure, I use too many programs to know exactly how everyone works For example:

Smartguardian: reads fan speeds on the fly, needs to reopen to see changes in cpu voltage
MBM5: reads fan speeds, needs to reopen to see changes in cpu voltage and speed
Clockgen: changes htt on the fly, needs to reopen to see changes in mem divider.

The Clockgen+CPU-Z issue is really curious, because you can end being able to screenshot the memory at 300+ 2-2-2-5 if you change the divider to a lower one (lets say 166) and raising HTT in Clockgen without closing Clockgen nor CPU-Z.

I'm currently at work, but I will try to take the screenshot at home and post it to prove (or not, I'm not 100% sure in CPU-Z)

[funkflix]

What do you mean by "the controller does not allow to go at more than 180.." 200 divider works fine @ 11 multi its true 1:1. Use the A64 mem calculator to calculate the tricky 183 divider. Destroyers mem did "only" run @ 261 Mhz.

It runs with 283MHz...! Try to catch this SuperPi with "real" 260MHz... Dunno, why the people do not believe this.

11 x 200 = 2200 | 2200 / 183 = 12,02 ~= 12 | 11 x 309 = 3400 | 3400 / 12 = 283,33MHz

[celemine1Gig]

It runs with 283MHz...! Try to catch this SuperPi with "real" 260MHz... Dunno, why the people do not believe this.

11 x 200 = 2200 | 2200 / 183 = 12,02 ~= 12 | 11 x 309 = 3400 | 3400 / 12 = 283,33MHz

Hi funkflix,

check this link:

http://math.gogar.com/athlon64.cgi?htt=307&showtable=1

Destroyer's RAM really "only" ran at ~260MHz. Even though it ran slower than he thought, it's still a impressive result and a really good SP time.

Check the bottom of the page where this quote can be read:

Fields in red are where the A64 produces different memory speeds than expected from the formula cpuspeed/ceil(multi*200/memclk)

[funkflix]

Destroyer's RAM really "only" ran at ~260MHz.

Nope!





SuperPi don't lie.

[celemine1Gig]

If you think so...