Some of you guys are asking what is the best ram for sandy bridge... i'm going to test some to find out. However this is not a value/performance thread (as some of these cost some major money), it's purely what these can acheive in terms of performance while keeping voltages @ 1.65v (4 24/7 use). After wards i'll partition this thread and do some benching with 1.7v+ 
For my testing method i use super pi 32m which is a very good indicator of 24/7 stability
Testbed:
Asus Maximus IV Extreme
2600k@ stock (for now)
Real Vdimm (mV) on MIVE bios 1598:
I have vdimm in bios set to 1.65 exactly... i have full phase set for vdimm, ocp disabled. Seems to overshoot a tiny bit
Corsair Dominator TR3X6G1600C8D
ver: 2.1
8-8-8- 1T@ 1600MHz (1.65v)[/I][/B]
I haven't changed any of the sub timings.. only the main 5, so i'm not sure if it will make a difference.
To sum up:
1600 8-8-8 1.65v ok
1600 7-7-7 1.65v ok
1600 6-6-6 1.65v no boot
1600 6-7-6 1.65v no boot
1600 6-x-x 1.65v no boot.
here is the surprising and disappointing part (i tried all the sb dividers)
1866 7-7-7 1.65v no boot
1866 7-8-7 1.65v no boot
1866 7-x-x 1.65v no boot
1866 8-x-x 1.65v no boot
1866 9-x-x 1.65v no boot
1866 10-x-x 1.65v no boot WOW!
i tried 2133 just for fun.
- nothing boots no matter timing
Gskill pi F3-17600CL7D-4GBPIS
manufactured date: feb 2011
7-10-10 1T @ 2200MHz (1.65v)[/I][/B]
Super pi 32m test (1.65v)
2250MHz 7-10-7 1T (1.65v)
2133MHz 7-10-7 1.65v
6-8-6 1866 (looking like it's gonna be psc xey will try 6-7-6
super pi 1m (am lazy, will update if any of these are unstable) stable
2250MHz 7-10-7 1.65v stable
just got these 10 minutes ago and basically did nothing but switched to 2133 strap, changed tRP to 7 and tRAS to 27. Still testing and will update as i go.
Corsair Dominator GT CMT4GX3M2A2000C8
ver 7.1
8-9-8 1T @ 2000MHz (1.65v)[/I][/B]
These are definitely PSC... doesn't like anything but tRCD+2
Super pi 32m (1.65v) cpu@stock
*update: 6-10-10 doesn't work for 2133 1.65v
@ any frequency >2133 cas 7 becomes unstable @ 1.65v.
## no boot @ 2250MHz 7-9-7, 7-10-7, 7-10-10, 8-9-8, 8-10-8,
Currently testing stability @ 2250MHz 9-10-9 1.65v
Tested
2250MHz 9-10-9 1.65v
cannot lower timings past this for 2250 (1.65v)
2133 7-9-7 1.65v
2133 7-10-7 1.65v
1866 6-8-6 1.65v
still testing
CMG4GX3M2A2000C8 (hyper mgh-e)
super pi 32m (1.65v) bios 1850
7-7-7 1866MHz 1.65v
super pi 1m (1.65v)
8-8-8 2133MHz 1.65v
you might be able to deduce that i was unable to get 2133 and these hypers to play nice
it's pretty well documented that sb does not like hyper ic... only if you have top end boards such as the mive and ud7/5 can you actually run them 32m stable.
Most that can have to run 1.7-1.72v vdimm, and a higher vccio ~1.14-1.2 ... so in accordance with my current testing method with 1.65v and a measly 1.1v for vccio, i was unable to over 1866 divider.
However these are 32m stable at 1600 6-6-5 1T and 1866 7-7-7 1T which is pretty darn good.. they will also scale up from those dividers if you increase blck and still be 32m stable (will get ss of that soon..) weird.
Corsair Dominator GT CMT6GX3M3A2000C8
super pi 32m 7-9-7-27 1T 94 @ 2133MHz (cpu@ 4.8ghz)
Gskill Ripjaws F3-17000CL7D-4GBXHD
screen shots comming soon, but seems like takes more volts to get 6-10-6 stable high frequency than my gskill pi 2200cl7. 6-9-6 not possible on standard cooling
The order of the ram in which i think is best for 24/7 usage and benching
1.Gskill pi F3-17600CL7D-4GBPIS:
This particular 2x2gb kit from gskill scales very well with it's rated voltage of 1.65v. For those of you who hesitate to increase vdimm due to worries of damaging your cpu, and who also want to get into benching, these are perfect for you. I've actually kept them stable at 2250MHz cas7 with only 1.62. They yielded the best super pi time and latency with cpu at the stock 3.8ghz. Aesthetically speaking they aren't to shabby looking
They are currently very scarce and hard to find .
2.Corsair Dominator GT CMT4GX3M2A2000C8:
This particular 2x2gb kit from corsair scales good with it's rated voltage of 1.65v. It does not compare to what was once the absolute best of the best label/brand name in "dominator gt" in previous years they used elpida IC. Due to the fallout with the defective hyper mnh-e ic (a large batch started dying with no cause), the majority of the RAM companies moved to using different integrated circuits for their top end stuff. This kit is indicative of what can be now considered premium performance in today's stream of marketed 'overclocking' ram. By that, i mean these can be had without to much trouble by going onto your favorite e-tailer and clicking buy... no need to search high and low on multiple forums and hundreds of ebay listings to find one. 7-9-7 2133 was achievable with the rated voltage of 1.65v.
3.CMG4GX3M2A2000C8 (hyper mgh-e):
this one is tough... it is a hyper mgh-e based module, which is both it's greatest asset and liability. To explicate further, hyper ic was/is considered the best ic for overclocking ram. However on this current platform 1155, it has trouble with compatibility issues and thus cannot achieve it's standard/full potential. Within my parameters of using just the stock voltage, the best i could get was near it's rated frequency of 2000MHz and it's rated timing 8-8-8. I am not certain where the issue lies... whether that issue is primarily a bios problem, or an inherent compatibility issue with sb imc itself. If it was the former i suspect that gigabyte, asus, or one of the other board manufacturers would have revised a bios by now that ameliorated the issue. From this observation i conclude that it lies within the new generation cpu itself. In terms of it's scaling with standard voltage, i cannot say it is very good since it barely reaches it's rating.. however in regards to tight timings this is the best out of the batch tested in this review. I got 6-6-5 1600MHz stable which was not possible with the others without a voltage increase.
4.Corsair Dominator TR3X6G1600C8D
I bought this hoping that the ic used would be some sort of elpida.. bbse, bdbg, or maybe hyper. The revision indicated it was so according to a table posted by corsair reps. 2.x was suppose to indicate ic by elpida. However once i tested it, it did not behave like elpida at all as it could not overclock to even the nearset sb memory divider which was 1866, regardless of how loose i set the timing. The only thing i could do with it was to decrease timing to 7-7-7 @ 1600MHz with the rated voltage of 1.65v. This leads me to beleive that the ic used was not elpida or powerchip since powerchip would have been able to increase in frequency with looser timings, nor could it have decreased tRCD to +0 without increasing it's rated voltage. I had youngpro who is one of the well known memory gurus tell me that it was samsung ic.. so i suspect he was correct by looking at it's behavior. This does not scale at all with voltage. It was a big disappointment.
Ocn member submissions
i have more ram that i will be testing in the near future, but testing ram is very time consuming... will update as i go
Edited by badatgames18 - 11/4/11 at 6:19pm
For my testing method i use super pi 32m which is a very good indicator of 24/7 stabilityTestbed:
Asus Maximus IV Extreme
2600k@ stock (for now)
Real Vdimm (mV) on MIVE bios 1598:
I have vdimm in bios set to 1.65 exactly... i have full phase set for vdimm, ocp disabled. Seems to overshoot a tiny bit
Corsair Dominator TR3X6G1600C8D
ver: 2.1
8-8-8- 1T@ 1600MHz (1.65v)[/I][/B]
I haven't changed any of the sub timings.. only the main 5, so i'm not sure if it will make a difference.
To sum up:
1600 8-8-8 1.65v ok
1600 7-7-7 1.65v ok
1600 6-6-6 1.65v no boot
1600 6-7-6 1.65v no boot
1600 6-x-x 1.65v no boot.
here is the surprising and disappointing part (i tried all the sb dividers)
1866 7-7-7 1.65v no boot
1866 7-8-7 1.65v no boot
1866 7-x-x 1.65v no boot
1866 8-x-x 1.65v no boot
1866 9-x-x 1.65v no boot
1866 10-x-x 1.65v no boot WOW!
i tried 2133 just for fun.
- nothing boots no matter timing
Gskill pi F3-17600CL7D-4GBPIS
manufactured date: feb 2011
7-10-10 1T @ 2200MHz (1.65v)[/I][/B]
Super pi 32m test (1.65v)
2250MHz 7-10-7 1T (1.65v)
2133MHz 7-10-7 1.65v
6-8-6 1866 (looking like it's gonna be psc xey will try 6-7-6
super pi 1m (am lazy, will update if any of these are unstable) stable
2250MHz 7-10-7 1.65v stable
just got these 10 minutes ago and basically did nothing but switched to 2133 strap, changed tRP to 7 and tRAS to 27. Still testing and will update as i go.
Corsair Dominator GT CMT4GX3M2A2000C8
ver 7.1
8-9-8 1T @ 2000MHz (1.65v)[/I][/B]
These are definitely PSC... doesn't like anything but tRCD+2
Super pi 32m (1.65v) cpu@stock
*update: 6-10-10 doesn't work for 2133 1.65v
@ any frequency >2133 cas 7 becomes unstable @ 1.65v.
## no boot @ 2250MHz 7-9-7, 7-10-7, 7-10-10, 8-9-8, 8-10-8,
Currently testing stability @ 2250MHz 9-10-9 1.65v
Tested
2250MHz 9-10-9 1.65v
cannot lower timings past this for 2250 (1.65v)
2133 7-9-7 1.65v
2133 7-10-7 1.65v
1866 6-8-6 1.65v
still testing
CMG4GX3M2A2000C8 (hyper mgh-e)
[/B]
super pi 32m (1.65v) bios 1850
7-7-7 1866MHz 1.65v
super pi 1m (1.65v)
8-8-8 2133MHz 1.65v
you might be able to deduce that i was unable to get 2133 and these hypers to play nice
it's pretty well documented that sb does not like hyper ic... only if you have top end boards such as the mive and ud7/5 can you actually run them 32m stable.
Most that can have to run 1.7-1.72v vdimm, and a higher vccio ~1.14-1.2 ... so in accordance with my current testing method with 1.65v and a measly 1.1v for vccio, i was unable to over 1866 divider.
However these are 32m stable at 1600 6-6-5 1T and 1866 7-7-7 1T which is pretty darn good.. they will also scale up from those dividers if you increase blck and still be 32m stable (will get ss of that soon..) weird.
Corsair Dominator GT CMT6GX3M3A2000C8
super pi 32m 7-9-7-27 1T 94 @ 2133MHz (cpu@ 4.8ghz)
Gskill Ripjaws F3-17000CL7D-4GBXHD
screen shots comming soon, but seems like takes more volts to get 6-10-6 stable high frequency than my gskill pi 2200cl7. 6-9-6 not possible on standard cooling
The order of the ram in which i think is best for 24/7 usage and benching
1.Gskill pi F3-17600CL7D-4GBPIS:
This particular 2x2gb kit from gskill scales very well with it's rated voltage of 1.65v. For those of you who hesitate to increase vdimm due to worries of damaging your cpu, and who also want to get into benching, these are perfect for you. I've actually kept them stable at 2250MHz cas7 with only 1.62. They yielded the best super pi time and latency with cpu at the stock 3.8ghz. Aesthetically speaking they aren't to shabby looking
They are currently very scarce and hard to find . 2.Corsair Dominator GT CMT4GX3M2A2000C8:
This particular 2x2gb kit from corsair scales good with it's rated voltage of 1.65v. It does not compare to what was once the absolute best of the best label/brand name in "dominator gt" in previous years they used elpida IC. Due to the fallout with the defective hyper mnh-e ic (a large batch started dying with no cause), the majority of the RAM companies moved to using different integrated circuits for their top end stuff. This kit is indicative of what can be now considered premium performance in today's stream of marketed 'overclocking' ram. By that, i mean these can be had without to much trouble by going onto your favorite e-tailer and clicking buy... no need to search high and low on multiple forums and hundreds of ebay listings to find one. 7-9-7 2133 was achievable with the rated voltage of 1.65v.
3.CMG4GX3M2A2000C8 (hyper mgh-e):
this one is tough... it is a hyper mgh-e based module, which is both it's greatest asset and liability. To explicate further, hyper ic was/is considered the best ic for overclocking ram. However on this current platform 1155, it has trouble with compatibility issues and thus cannot achieve it's standard/full potential. Within my parameters of using just the stock voltage, the best i could get was near it's rated frequency of 2000MHz and it's rated timing 8-8-8. I am not certain where the issue lies... whether that issue is primarily a bios problem, or an inherent compatibility issue with sb imc itself. If it was the former i suspect that gigabyte, asus, or one of the other board manufacturers would have revised a bios by now that ameliorated the issue. From this observation i conclude that it lies within the new generation cpu itself. In terms of it's scaling with standard voltage, i cannot say it is very good since it barely reaches it's rating.. however in regards to tight timings this is the best out of the batch tested in this review. I got 6-6-5 1600MHz stable which was not possible with the others without a voltage increase.
4.Corsair Dominator TR3X6G1600C8D
I bought this hoping that the ic used would be some sort of elpida.. bbse, bdbg, or maybe hyper. The revision indicated it was so according to a table posted by corsair reps. 2.x was suppose to indicate ic by elpida. However once i tested it, it did not behave like elpida at all as it could not overclock to even the nearset sb memory divider which was 1866, regardless of how loose i set the timing. The only thing i could do with it was to decrease timing to 7-7-7 @ 1600MHz with the rated voltage of 1.65v. This leads me to beleive that the ic used was not elpida or powerchip since powerchip would have been able to increase in frequency with looser timings, nor could it have decreased tRCD to +0 without increasing it's rated voltage. I had youngpro who is one of the well known memory gurus tell me that it was samsung ic.. so i suspect he was correct by looking at it's behavior. This does not scale at all with voltage. It was a big disappointment.
Ocn member submissions
CL3P20's gskill pi
Gskill F3-16000CL6D-4GBPIS
2000mhz, 6-9-6-24, 1.65v
[/QUOTE]
Tests coming soonGskill F3-16000CL6D-4GBPIS
2000mhz, 6-9-6-24, 1.65v
[/QUOTE]i have more ram that i will be testing in the near future, but testing ram is very time consuming... will update as i go
Edited by badatgames18 - 11/4/11 at 6:19pm
Sandy
(21 items) |
|
Sandy
(21 items) |
