OCZ 8500 Low-Voltage Blade

[Sayno]

I tried the 4:5 divider only once so far and I couldn't even get to desktop, but maybe it was the BIOS. Is 4:5 a strong divider on the P45? I thought 5:6 was optimal.

Can't seem to boot at PL7 with anything lol - still have a lot to learn!

1:1 is best if your mobo can get the FSB high enough.

I've broke 10k read and 11k write and copy with ram at 1080 1:1. 24/7 stable I might add.

This is just cheap $35 micro center special stuff though

http://img268.imageshack.us/img268/7066/10gbz.jpg

[BeepBeep2]

That's crazy memory bandwidth from an older cpu/mobo.

Well, it has it's advantages and weaknesses. Most of these 590 SLI boards won't even get past 1000. (Of course, without Xtreme tweaking.) When I popped these in for the first time I could not even boot into Windows 7 64-bit @ 6-6-6-18 1040 Mhz 1.85v.

Gotta love the latency though...nearly half of intel platforms and just the same or more bandwidth at 2T.

[-n7-]

OCZ2P1150LV4GK arrived!

w00!

It's been far too long since n7 last had new RAM to play with.

New babies!


Lined up with my Mushkin.


I'm too scared to ruin the fun by putting in all 8 GB at once, so i'm playing with kit #1 for now.
Subtimings aren't tweaked, so if i was worried about numbers, i could get Everest a bit higher. Not the concern for now though.

DDR2-1200 5-5-5-15 1.8v (tRD 8) good for 5 passes of LinX.

[BeepBeep2]

Ooh...same kits I just started playing with. Of course I've got 1125 MB/s copy 10868 MB/s read and I'm only @ 1100 5-5-5-12 1.81v
To me, it's the bandwidth that counts...too bad, my CPU and motherboard are limiting. I also prefer these heatspreaders over the blade ones. I like shiny.

Try PL7 and PL6 when you get the chance......I'm pretty sure this RAM can do a lot more than 1200 and was only stable at my motherboard's lowest drive strength settings.

[Don_Dan]

Specs in sig, I'm currently doing 1100 5-5-5-12 2T tRC 17 @ 1.81v

System will not post @ 1100 5-5-5-10 2T tRC 15 @ 1.83v... Somehow completely stable at the above though.

Your tRC is too tight in my opinion. Do you have a possibility to change tRFC?
If not try a higher tRC ~30, as by JEDEC spec tRFC >= 2 * tRC.
These kits need high tRFC to reach high clockspeeds.

OCZ2P1150LV4GK arrived!

It's been far too long since n7 last had new RAM to play with.

So you finally got some too....
Looking forward to more results!

[LennyRhys]

1:1 is best if your mobo can get the FSB high enough.

I've broke 10k read and 11k write and copy with ram at 1080 1:1. 24/7 stable I might add.

High mhz/lower fsb is much better than 1:1 on P45 - here's 4:5 at 525 FSB vs your 1:1 at 540 fsb

(courtesy of radja)

DDR2-1200 5-5-5-15 1.8v (tRD 8) good for 5 passes of LinX.

So not fair... I'm guessing the plats aren't binned as tightly as the blades, because my 9200s need more than EVP 1.85v for 1200mhz.

Very nicely done btw.

[saxovtsmike]

you´ve got to love those P45 Boards, 500+ fsb on 45nm quads, thats the dreamland far far away for most X48 users.
Best I got out of my Blades was nealry 9.6k rad, 9.4k write and 9.6k copy with 450 fsb and 600 ram. Sadly that´s with PL6 which is non by far stable.

[BeepBeep2]

Your tRC is too tight in my opinion. Do you have a possibility to change tRFC?
If not try a higher tRC ~30, as by JEDEC spec tRFC >= 2 * tRC.
These kits need high tRFC to reach high clockspeeds.



So you finally got some too....
Looking forward to more results!

Nope, you can't change tRFC on AMD. And I gain a good 150 MB/s going from 26 tRC (Auto and maximum) to 17. It's a good trick if the ram runs stable. My main issue with clocking was the drive strength. They were set on auto, And going down it was like: 1.50x 1.50x 1.50x 1.50x 1.0x 1.0x

Now they are all 1.0x so I had to lower 4 of them about 50% to get them to run correctly.

[-n7-]

So not fair... I'm guessing the plats aren't binned as tightly as the blades, because my 9200s need more than EVP 1.85v for 1200mhz.

Very nicely done btw.

My board is generally known to overvolt on vDIMM by about 0.08v.

So 1.8v is likely around 1.88v.

Below is pretty much the highest i could get with the first 2x2 GB.
I imagine i could get higher with more vDIMM, but i'm not going over 1.8v for these.

DDR2-1230 5-5-5-18 loose subtimings (not stable)


I actually had better luck on the 266 strap (2:3) than on 333 (5:8), which was weird. On 5:8, i couldn't even get into the OS @ DDR2-1230 5-5-5.

I tried 6-6-6 as well, but it did not get me anywhere.
Couldn't even boot into the OS w/o missing files or BSODs @ DDR2-1250.

So no scaling with CAS 6 at all
Really too bad...

DDR2-1215 5-5-5 i could still get thru 5 passes of LinX, but that was about it.


Here's as good as it's going to get for me in Everest:



I've now put the second kit in for 8 GB.
Headache has begun with 8 GB, as usual.

With 8 GB, just in Memtest86+ alone, DDR2-1160 or higher is not stable

So as i'm not willing to go over 1.8v for now, my dreams of DDR2-1200 8 GB are pretty crushed.

Trying to get things stable with 8 GB @ 4:5 at this point.

9x456 1.24375v (4104 MHz) LLC enabled
4:5 266 strap (DDR2-1140 5-5-5-18-60) Clock Twister @ Lighter tRD @ 8

HCI 4x1400 MB 300% stable & LinX max RAM 25 passes stable.
Course, after these two, i got a BSOD just while surfing the web...really hoping that was because i just need more NB voltage or vcore (LLC bumps vcore up quite a bit under load, so it's actually a lot lower idle).

[Sayno]

High mhz/lower fsb is much better than 1:1 on P45 - here's 4:5 at 525 FSB vs your 1:1 at 540 fsb

Interesting, I will give that a try tonight. Any Idea what his PL was? My screenshot is PL9. The latency on his ram is quite impressive to say the least. I'm not sure if I will be able to get this ram running at 1300mhz though. . . its only 800mhz rated.

[Hoss331]

I tried 6-6-6 as well, but it did not get me anywhere.
Couldn't even boot into the OS w/o missing files or BSODs @ DDR2-1250.

So no scaling with CAS 6 at all
Really too bad...

Try 5-6-6, I didnt have much luck with 6-6-6 either.

[18 is # 1]

Try 5-6-6, I didnt have much luck with 6-6-6 either.

Me also too as well ditto.

[LennyRhys]

I didnt have much luck with 6-6-6 either

That's because 6-6-6 is evil

[18 is # 1]

1066 LV Platinums are on sale at Amazon for $42 shipped! http://www.amazon.com/OCZ-PC2-8500-P...7584970&sr=1-2
Very tempting.

[Don_Dan]

Nope, you can't change tRFC on AMD. And I gain a good 150 MB/s going from 26 tRC (Auto and maximum) to 17. It's a good trick if the ram runs stable. My main issue with clocking was the drive strength. They were set on auto, And going down it was like: 1.50x 1.50x 1.50x 1.50x 1.0x 1.0x

Now they are all 1.0x so I had to lower 4 of them about 50% to get them to run correctly.

tRC was good to generate bandwidth on 680i too where you didn't have tRFC either, but can you get a higher overclock at a higher tRC?

I actually had better luck on the 266 strap (2:3) than on 333 (5:8), which was weird. On 5:8, i couldn't even get into the OS @ DDR2-1230 5-5-5.

I had big problems on 5:8 divider too, very weird. I couldn't boot at anything over DDR2-1160, though almost DDR2-1200 was stable in Windows when I setFSB'd to that speed.

[Ket]

I might get a kit of these when I'm done testing my aXeRAM. I want a 4GB kit that can do 1200+MHz.

[-n7-]

All last night i was fighting, trying to get 8 GB higher.

In Memtest86+ test 5 quick checks, i was able to get past DDR2-1140, but the problem is Memtest86+ getting errors doesn't mean things are actually stable when it comes to 8 GB.

4 GB is seriously such a piece of cake for overclocking LOL.

8 GB...such a different annoying world, at least if you are really testing for stability.

So yeah, in Memtest86+, i had DDR2-1173 5-6-5-18-70 get thru a few test 5 runs.
DDR2-1187 5-6-5-18-70 1.82v also.

But it wasn't stable in the OS, & having to use 400 strap (3:4) puts out worse Everest numbers than 266 strap (4:5), so running say 9x450 DDR2-1200 is hardly any faster than 9x456 DDR2-1140, not to mention i cannot get as high with CPU using 3:4 either.

9x450 3:4 (DDR2-1200) - this i cannot do with 8 GB, but it shows the numbers.


And here's what i'm currently running
9x456 4:5 (DDR2-1140)


Faster CPU speed more than makes up for a tiny loss in Everest numbers.

And it's doable stably.


So in short, n7 is settling with 4:5 operation, which means DDR2-1112 to 1140 range depending on whether i sit on 4 or 4.1 GHz.

[OC Nub]

Anything over 1200mhz seems to be very hard with this set. Using 4:5 divider pl8 on the UD3P board.



n7: Nice vcore BTW for 4.1ghz.

[18 is # 1]

Anything over 1200mhz seems to be very hard with this set. Using 4:5 divider pl8 on the UD3P board.
:

That's just about what my 1066 Blades did.

[MongGrel]

Me too, at least more.

Egg and their silly you can only buy freaking 4G at a time stuff

[xguntherc]

So what is the most important score on the everest test, the copy?
I don't have the blades (I'm getting them soon) but for now, I have some decent scores on my stuff at stock 1066 Mhz, so I was wondering what I need to change. In everest I got

Read: 10940 Write: 9424 and Copy: 8069.. everyone else in this thread have closer numbers. like all in the mid or high 9's. why is my read so much better than copy, how can I even that out. I'm running 1066 Mhz 5-5-5-15

[Don_Dan]

Read: 10940 Write: 9424 and Copy: 8069.. everyone else in this thread have closer numbers. like all in the mid or high 9's. why is my read so much better than copy, how can I even that out.

You can't, your read bandwidth will always be much better than your copy bandwidth. Please post a screenshot including MemSet.

PM incoming in 5 minutes, didn't forget you, was very busy on the weekend.

[xguntherc]

I just tried to run memset 4.0 and I got some error about that I must use Normal Size 96ppp in this software, please go to office>properties> change to normal or some junk. whats that about. lol

[-n7-]

So what is the most important score on the everest test, the copy?
I don't have the blades (I'm getting them soon) but for now, I have some decent scores on my stuff at stock 1066 Mhz, so I was wondering what I need to change. In everest I got

Read: 10940 Write: 9424 and Copy: 8069.. everyone else in this thread have closer numbers. like all in the mid or high 9's. why is my read so much better than copy, how can I even that out. I'm running 1066 Mhz 5-5-5-15

Write/copy are largely FSB based.
Intel chipsets do better for write/copy.

nForce does well with read if you use 1T, & latency will be great as well.

[xguntherc]

well I was able to get over 11K on read, and 9600 on write, and just under 9k on copy with a few timings changes.