2GB OCZ PC2-8000 Titanium Alpha VX2 @560mhz 4-3-3-4 / @600mhz 5-3-3-3

[eva2000]

with my 2x1GB OCZ PC2-8000 Titanium Alpha VX2 memory, I surprised myself in breaking my personal best single Super Pi 32M time at 3600Mhz core2 duo cpu clocks - and my PB was on 965P/975x of around 13min 37-39 seconds.

But this is on Asus 680i Striker of all boards and with sub 600mhz memory clocks. But guess tight timings + tight sub timings helped with just 533mhz 4-3-3-4 2T at 2.35v bios set vdimm

System:

• Intel Core 2 Extreme X6800 L627A872 B2 Step 6
• Swiftech Apogee/Radiical Triple Fan Single Pass Rad/MCP655 + Kayl 100W Custom PSU
• Asus 680i Striker Extreme 0402 bios
• 128MB Gainward FX5200 PCI card
• 2GB OCZ PC2-8000 Titanium Alpha VX2
• 4x80GB Hitachi 7K80 SATAII Raid 0 on SATA3 to SATA6 ports
• Pioneer DVD Burner
• 600W OCZ EvoStream

• Old PB Single Super Pi 32M time on Asus 680i Striker Extreme = 13m 52s 313ms
• New PB Single Super Pi 32M time on Asus 680i Striker Extreme = 13m 33s 812ms and 13m 32s 907ms

Old PB Single Super Pi 32M time on Asus 680i Striker Extreme


New PB Single Super Pi 32M time on Asus 680i Striker Extreme


better run


1M


Everest Ultimate Bandwidth


Yes the X6800 being B2 stepping might have helped too against my old E6600 B1, but I'm putting my money on the tight subtimings being the main factor in such fast times

[wa77ss]

That VX2 stuff is beast!

[Gautam]

Do you really think that that tRCD 3 is really tRCD 3?

I know we see people doing insane 4-3-3 speeds on 680i every day, even past 600. But if you try those parameters on any other chipset, chances are you'll lose a good 100-200MHz off your clocks.

Anyways, good looking times, didn't mean to put a damper, but the tRCD 3 on 680i has been bothering me for quite a while.

[eva2000]

Gautam i would suspect the same that tRCD might not be really = 3 IF I wasn't getting such fast super pi 32m times at 3.6ghz. So I think tRCD is really = 3

[Gautam]

It's a good time, but I don't think its fast enough to show that that's indeed tRCD 3...it's still slower than this by quite a margin.

[ozzimark]

Gautam i would suspect the same that tRCD might not be really = 3 IF I wasn't getting such fast super pi 32m times at 3.6ghz. So I think tRCD is really = 3

try benching 1m 4-4-3, 4-3-3, 3-3-3 and 3-4-3 at the same speeds? that'll tell the truth imo

[eva2000]

will do but for now

560mhz 4-3-3-4 2T 3-1-3-6-34 at 2.5v vdimm

Yup tight ass main and sub timings on Asus 680i Striker really do make a huge difference. Who would of thought I could pull off sub 13min 30s Super Pi 32M times at 3600mhz with a 680i chipset motherboard without even clocking memory above 560mhz (DDR2-1120) speeds!

Jumped all the way to 560mhz 4-3-3-4 2T with sub timings of 3-1-3-6-34 at 2.5v vdimm bios set. According to memtest86+ v1.70, reported bandwidth at 560mhz 4-3-3-4 2T was 5413MB/s versus at 533mhz 4-3-3-4 2T of 5209MB/s.



Digital camera snapshots of 32M run with 3x cpuz windows open + memset window.









Final result = 13m 28s 391ms



Re-ran 32M but with cpuz and memset windows closed = 13m 27s 719ms !



1M time = 13.922s



Everest Ultimate Bandwidth with 43.3ns latency!



Hexus Pi Fast



All I can say is bloody fast for 680i chipset

[guess2098]

George you can still make it faster



this is with copy trick


i ran this a while back ago, detail timing already forget.......

[eva2000]

Dam that's fast Is that with winxp or win2k3 ?

copy trick and maxmem = 600 both result in same times for me on other platforms for some reason with winxp heh...

edit: oh i see ram is at 600mhz a good 40mhz more hehe

[guess2098]

it is 2k3, yeah it is 600mhz on 2.5v
and OS is very clean

btw that ram died at the 2.9v on commando XD

[guess2098]

also here is the one without copy trick

[eva2000]

Nice set of memory you had there Denny, shame it died man

I think i'll need to loosen the sub timings to hit 600mhz and see how that goes. Guess win2k3 would help too but I'm on WinXP Pro SP2 heh.

[dinos22]

nice RAM danny too bad it died

[eva2000]

Denny i'm getting close

Managed to clock ram up to 600mhz 5-3-3-3 2T but with even further tighten the sub timings even more but had to loosen CAS from 4 to 5 in order to keep vdimm at 2.5v bios set ceiling (within OCZ warrantied max vdimm of 2.52v)

Memtest86+ v1.70 reported bandwidth

• @533mhz 4-3-3-4 2T 3-1-3-6-30 at 2.35v vdimm = 5,209MB/s
• @560mhz 4-3-3-4 2T 3-1-3-6-34 at 2.5v vdimm = 5,413MB/s
• @600mhz 5-3-3-3 2T 2-1-2-8-24 at 2.5v vdimm = 5,625MB/s

600mhz 5-3-3-3 2T 2-1-2-8-24 at 2.5v vdimm



Super Pi 1M


Super Pi 32M


Everest Ultimate Bandwidth with 42.9ns latency!


Hexus Pifast


That's fastest I've managed to run Super Pi 32M at 3600mhz cpu speeds on any core2 platform to date (including Asus P5B Deluxe/P5W DH, Intel D975XBX, DFI ICFX3200) !

[eva2000]

Denny I'm there now

Yup further tweaked managed to shave a few seconds off

[eva2000]

Yeah very fast indeed

4000Mhz - 10x400FSB / 3950mhz - 10x395FSB

The Asus 680i Striker Extreme just runs my X6800 hotter than on Asus P5B Deluxe at higher vcore settings ~10-13C hotter at idle.

I tried X6800 @10x400FSB = 4000Mhz at 1.56-1.62v vcore and idle coretemp reported 57-67C versus same X6800 on Asus P5B Deluxe at 4092mhz at 1.65v vcore only reporting coretemps idle at 45-47C.

So only could manage 1M times for X6800 @4000mhz at 1.6v (1.57v load) on Asus 680i Striker Extreme. Higher vcore just meant reboots in windows under load (maybe vdroop too much at higher vcore like on P5B Deluxe ? )

Super Pi 1M


Everest Bandwidth with 42.7ns latency


But at much lower vcore of 1.56v idle/1.54v load, coretemp reports more reasonable idle temps and I can manage 3967-3972mhz at 1.56v idle for 32M with lower mem clocks but once hit 600mhz 5-3-3-x timings seems tight sub timings cause instability until I drop cpu down to 3950mhz at 1.57v idle!

@592.5mhz 5-3-3-5 3-1-6-8-28 2T timings at 2.5v vdimm bios set, Memtest86+ v1.70 reports 5540MB/s memory bandwidth.

Super Pi 1M


Super Pi 32M


Everest Bandwidth with 43.0ns latency


Compared to my old previous max 32M on Asus 680i Striker

[dinos22]

Higher vcore just meant reboots in windows under load (maybe vdroop too much at higher vcore like on P5B Deluxe ? )

i've got the same issue here

[guess2098]

nice! G.J man!
George do you use 2K3 yet?

[eva2000]

i've got the same issue here

hmmm could be vdroop issue then ?

Denny, only use winxp pro sp2 - i've tried win2k3 (trial cd) twice in the past but each time pi times were slower than on winxp see http://www.xtremesystems.org/forums/...ad.php?t=90566 i just don't know.. winxp pro sp2 seems fastest for me

[OneyedK]

Hi, very interesting thread!!! Thanks!!!

I was reading:
http://www.simmtester.com/page/news/...ws.asp?num=139
and dug up:
http://download.micron.com/pdf/datas...r2/1GbDDR2.pdf

So after reading the spd table (SPDtool 1.55, OCZ2TA1000VX21G), I noticed tRP and tRCD are programmed for -25E chips, while tRC is programmed for -3E chips. Now, before I take off the HSP to see what chips are really under the hood, I'd like to know if these timings really matter? I'm using them on P5WDG2 WS Pro (975X) and cannot boot with two sticks after a clear-cmos.
Other timings in the spd-table might differ from the datasheets too, but maybe the spd-table is not so important if you use memset?!

Thanks for any help, hope this is not offtopic, but I didn't want to start a new Titatium Alpha VX2 topic with one like this existing...

[eva2000]

don't think it makes much diff


520mhz 4-3-3-4 2-1-2-6-20 2T at 2.3v

Now to look for a compromise between max memory clocks with high vdimm and more reasonable lower vdimm and suitable memory clocks which have very tight sub timings.

I ended up with 520mhz 4-3-3-4 2-1-2-6-20 2T at 2.3v vdimm bios set. Only ~1.39 seconds slower in 32M Super Pi than 533mhz 4-3-3-4 3-1-3-6-30 2T at 2.35v run but uses 0.05 vdimm less! Interestingly, 520mhz run was faster in 1M Super Pi due to tighter sub timings at 13.938s (520mhz) vs 13.969s (533mhz).

Super Pi 1M


Super Pi 32M


Everest Ultimate Bandwidth

[jozzy]

@ eva2000 : seems like the OCZ PC2-8000 Titanium Alpha VX2 using D9-GMH right?

[eva2000]

not sure as i don't intend to remove the HS at all heh.. either D9GKX or D9GMH

Well 1 second faster at 3600mhz, and also did 3612mhz and 3647mhz runs to see what it takes on 680i striker to hit sub 13min 05s times for 32M

[Blknoel]

I don't think it's GMH, it requires too little voltage to reach those high clocks/tight timings ^^

GMH typically performs at 500Mhz 4-4-4 or 4-4-3 with 2.3v , this is 520+ on 4-3-3 with 2.3v, gotta be high binned GMH at least, or more sensibly, GKX.

Where did you get it and for how much?

[dinos22]

I don't think it's GMH, it requires too little voltage to reach those high clocks/tight timings ^^

GMH typically performs at 500Mhz 4-4-4 or 4-4-3 with 2.3v , this is 520+ on 4-3-3 with 2.3v, gotta be high binned GMH at least, or more sensibly, GKX.

Where did you get it and for how much?

he's on 680i you do realise that

completely different beast to intel chipsets