• Welcome to Overclockers Forums! Join us to reply in threads, receive reduced ads, and to customize your site experience!

2GB OCZ PC2-8000 Titanium Alpha VX2 4-4-4-15 [56k users beware]

Overclockers is supported by our readers. When you click a link to make a purchase, we may earn a commission. Learn More.

eva2000

Member
Joined
Aug 5, 2002
Location
Brisbane, Australia
These Micron D9GKX based, OCZ PC2-8000 Titanium Alpha VX2 memory modules warrantied at 2.4v + 5% (2.52v vdimm) are so hard to find these past few months. But OCZ was kind enough to send me a pair to play with. Thanks Tony đź‘Ť

titanium_alphavx2_001.jpg


titanium_alphavx2_028.jpg


titanium_alphavx2_005.jpg


titanium_alphavx2_025.jpg


Ordered a Asus P5B Deluxe/Wifi motherboard just waiting for it to arrive so can play with my Intel Core 2 Quad Q6600 Kentsfield cpu properly as well.

Should be a nice combination once my Hipro Maximizer DDR2 PCB arrives as well :D

HiproMaximizer_001.jpg


DDR1 PCB..
HiproMaximizer_012.jpg


HiproMaximizer_014.jpg
 
2GB OCZ PC2-8000 Titanium Alpha VX2 Micron D9GK

System:

Re-ran some of the single Super Pi v1.5 32M Pi times with stock @2400Mhz 9x266fsb with 2:3 divider and OCZ Titanium AlphaVX2 @400mhz 3-3-3-8 at 2.29v vdimm.



Super Pi 1M & 32M




(Due to higher async upclocked memory speed and tighter timings of 3-3-3-8, the 32M time is 98+ seconds faster than 1:1 266mhz 4-4-4-8 timings!)

Super Pi 8M & 16M
superpi-8m_single.png
superpi-16m_single.png


(Due to higher async upclocked memory speed and tighter timings of 3-3-3-8, the 16M time is ~45 seconds faster than 1:1 266mhz 4-4-4-8 timings!)

Dual 32M


(Due to higher async upclocked memory speed and tighter timings of 3-3-3-8, the dual 32M Pi time is ~150 seconds faster than 1:1 266mhz 4-4-4-8 timings!)
 
Swapped out my Intel Core 2 Quad Q6600 for now and plucked in my Intel Core 2 Duo E6600 Conroe cpu so I can properly test this 2x1GB OCZ PC2-8000 Titanium Alpha VX2 Micron D9GKX memory :)

titanium_alphavx2.jpg

System:

What better way to see how fast this 2x1GB OCZ PC2-8000 Titanium Alpha VX2 Micron D9GKX is than by benching with Super Pi v1.5 :)

Let's start with 1:1 divider and 3600Mhz 9x400fsb with 400mhz memory clock and see.

2GB Gskill PC6400PHU2 HZ Micron D9GMH 0629 batch​

First a comparison of my fastest time with 2GB Gskill PC6400PHU2 HZ Micron D9GMH 0629 batch 9x400fsb 1:1 with memory at 400mhz 4-4-3-5 with tightest memset tweaked timings

on Intel D975XBX Bad Axe rev304 with 2GB Gskill PC6400PHU2 HZ Micron D9GMH 0629 batch and memset tweaked to tightest possible timings
superpi-32m_single_14m15s719ms.png


even with E6700 @9x360fsb 4:5 with 2GB Gskill PC6400PHU2 HZ Micron D9GMH 0629 batch @450mhz 4-4-3-4 only managed 14m 05.406s
superpi-32m_single_14m05s406ms_memset.png


2x1GB OCZ PC2-8000 Titanium Alpha VX2 Micron D9GKX

So what can 2x1GB OCZ PC2-8000 Titanium Alpha VX2 Micron D9GKX do on same motherboard, Intel D975XBX Bad Axe rev304 ?

how about 9x400fsb 1:1 with memory @400mhz 3-3-3-8 at 2.29v :)

untweak straight boot from bios
superpi-32m_single_tnn.png


straight boot from bios + memset v3.0 tweaked
superpi-32m_single_13m59s437ms_tnn.png


Sub 14min for the first time with 1:1 divider and just 9x400fsb! :cool:

So OCZ PC2-8000 Titanium Alpha VX2 with tightest 1:1 400mhz timings of 3-3-3-8 is 16.282 seconds faster than Gskill PC6400PHU2 HZ with tightest 1:1 400mhz timings of 4-4-3-5.


.
 
Now to see how high this memory can go at looser timings. Seems 550mhz 4-4-4-12 at 2.39v with fairly tight memset timings for tRFC is at least Super Pi single and dual 32M stable. Clockgen was used for the tests as well.

click on images for full screen shots



Single 32M


Dual 32M
half way mark
superpi-32m_halfway_tn.png


end result


@560mhz 5-4-4-8 at 2.39v​

Change to CAS 5 and loosened tRFC from 28 to 38 while rest of default memset v3.0 timings remain the same.





Max Super Pi 1M @583mhz 5-5-5-15 at 2.39v​



Max CPUZ Validation @585mhz 5-5-5-15 at 2.39v​



With 2.39v vdimm via pencil mod on Intel D975XBX Bad Axe rev304 with 2:3 divider on 1067fsb strap, this set of 2x1GB OCZ PC2-8000 Titanium Alpha VX2 Micron D9GKX managed
  • Max single and dual 32M Super Pi @550mhz 4-4-4-12 at 2.39v = DDR2 1100Mhz = PC2-8800
  • Max single 32M Super Pi @560mhz 5-4-4-8 at 2.39v = DDR2 1120Mhz = PC2-8960
  • Max single 1M Super Pi @583mhz 5-5-5-15 at 2.39v = DDR2 1166Mhz = PC2-9328
  • Max CPUZ Validation @585mhz 5-5-5-15 at 2.39v = DDR2 1170Mhz = PC2-9360

I'm sure I can get alot higher mem clocks with the Asus P5B Deluxe I ordered for delivery next week :)


.
 
Last edited:
You've got some nice equipment there with some very impressive results. Can't wait to see what they can do on the P5B Deluxe, good luck!
 
thanks guys

AFAIK, Gskill PC8000 HZ and OCZ PC8000 Titanium Alpha VX2 both use Micron D9GKX modules and probably get the same clocks just OCZ warrantied for 2.52v vdimm max while Gskill warrantied for 2.3v vdimm max. I haven't tested the Gskill PC8000 HZ so cant say personally but OCZ ones from testing i done so far are sweet :)
 
Nice eva! I've done a lot of reading about this RAM since I read your thread and I might just have to get some. Impressive stuff.
 
Why don't people ever do real stability testing instead of just SPi? I know SPi is easy and it might be all you care about but it's misleading to people who want real stability to just show SPi stability when the differences between that and real stability can be large, there's no way to judge what the 24/7 stability is. Running memtest or orthos when it's time to sleep is easy too. ;)
 
Last edited:
MadMan007 said:
Why don't people ever do real stability testing instead of just SPi? I know SPi is easy and it might be all you care about but it's misleading to people who want real stability to just show SPi stability when the differences between that and real stability can be large, there's no way to judge what the 24/7 stability is. Running memtest or orthos when it's time to sleep is easy too. ;)
super pi is used to compare performance of memory not as a absolute stability in above tests... at same clock speed and timings, there's no way to differentiate between different brands of memory with memtest and/or prime.

Also find me 3 folks on this forum that run 2.39v vdimm 24/7 on their DDR2 memory... someone show me screenies for anyone who's done 7 day 24/7 (168 hrs) prime large ffts at 2.39+ vdimm on their DDR2 memory :)
 
I understand what SPi is used for, and from what you say it's 'to differentiate different brands of memory at the same speed and timing.' But looking at the 'final report' in the first post quoted below just shows clocks and voltage, which certainly makes it look like you use SPi as a stability tester. All I'm saying is why not let a real memory test run when you're not using the computer to give some results that are useful for actual computer use and give a better quality of memory review.

eva2000 said:
With 2.39v vdimm via pencil mod on Intel D975XBX Bad Axe rev304 with 2:3 divider on 1067fsb strap, this set of 2x1GB OCZ PC2-8000 Titanium Alpha VX2 Micron D9GKX managed
  • Max single and dual 32M Super Pi @550mhz 4-4-4-12 at 2.39v = DDR2 1100Mhz = PC2-8800
  • Max single 32M Super Pi @560mhz 5-4-4-8 at 2.39v = DDR2 1120Mhz = PC2-8960
  • Max single 1M Super Pi @583mhz 5-5-5-15 at 2.39v = DDR2 1166Mhz = PC2-9328
  • Max CPUZ Validation @585mhz 5-5-5-15 at 2.39v = DDR2 1170Mhz = PC2-9360

I'm sure I can get alot higher mem clocks with the Asus P5B Deluxe I ordered for delivery next week :)

.

I don't know what the point of your second statement is. Do you test your memory for that long? or..? I'm not saying 7 days is needed but there's definitely a difference between SPi stable and real stability.

Don't get me wrong I appreciate the memory review I just think it could be more complete and informative with some additional testing.
 
Last edited:
It normally takes me around 1-3 weeks to properly test memory including 72+ hr memtest/prime testing but due to time constraints i prefer shorter tests...

and yes i'm focused on super pi as i want to better my world ranking which i'm sitting at #13 http://www.hwbot.org//hallOfFame.do?type=result&applicationId=7 and #10 right now http://holicho.lib.net/pi3355/pi3355.htm

yes there's a diff between real stability and super pi... but that's for another day

FYI, depending on ram and board i've had dual 32M passable memory clocks that can run 18+hrs of video encoding without problems.

It's hard to please everyone.. folks complain when folks show only 1M pi, saying run 32M, then complain say run dual 32M, then complain and say run prime, then complain and say prime wasn't run long enough heh

How long is long enough differs for most folks - i run memtest for over 100hrs sometimes before even loading windows LOL :)

Some very old screenies with DDR1
memtest_002.jpg


memtest_004.jpg


memtest_005.jpg
 
Yup Steve.. lovely sticks.. setting up my P5B Deluxe right now with these OCZ Titanium Alpha VX2.. 7x500fsb 1:1 500mhz 4-4-4-12 straight boot from bios hehe
 
Changed motherboards for some stressprime action too

System:

Testing the memory with the following fixed settings with 2:3 divider and used Clockgen for the more incremental values sometimes:

PCI Express Frequency: 100
PCI Clock Synchronization Mode: 33.33Mhz
Spread Spectrum: Disabled
CPU Vcore Voltage: 1.4375 (1.384) - set higher than needed for test purposes
FSB Termination Voltage: 1.3v
NB Vcore: 1.45v
SB Vcore (SATA/PCIE): AUTO
ICH Chipset Voltage: AUTO

Memory Remap Feature: Disabled
Configure DRAM Timing by SPD: Disabled
DDRAM CAS# Latency: 4
DRAM RAS# to CAS# Delay: 4
DRAM RAS# Precharge: 4
DRAM RAS# Activate to Precharge: 12
DRAM Write Recovery Time: 6
DRAM TRFC: 42
DRAM TRRD: 10
Rank Write to Read Delay: 10
Read to Precharge Delay: 10
Write to Precharge Delay: 10
Static Read Control: Disabled

Modify Ratio Support: Enabled
Ratio CMOS Setting: 9x
CIE Support: Disabled
Max CPUID Value Limit: Disabled
Vanderpool Technology: Disabled
CPU TM function: Disabled
Execute Disable Bit: Disabled
PECI: Disabled

Max Stressprime Orthos Large FFTs (~1.3GB memory tested)

click images for full screenshots

@551mhz 4-4-4-12 at 2.45v bios (AI Suite reports 2.40v)


Stressprime Orthos Large FFTs


Single 32M​


Dual 32M​
 
Max Super Pi 1M @5-4-4-12 at 2.45v
Loosened CAS timing from 4-4-4-12 to 5-4-4-12 at 2.45v with 1.65v NB voltage and managed to break DDR2-1200Mhz barrier :)

PCI Express Frequency: 100
PCI Clock Synchronization Mode: 33.33Mhz
Spread Spectrum: Disabled
CPU Vcore Voltage: 1.5 (1.448) - set higher than needed for test purposes
Memory Voltage: 2.45v
FSB Termination Voltage: 1.3v
NB Vcore: 1.65v
SB Vcore (SATA/PCIE): AUTO
ICH Chipset Voltage: AUTO

Memory Remap Feature: Disabled
Configure DRAM Timing by SPD: Disabled
DDRAM CAS# Latency: 5
DRAM RAS# to CAS# Delay: 4
DRAM RAS# Precharge: 4
DRAM RAS# Activate to Precharge: 12
DRAM Write Recovery Time: 6
DRAM TRFC: 42
DRAM TRRD: 10
Rank Write to Read Delay: 10
Read to Precharge Delay: 10
Write to Precharge Delay: 10
Static Read Control: Disabled

Modify Ratio Support: Enabled
Ratio CMOS Setting: 9x
CIE Support: Disabled
Max CPUID Value Limit: Disabled
Vanderpool Technology: Disabled
CPU TM function: Disabled
Execute Disable Bit: Disabled
PECI: Disabled

@600Mhz 5-4-4-12 at 2.45v​





@605Mhz 5-4-4-12 at 2.45v​



 
Async Memory Overclocking 4:5

Now to test out 4:5 cpu/memory divider on Asus P5B Deluxe. As I have read, 4:5 divider on Asus P5B Deluxe is alot harder to use and will not overclock memory as high as the max achievable on 2:3 divider. Seems part of the key is NB/vMCH voltage which needs to be raised for stability on 4:5 divider. At 400FSB 4:5 500Mhz memory from bios, I still needed to raise NB volts from 1.45v to 1.55v in order to eliminate 1000s of errors in memtest86+ v1.65's Test #4.

  • Raising NB volts to 1.55v allowed me to hit 420FSB 4:5 525mhz 4-4-4-12 at 2.45v vdimm.
  • Raising NB volts from 1.55v to 1.65v allowed me to hit 432FSB 4:5 540mhz 4-4-4-12 at 2.45v vdimm straight boot from bios.
Then I tried booting from bios at 400FSB 4:5 500mhz 4-4-4-12 at 2.45v and using clockgen to raise FSB from 400FSB to 432FSB, but at around 421FSB I got BSOD from within windows while doing Systool dual 32M Pi test. I rebooted and tried again with exact same BSOD.

So I decided to change memory timings from 4-4-4-12 to 5-4-4-12 and boot into windows at 400FSB 4:5 500mhz 5-4-4-12 at 2.45v and managed to clockgen up to only 426FSB 4:5 533mhz 5-4-4-12 at 2.45v.

The end result is 4:5 divider tests results with two configurations:
  1. Config #1 = Straight bios set boot at 9x432FSB = 3888Mhz with 4:5 memory at 540mhz 4-4-4-12 at 2.45v
  2. Config #2 = Boot from bios set at 9x400FSB = 3600Mhz with 4:5 memory at 500mhz 5-4-4-12 at 2.45v which clockgen to 9x426FSB = 3839Mhz with 4:5 memory at 533mhz 5-4-4-12 at 2.45v. Taking advantage of the faster NB strap at =<400FSB.
Config #1
PCI Express Frequency: 100
PCI Clock Synchronization Mode: 33.33Mhz
Spread Spectrum: Disabled
Memory Voltage: 2.45v
CPU Vcore Voltage: 1.625 (1.552)
FSB Termination Voltage: 1.3v
NB Vcore: 1.65v
SB Vcore (SATA/PCIE): AUTO
ICH Chipset Voltage: AUTO

Memory Remap Feature: Disabled
Configure DRAM Timing by SPD: Disabled
DDRAM CAS# Latency: 4
DRAM RAS# to CAS# Delay: 4
DRAM RAS# Precharge: 4
DRAM RAS# Activate to Precharge: 12
DRAM Write Recovery Time: 6
DRAM TRFC: 42
DRAM TRRD: 10
Rank Write to Read Delay: 10
Read to Precharge Delay: 10
Write to Precharge Delay: 10
Static Read Control: Disabled

Modify Ratio Support: Enabled
Ratio CMOS Setting: 9x
CIE Support: Disabled
Max CPUID Value Limit: Disabled
Vanderpool Technology: Disabled
CPU TM function: Disabled
Execute Disable Bit: Disabled
PECI: Disabled
Config #2
PCI Express Frequency: 100
PCI Clock Synchronization Mode: 33.33Mhz
Spread Spectrum: Disabled
Memory Voltage: 2.45v
CPU Vcore Voltage: 1.625 (1.552)
FSB Termination Voltage: 1.3v
NB Vcore: 1.65v
SB Vcore (SATA/PCIE): AUTO
ICH Chipset Voltage: AUTO

Memory Remap Feature: Disabled
Configure DRAM Timing by SPD: Disabled
DDRAM CAS# Latency: 5
DRAM RAS# to CAS# Delay: 4
DRAM RAS# Precharge: 4
DRAM RAS# Activate to Precharge: 12
DRAM Write Recovery Time: 6
DRAM TRFC: 42
DRAM TRRD: 10
Rank Write to Read Delay: 10
Read to Precharge Delay: 10
Write to Precharge Delay: 10
Static Read Control: Disabled

Modify Ratio Support: Enabled
Ratio CMOS Setting: 9x
CIE Support: Disabled
Max CPUID Value Limit: Disabled
Vanderpool Technology: Disabled
CPU TM function: Disabled
Execute Disable Bit: Disabled
PECI: Disabled

Results:
  • Super Pi single and dual 32M was 27 seconds and ~1min faster respectively with clockgen raised 9x426FSB 4:5 533mhz 5-4-4-12 configuration. And seems the trend continues with other bandwidth intensive/specific benchmarks.
  • Everest latency was 65.5ns vs 45.5ns
  • Mbench Access latency was 58.9ns vs 49.2ns
  • Sandra XI Random Access Latency was 66ns vs 60ns

==================================================​
Super Pi single 32M
3888Mhz - 9x432FSB
4:5
540mhz 4-4-4-12 at 2.45v


Super Pi single 32M
3839Mhz - 9x426FSB
4:5
533mhz 5-4-4-12 at 2.45v
(clockgen from 400FSB)


==================================================​
Super Pi dual 32M
3888Mhz - 9x432FSB
4:5
540mhz 4-4-4-12 at 2.45v


Super Pi dual 32M
3839Mhz - 9x426FSB
4:5
533mhz 5-4-4-12 at 2.45v
(clockgen from 400FSB)


==================================================​
Everest Ultimate v3.50 Bandwidth
3888Mhz - 9x432FSB
4:5
540mhz 4-4-4-12 at 2.45v
everest-bandwidth.jpg


Everest Ultimate v3.50 Bandwidth
3839Mhz - 9x426FSB
4:5
533mhz 5-4-4-12 at 2.45v
(clockgen from 400FSB)
everest-bandwidth.jpg


==================================================​
MBench v1.0
3888Mhz - 9x432FSB
4:5
540mhz 4-4-4-12 at 2.45v
Code:
---- MBench (C) ver 1.0 (Feb 2002) ----
---- System memory benchmark ----
---- www.x86-secret.com ----

Intel P6 processor (CPUID = 6f5) @ 3888.1MHz
Instruction set support : MMX SSE SSE2 

Access latency			 58.9 ns (229 clocks)

Read datarate (INT)		7274 Mb/s
Write datarate (INT)		2986 Mb/s

Read datarate (MMX)		7898 Mb/s
Write datarate (MMX)		2935 Mb/s

Read datarate (SSE)		8181 Mb/s
Write datarate (SSE)		7877 Mb/s

MBench v1.0
3839Mhz - 9x426FSB
4:5
533mhz 5-4-4-12 at 2.45v
(clockgen from 400FSB)
Code:
---- MBench (C) ver 1.0 (Feb 2002) ----
---- System memory benchmark ----
---- www.x86-secret.com ----

Intel P6 processor (CPUID = 6f5) @ 3600.0MHz
Instruction set support : MMX SSE SSE2 

Access latency			 49.2 ns (177 clocks)

Read datarate (INT)		7544 Mb/s
Write datarate (INT)		3428 Mb/s

Read datarate (MMX)		8071 Mb/s
Write datarate (MMX)		3436 Mb/s

Read datarate (SSE)		8430 Mb/s
Write datarate (SSE)		7247 Mb/s

==================================================​
Sandra XI Random Access Latency
3888Mhz - 9x432FSB
4:5
540mhz 4-4-4-12 at 2.45v
latency.png


Sandra XI Random Access Latency
3839Mhz - 9x426FSB
4:5
533mhz 5-4-4-12 at 2.45v
(clockgen from 400FSB)
latency.png
 
Back