The Official OCZ VX Thread - Strictly Test Data Only!!!

[conrad.maranan]

Forgive the illustration because I'm in a hurry and my digital camera seems to have conked out on me. Anyway, this is just to give you an idea of what type of cooling I am using for my DIMM region. This crude but simple modification allowed my VX a net increase of 5MHz (from 258MHz to 263MHz), which was stable enough to run many loops of Memtest, Super PI all the way to 32M, and just over 12 hours of Prime Torture Test Blend Priority 10.

If you're looking for a few extra megahertz from your VX, you have a fairly good chance of unveiling them with improved cooling.

Now please post up a rock-solid, stable, and error-free 270MHz in dual channel 2-2-2. I know someone out there has a lucky set.

[HiJon89]

3.51VDIMM measured with a digital multimeter. Any more voltage than this will cause errors. I'm pretty sure there's a little bit more to be gained from the VX modules, but at this point, the CPU is running way too hot. I've seen 265MHz 2-2-2-10 on my 3000+ Winchester, so I know that the VX definitely has potential.

Why don't you try dropping the multiplier to 9?

[conrad.maranan]

Why don't you try dropping the multiplier to 9?

I've tried all of the multipliers already - several times.

[HiJon89]

I've tried all of the multipliers already - several times.

Oh, if you lowered the multiplier and found the highest memory speeds, why do you think they still have some more Mhz left in them?

[conrad.maranan]

The CPU I'm testing with is at a blistering 71C to 72C under load at 10x263MHz. I've done 265MHz in dual channel with my Winchester, which did not run even close to 60C under load.

I'm confident there's more to be realized if I can control the temperatures. If the CPU is hot, so is the memory controller. If the memory controller is running way too hot, the chances are you're not getting the most out of your RAM.

[amilian]

Forgive the illustration because I'm in a hurry and my digital camera seems to have conked out on me. Anyway, this is just to give you an idea of what type of cooling I am using for my DIMM region. This crude but simple modification allowed my VX a net increase of 5MHz (from 258MHz to 263MHz), which was stable enough to run many loops of Memtest, Super PI all the way to 32M, and just over 12 hours of Prime Torture Test Blend Priority 10.

If you're looking for a few extra megahertz from your VX, you have a fairly good chance of unveiling them with improved cooling.

Now please post up a rock-solid, stable, and error-free 270MHz in dual channel 2-2-2. I know someone out there has a lucky set.

Awesome idea. The heat being blown out from the heatsink hits the ram directly, so this wind tunnel concept is a great way to keep that heat off the sticks, and actually deliver cooler air to them!
Thanks for sharing!

[andyOCZ]

I like!! Excellent overclock too!

[Rabbi_NZ]

Awesome idea conrad!
How do you fix the tunnel to the board?


On another note... I've been fiddling with more timings last night and found a few more interesting bits of info...

1) I can run my modules as tight as 1T-1.5-2-0-2-7-15 (CPC-CAS-Trcd-Tras-Trp-Trc-Trfc)... but cant drop my Trfc any further. I usually see people running Trc & Trfc within 2-4 of each other but 15 is the lowest I can get Trfc, whereas Trc goes to the lowest setting of 7 with no problems.
What is everyone else running?
Maybe it's my memcontroller.... these settings are stable up to 260 w/ 3.3vDimm (my max vDimm for now).

2)Tref is a biggy!
I only ran the settings that related to 200MHz (1:1) ratio... but here are my results with the settings I tried (CPU @ 10x256):

3120 : A close 2nd. Sandra bandwidth = 397x/397x , SuperPI 1MB = 33s-34s
0064 : No POST
1560 : 3rd best. Sandra bandwidth = 393x/393x , SuperPI 1MB = 34s
0032 : Horrendous Bandwidth!!! Sandra bandwidth = 500/500 , SuperPI 1MB = 1:02 m
0780 : 4th best. Sandra bandwidth = 381x/381x , SuperPI 1MB = 34s
0016 : No POST
3072 : Best for me. Sandra bandwidth = 397x/397x , SuperPI 1MB = 33s
0128 : Sandra bandwidth = 252x/252x , SuperPI 1MB = 35s

[trans am]

get that camera working soon, this I gotta see.

[Rabbi_NZ]

Good job, Alex. What's your maximum Super PI 32M stable with just one stick?

Didn't have time to do many 32mb runs or fiddle with multis today but here's my experience with SuperPI 1MB with my one better stick (does bout 1-2MHz better than the other)... you're hosting these so it's old news! but act interested hehehe


This will give you a good idea of how much speed we need to gain to lower SuperPI with Socket754+OCZ VX.... not sure bout 939...

Thanks again to Conrad for his generous hosting

10x270 1.5-2-2-0-7-15 SuperPI 1x512 OCZ VX
10x269 1.5-2-2-0-7-15 SuperPI 1x512 OCZ VX
10x267 1.5-2-2-0-7-15 SuperPI 1x512 OCZ VX
10x265 1.5-2-2-0-7-15 SuperPI 1x512 OCZ VX
10x263 1.5-2-2-0-7-15 SuperPI 1x512 OCZ VX

[Edit] I have a feeling 10x271 would give occasional 31s and 10x272 would give consistent 31s... but my CPU is getting maxed there.
I ran 10x271 once and got 32s but it was close... and then it rebooted... I think more VCore would stabilise it enough for more runs tho.

[dutchman.pt]

get that camera working soon, this I gotta see.

Make that 2.

Show us that crazy wind tunnel that you got there, conrad.

[cinders]

Awesome idea. The heat being blown out from the heatsink hits the ram directly, so this wind tunnel concept is a great way to keep that heat off the sticks, and actually deliver cooler air to them!
Thanks for sharing!

Yes, awesome idea....

[cinders]

There's some pics of mine on this thread

http://www.xtremesystems.org/forums/...ad.php?t=47413

post #19

mine sits at 35 degrees C when run constantly at 3.4v

without it they hit 60 before i hit the switch....

[conrad.maranan]

you're hosting these so it's old news! but act interested hehehe

Show me stability all the way to 32M at those speeds and you'll have my 100% undivided attention, Alex.

I've been very busy with social obligations this weekend, but I managed to purchase a few sheets of acrylic from Home Depot this afternoon. I'll build a much more refined wind tunnel and will resume testing after completion. Pictures will follow as well.

Have any of the OCZ employees hit 270MHz or greater in dual channel yet? Please let us know where you guys are currently at and if you feel that your memory is still climbing after time.

[Rabbi_NZ]

Show me stability all the way to 32M at those speeds and you'll have my 100% undivided attention, Alex.

hahahahaha I love it! Any other forum would have people offering me their first born after seeing those screenies... here I get challenged for more... that's why this place is so great.

Will try for the highest 1:1 SuperPI 32MB but at 3.3v it really is difficult to stabilise these sticks past about 260.

What timings you running Conrad? I've done a fair bit of testing with regard to timings lately... wouldn't mind comparing some info

[DevilsRejection]

you think i could get 240 @ 2-2-2-10 1T at 3.2V or less?

[TEDY]

if you ask me TCCD much better then this VX....same price

VX+BOOSTER = 280$

TCCD = from 221 to 350$...depends

[conrad.maranan]

http://xtremesystems.org/forums/show...312#post629312

hahahahaha I love it! Any other forum would have people offering me their first born after seeing those screenies... here I get challenged for more... that's why this place is so great.

Will try for the highest 1:1 SuperPI 32MB but at 3.3v it really is difficult to stabilise these sticks past about 260.

What timings you running Conrad? I've done a fair bit of testing with regard to timings lately... wouldn't mind comparing some info

The timings I used for 10x263MHz 1:1 1T 2-2-2-10 are all listed here. This is 100% rock-solid and Prime Torture Test stable. I can run any CPU and memory intensive application without incident.

you think i could get 240 @ 2-2-2-10 1T at 3.2V or less?

While waiting for the replacement for my DDR Booster last week, I was able to do 230MHz 100% stable at 2.85VDIMM. Based on what I was able to do, I will guess that 240MHz 2-2-2-10 1T at 3.2VDIMM is highly probable. Of course, there's only one way for you to find out.

if you ask me TCCD much better then this VX....same price

VX+BOOSTER = 280$

TCCD = from 221 to 350$...depends

I guess it all boils down to what you want: high frequencies at 2.5-3-3-x timings or mid-high frequencies at 2-2-2-x timings.

[Rabbi_NZ]

In this order:
CPC-Tcl-Trp-Trcd-Tras-Trc-Trfc-Trrd-Twr-Twtr-Trtw-Tref-Twcl-Drive Strength

Yours Conrad:
1T-2.0-2-2-10-10-15-2-2-2-1-0780?(15.6uS)-?-?
Mine:
1T-1.5-2-2-00-07-15-0-3-1-2-3072 (03.9uS)-1-4

I'm not sure bout your Tref "value" but I think that is it... I got it from here

Those are my 260 3.3v timings, 100% stable... but for more MHz I think I really need more voltage.

How low can you get your Trfc dude? Mine is seriously unstable below 15, but strange thing is Trc will go down to 7 no problem.

I did a write-up here @ DFI-Street about my timings and what I've found to be the best so far... I think you'll find some of it interesting

[andyOCZ]

I just wanted to make sure that we all are up on the subject of some A64 90nm CPU's earlier than week 41 failing Prime95.

The Official thread is here: http://www.xtremesystems.org/forums/...962#post631962

I won't go into details here, but please if you are using Prime95 to test while running an A64 Winchester 90nm take a minute to read this. Many CPU's fail at stock settings. Always test at BIOS defaults for 7+ hours before using Prime95.

Thanks

[cuddles]

260 Mhz @ 1.5-2-2-0..... *faints*

[lalPOOO]

Anyone have results from an intel chipset? (875) I'd really like to pick up a set of these...

[conrad.maranan]

The DFI 754 board definitely has more options as far as memory timings are concerned. This is one thing that I truly miss most about that board. I do not know what my Twcl value is at, but Drive Strength is at Weak in BIOS. The only two options available on the MSI K8N Neo2 Platinum are Normal and Weak.

Anyway, the values I have on that screenshot are the absolute tightest timings that I can run 100% stable. I can slacken up the numbers a bit to gain 1MHz or 2MHz, but I am not willing to sacrifice performance. Setting CAS Latency to 1.5 yielded no increase in Sandra bandwidth, no decrease in Everest latency, and no decrease in Super PI 32M when I tested.

I finished the rough version of my wind tunnel this morning. I'll cut up three more pieces of acrylic tomorrow and start on the final version. I have other projects that need tending to, so it'll probably be another week before I have everything complete and ready to install.

Cheers!

[ed.howell]

Anyone have results from an intel chipset? (875) I'd really like to pick up a set of these...

I would like to know the same thing??

[Orker]

Anyone have results from an intel chipset? (875) I'd really like to pick up a set of these...

Yeh first page of this thread.. eva2000 was testing with intel