A-Data Xtreme DDR3-2000X 6GB testing/Hyper lovin' ? - Page 2

BenchZowner · 11-02-2009, 04:16 PM

I'm running the tests for the review right now.
I'll pop the HS off after the measurements ( probably tomorrow, got lots of measurements to take ).

eva2000 · 11-03-2009, 03:53 AM

Bill hope you can share some of your voltages used as I'm having weird experience in that 7-8-7-24 1t timings is much more stable in memtest86+ 4.0 than 8-9-8-24. And 8-9-8-24 is more stable than 9-9-9-24 all at same voltages!!!!!!! Testing with i7 920 3845B260 and P6T6 WS Revo

BenchZowner · 11-03-2009, 04:09 AM

I haven't tried really loose timings like 9T tRCD ( mostly because this CPU is IMC & BCLK limited ).
But... on a Gigabyte GA-EX58-UD5 ( F9e BIOS ) they're stable at:

DDR3-2100 8-8-8-25-1-2-5-8-19-7-80-4-1-1 B2B 10 @ 210MHz BCLK @ 1.56V ( BIOS )
DDR3-2120 7-8-7-23-1-2-5-8-19-7-72-4-1-1 B2B 8 @ 212MHz BCLK @ 1.84V ( BIOS )
DDR3-2070 7-7-5-22-1-2-5-8-19-7-72-4-1-1 B2B 8 @ 207MHz BCLK @ 1.8V ( BIOS )
DDR3-1870 6-7-6-20-1-2-4-7-17-7-52-4-1-1 B2B 4 @ 187MHz BCLK @ 1.82V ( BIOS )

eva2000 · 11-03-2009, 06:01 AM

Thanks Bill, seems I might have different behaving modules hence why I found 12GB config a bit tricky.

Starting prelim memtest86+ v4.0 tests and interesting 4 of the 6 modules (modules #1-4) can handle 2000mhz 7-8-7-24 with 3 of them at lower QPI/DRAM but 1 needs higher QPI/DRAM (module #3). The other 2 modules (modules #5 and #6) error out slightly in 7-8-7-24 at 2000mhz but handle 7-9-7-24 and 8-9-8-24 better with much lower vdimm than the other 4 modules <1.62v. Pairing up modules #1 + #2 + #4 can do 2000mhz 7-8-7-24 around 1.68v. Pairing up modules #5 + #6 + #3 can do 2000mhz 7-9-7-24 at 1.62v and like lower vdimm over high vdimm.

First bench of modules 5 + 6 + 3, easily done at DDR3-2097Mhz 8-9-8-24 2T at 1.64v.

Not bad considering 12GB cost around AUD$396 or US$352

BenchZowner · 11-06-2009, 06:15 AM

HeatSpreader removed, chips revealed and...
... hmm...

Any durability tests and maximal voltages tested, anybody ?

eva2000 · 11-06-2009, 06:39 AM

0911 seems pretty old hypers indeed. Not sure if mine are the same though.

BenchZowner · 11-06-2009, 06:41 AM

I hope these are not the degrading cr*p from Elpida

eva2000 · 11-06-2009, 06:48 AM

Only time will tell, seems some of my modules are elpida bbse going by inability to do 7-8-7-x and liking 7-9-7-x and 8-9-8-x

Ronbo56 · 11-07-2009, 06:27 AM

Quote:

Originally Posted by BenchZowner

HeatSpreader removed, chips revealed and...
... hmm...

Any durability tests and maximal voltages tested, anybody ?

If you guys remember the Kingston Hyper X thread, the old hypers were the MHN the new hypers are the MGH.

eva2000 · 11-07-2009, 07:21 AM

whoops yeah overlooked the fact that Bill's photo is of MGH-E Hypers! so could be okay ?

Interesting findings for me especially has I have a problematic module #3 in first kit1 which likes higher QPI/DRAM and VDIMM than module #5 and #6.

System:

Intel Core i7 920 3845B260
EK Supreme LT 1366 / FESER 480 / MCP655
Asus P6T6 WS Revolution
A-Data Xtreme Series 2000X 9-9-9-24 1.6-1.8 rated memory
512MB Inno3D 8800GT OC Edition 190.62WHQL
750GB Samsung SATA
Pioneer DVR-217 DVD-RW
Corsair HX1000 PSU
Dual Boot - WinXP Pro SP3 (nLite fully updated) with Vista Theme & Win7 Ultimate 64bit RTM

DDR3-2000Mhz 8-9-8-24 1T at 1.62v
(Kit 1: module #5 + #3 + #6)

DDR3-2137Mhz 8-9-8-24 1T at 1.64v
(Kit 1: module #5 + #3 + #6)

eva2000 · 11-08-2009, 02:10 AM

Hmm could these indeed be a MGH-E Hypers ? Lower max 7-8-7-x than original Hypers but higher than Elpida BBSE are capable of ?? Or good Elpida BBSE ?

DDR3-1960Mhz 7-8-7-24 1T at 1.62v
(Kit 1: module #5 + #3 + #6)

From testing individual modules #5 and #6 they don't like DDR3-2000Mhz 7-8-7-x at all, while module #3 can handle 7-8-7-x but needs higher QPI/DRAM and vdimm voltage than modules #1, #2 and #4. So decided to see how high Kit 1 with modules #5 + #3 + #6 can go with 7-8-7-24 1T timings. I found the following from using memtest86+ v4.0 and HyperPi 8x32M tests:

With same tFAW = 34, B2B = 4 and DDRAM Data/CTRL VREF A/B/C tweaks as my 8-9-8-24 tests, I managed to get much less errors at DDR3-2000Mhz 7-8-7-24 1T. With tweaks around 50-65 errors, without tweaks = 10000s of errors in memtest86+ v4.0.
But as modules #5 and #6 like lower than 1.64v vdimm memory voltage, couldn't get DDR3-2000Mhz stable.
The highest I could go for memtest86+ v4.0 and HyperPi 8x32M was DDR3-1960Mhz 7-8-7-24 1T at 1.62v.

eva2000 · 11-08-2009, 05:17 AM

Definitely a different beast compared to Bills above kits,

DDR3-1900Mhz 6-8-6-24 1T at 1.62v
(Kit 1: module #5 + #3 + #6)

Testing CAS6 clocks, with 6-7-6-24 1T at 1.62v max memory frequency came to DDR3-1720Mhz (860mhz), but loosen tRCD from 7 to 8 and max memory frequency jumped to DDR3-1900Mhz 6-8-6-24 1T at 1.62v!

BenchZowner · 11-08-2009, 05:34 AM

George A-data sent you those or you bought them ?
If the answer to the second part of the question is yes, then it's definitely using different chips if not binning as well.

A-data told me that these kits are due for release in a week/2 weeks, if that info is correct, then I think they should change their model number to separate them from the older DDR3-2000X kits.

eva2000 · 11-08-2009, 05:51 AM

I bought them last week 12GB 2000X for AUD$396 or US$352 - too cheap for Hyper's probably hehe but still nice value

12GB - 6x2GB A-Data 2000X
@DDR3-1702Mhz 6-7-6-24 2T at 1.62v
(Kit 1 + 2: module #1 + #5 + #4 + #3 + #2 + #6)

It paid off to individually test all 6x 2GB A-Data 2000X modules and sort out the problematic module #3 as now it's sorted out, clocking 12GB - 6x2GB config is so much easier, straight into 12GB A-Data 2000X @DDR3-1702Mhz 6-7-6-24 at 1.62v. Kept CMD Rate at auto for 2T just for starters.

Stable for Super Pi 32M at least. Need to work on HyperPi 8x32M tommorrow

click image for full screenshot