eva2000
Gawd
- Joined
- Apr 14, 2002
- Messages
- 539
Lots of folks have reported the new Elpida 50mm Hyper IC modules used for the high end triple channel memory kits fairs pretty well for overclocking on X48 DDR3 platform. So decided to check it out with my 3x2GB Kingston HyperX T1 DDR3-2000Mhz CAS8 Triple Channel kit (KHX16000D3ULT1K3/6GX) on Asus Rampage Extreme as my DFI UT X58-T3EH8 is occupied for other tests.
System:
Lots of folks have reported 3x2GB Kingston HyperX T1 DDR3-2000Mhz CAS8 Triple Channel kits having a module failure or a poor module within their 3x2GB kit on X58 platform. So I decided to kill 2 birds with 1 stone so to speak. I will test each 2GB module individually on Rampage Extreme's optimal dimm slots A2+B2. Each 2GB module will be tested on A2 and then B2 dimm slots using quick Memtest86+ v2.11 loop test #5 for 10 passes minimum. Not a stressful test but enough to weed out which 2GB modules are the better of the 3 modules to be paired for dual channel 2x2GB operation on Rampage Extreme.
I used a bit of masking tape to label each 2GB module - #1, #2 and #3 modules respectively.
Results for 8.5x cpu multiplier and main 8-8-8-24 timings on 266FSB Strap 1:2 divider.
At certain frequency subtimings for tRTR_S/RTR_D and tWTW_S/WTW_D had to be reduced to 5/9 for each pair for read to read same and different ranks and write to write same and different ranks respectively to pass test #5 loop. This reduced memtest reported bandwidth by ~10.5%. Still great once you factor in dual channel bandwidth will be higher.
i.e. at 950mhz and 960mhz respectively bandwidth dropped from AUTO subtimings to manually set to above from 6145MB/s to 5500MB/s and 6210MB/s to 5558MB/s.
With looser subtimings for at:
Module #1 in A2 dimm slot
Module #2 in A2 dimm slot
Module #3 in A2 and B2 dimm slot
Module #3 cut short my tests for B2 dimm slot testing, so will do that later with module #1 and #2 in B2 and update this post with the results later on.
Conclusion:
As you can see module #1 didn't fair that well compared to module #2. At 950mhz it already required the 4 mentioned subtimings to be loosened at least 10mhz earlier/lower than when module #2 needed to touch those subtimings. Module #2 continued to scale all the way to 990Mhz at same vdimm voltage until 1000Mhz which needed vdimm bump from 1.72306v to 1.74956v. I didn't apply much more vdimm over that for 2GB single module testing so as to leave room for when dual channel is tested which will require more vdimm usually. Module #3 on the other hand was a complete dud which I'd deem faulty by any definition
As you can see there are differences amongst the 3x 2GB modules even on X48. Will these differences translate over on X58 in triple channel mode ? Are these differences one explanation as to reason some folks reporting these 3x2GB Kingston HyperX T1 2000 CAS8 kits have one module failing or not hitting rated speeds ??
Module #3 was a disappointment and could be reason why on DFI UT X58-T3EH8 i got 4GB detected out of 6GB in a few different module/dimm slot combinations. If the modules were placed in a certain dimm slot configuration, all 6GB was detected at DDR3-2000Mhz 8-8-8-24 on DFI UT X58-T3EH8. Same with certain CPU VTT and VDIMM combinations 6GB was detected instead of 2GB or 4GB (memory tip #6). But didn't test this kit much as Corsair Dominator GT kit was busy on DFI UT X58-T3EH8.
I'd say worth looking into testing each module individually if you are having 3x2GB issues on X58 platform as well. Of course module #3 would be deemed faulty for not running rated timings at least 8-8-8-24 for lowly DDR3-1066mhz to DDR3-1333Mhz underclock!.
System:
- E3110 Q750A158
- OCZ Vendetta 2 Air
- Asus Rampage Extreme 1104 bios
- 128MB Gainward FX5200 PCI
- 3x 2GB Kingston HyperX T1 DDR3-2000Mhz CAS8 Triple Channel kit (KHX16000D3ULT1K3/6GX)
- 1TB Samsung SATAII
- Pioneer 216 DVD-RW
- Corsair HX1000
- WinXP Pro SP3 nLite fully updated
Lots of folks have reported 3x2GB Kingston HyperX T1 DDR3-2000Mhz CAS8 Triple Channel kits having a module failure or a poor module within their 3x2GB kit on X58 platform. So I decided to kill 2 birds with 1 stone so to speak. I will test each 2GB module individually on Rampage Extreme's optimal dimm slots A2+B2. Each 2GB module will be tested on A2 and then B2 dimm slots using quick Memtest86+ v2.11 loop test #5 for 10 passes minimum. Not a stressful test but enough to weed out which 2GB modules are the better of the 3 modules to be paired for dual channel 2x2GB operation on Rampage Extreme.
I used a bit of masking tape to label each 2GB module - #1, #2 and #3 modules respectively.
Results
Results for 8.5x cpu multiplier and main 8-8-8-24 timings on 266FSB Strap 1:2 divider.
At certain frequency subtimings for tRTR_S/RTR_D and tWTW_S/WTW_D had to be reduced to 5/9 for each pair for read to read same and different ranks and write to write same and different ranks respectively to pass test #5 loop. This reduced memtest reported bandwidth by ~10.5%. Still great once you factor in dual channel bandwidth will be higher.
i.e. at 950mhz and 960mhz respectively bandwidth dropped from AUTO subtimings to manually set to above from 6145MB/s to 5500MB/s and 6210MB/s to 5558MB/s.
With looser subtimings for at:
- tRTR_S= 5
- tRTR_D = 9
- tWTW_S = 5
- tWTW_D = 9
- 970mhz = 5616MB/s
- 990mhz = 5732MB/s
- 1000mhz = 5790MB/s
Module #1 in A2 dimm slot
- 940mhz PASSED - 1.67006v vdimm, 1.60372v NB, NB DDRVREF -20.0mv with A2 Clk Skew = Advance 250ps
- 950mhz PASSED tRTR_S/RTR_D 5/9 tWTW_S/WTW_D 5/9 - 1.72306v vdimm, 1.60372v NB, NB DDRVREF -17.5mv with A2 Clk Skew = Advance 250ps
- 960mhz FAILED 1PE10 tRTR_S/RTR_D 5/9 tWTW_S/WTW_D 5/9 - 1.73631v vdimm, 1.60372v NB, NB DDRVREF -17.5mv with A2 Clk Skew = Advance 350ps
Module #2 in A2 dimm slot
- 940mhz PASSED - 1.67006v vdimm, 1.60372v NB, NB DDRVREF -17.5mv with A2 Clk Skew = Advance 250ps
- 950mhz PASSED - 1.67006v vdimm, 1.60372v NB, NB DDRVREF -17.5mv with A2 Clk Skew = Advance 350ps
- 960mhz PASSED tRTR_S/RTR_D 5/9 tWTW_S/WTW_D 5/9 - 1.72306v vdimm, 1.60372v NB, NB DDRVREF -17.5mv with A2 Clk Skew = Advance 350ps
- 970mhz PASSED tRTR_S/RTR_D 5/9 tWTW_S/WTW_D 5/9 - 1.72306v vdimm, 1.60372v NB, NB DDRVREF -17.5mv with A2 Clk Skew = Advance 350ps
- 980mhz PASSED tRTR_S/RTR_D 5/9 tWTW_S/WTW_D 5/9 - 1.72306v vdimm, 1.60372v NB, NB DDRVREF -17.5mv with A2 Clk Skew = Advance 350ps
- 990mhz PASSED tRTR_S/RTR_D 5/9 tWTW_S/WTW_D 5/9 - 1.72306v vdimm, 1.60372v NB, NB DDRVREF -17.5mv with A2 Clk Skew = Advance 350ps
- 1000mhz PASSED tRTR_S/RTR_D 5/9 tWTW_S/WTW_D 5/9 - 1.74956v vdimm, 1.60372v NB, NB DDRVREF -17.5mv with A2 Clk Skew = Advance 350ps
Module #3 in A2 and B2 dimm slot
- WTF even at 666mhz 8-8-8-24 = DET DDRAM error on POSTLED code couldn't boot system in A2 dimm slot.
- Moved to B2 dimm slot and only after clearing CMOS got CMOS ERROR POSTLED code.
- So installed module #1 in A2 and problematic module #3 in B2 dimm slots and it booted into bios fine after clearing CMOS. SPD values read 9-9-9-24 tRFC 74 with 1.64v default vdimm set by bios.
- So in B2 dimm slot tried 8-8-8-24 and again = DET DDRAM error no boot.
- So in B2 dimm slot change to 9-9-9-24 timinsg and it booted up into memtest fine at 1.67v but to get into windows needed 1.7v but windows seems like it had some graphic/display corruption!
- Down clocked memory to DDR3-1066Mhz 9-9-9-24 and windows didn't have corruption enough to get a screenshot of cpuz spd for both module #1 and #3 together as i wanted to grab the SPDTool profile for them but system rebooted due to module #3 instability before I could open up SPDTool
- Looks like module #3 is really a dud
Module #3 cut short my tests for B2 dimm slot testing, so will do that later with module #1 and #2 in B2 and update this post with the results later on.
Conclusion:
As you can see module #1 didn't fair that well compared to module #2. At 950mhz it already required the 4 mentioned subtimings to be loosened at least 10mhz earlier/lower than when module #2 needed to touch those subtimings. Module #2 continued to scale all the way to 990Mhz at same vdimm voltage until 1000Mhz which needed vdimm bump from 1.72306v to 1.74956v. I didn't apply much more vdimm over that for 2GB single module testing so as to leave room for when dual channel is tested which will require more vdimm usually. Module #3 on the other hand was a complete dud which I'd deem faulty by any definition
As you can see there are differences amongst the 3x 2GB modules even on X48. Will these differences translate over on X58 in triple channel mode ? Are these differences one explanation as to reason some folks reporting these 3x2GB Kingston HyperX T1 2000 CAS8 kits have one module failing or not hitting rated speeds ??
Module #3 was a disappointment and could be reason why on DFI UT X58-T3EH8 i got 4GB detected out of 6GB in a few different module/dimm slot combinations. If the modules were placed in a certain dimm slot configuration, all 6GB was detected at DDR3-2000Mhz 8-8-8-24 on DFI UT X58-T3EH8. Same with certain CPU VTT and VDIMM combinations 6GB was detected instead of 2GB or 4GB (memory tip #6). But didn't test this kit much as Corsair Dominator GT kit was busy on DFI UT X58-T3EH8.
I'd say worth looking into testing each module individually if you are having 3x2GB issues on X58 platform as well. Of course module #3 would be deemed faulty for not running rated timings at least 8-8-8-24 for lowly DDR3-1066mhz to DDR3-1333Mhz underclock!.