ADATA Supreme DDR3-1333 Laptop RAM Review

Add Your Comments

With systems getting smaller and smaller, many people are having to rely on laptop memory sticks (SO-DIMM) to fit everything on the motherboard.  So, when you change over to one of these tiny systems, are you compromising your ability to tweak and tune the system for it’s top performance? Luckily, the answer is: not necessarily.  With any luck, these ADATA Supreme DDR3-1333 sticks can live up to their name.

ADATA Supreme DDR3-1333 SO-DIMMs

ADATA Supreme DDR3-1333 SO-DIMMs

Just like desktop RAM, laptop RAM is offered in a variety of speeds and CAS latencies to fit every budget.  When placed in a typical laptop, heat and power draw will be major limiting factors but a Mini-ITX system can offer the options you are used to seeing with a regular ATX desktop system. So, it would seem reasonable to think that laptop RAM should be overclockable and tweakable just like it’s desktop counterparts.

Specifications

ADATA Supreme DDR3-1333 SO-DIMMs come in 1 GB, 2 GB, and 4 GB modules. I’m reviewing two 2GB modules.

Suitable for Notebook
Module Specification 204Pin SO-DIMM
Density 1 GB, 2 GB, 4 GB
IC Configuration 128×8, 256×8
CAS Latency 9
Working Voltage 1.5 V ± 0.075 V

These don’t have any heatsinks so you’ll have to be careful with the voltage.  These should run at 1.65 V, just like desktop modules, for a short period of time but I would not expect them to last if run at that setting for an extended period.

Hynix ICs

Hynix ICs

The Hynix ICs are designated as H5TQ1G83BFR  G7C, which should be DDR3-1066 CL7 according to their product page.  If this is true, then ADATA worked some magic getting them to run stable with stock settings at DDR3-1333 CL9.  The product page also hints that these should be able to run at DDR3-1600 CL10 with a little help.

Speed binning table courtesy hynix.com

Speed binning table courtesy hynix.com

Editors Note – 12/3/2010: We just received a quote from ADATA with respect to their memory binning – “Most of our DDR3 1333 modules are overclocked from 1066 IC. thus it is capable of a higher speed/CAS latency but at the same time you’ll sacrifice on frequency.”

Unfortunately, the Mvix Minix 890GX motherboard I used to test these would not allow for FSB (base clock) overclocking, so I was left with simply changing the clock ratios and latencies.  This also led to the side effect of receiving a “DIMM CL/T mismatch. DIMM(s) Running in Minimum Mode” error whenever I tried to use a ratio higher than the rated speed (800 MHz, for instance).  When I got this error, the computer would still boot fine and would pass memtest86+, but the memory bandwidth certainly took a hit.  CPU-Z would also show the running speed at 800 MHz (DDR3-1600) even though the tests would come out much slower than stock speeds.  With the locked base clock, I could not test to see if this error was simply caused by increasing the ratio or if running the sticks above spec in any way was forbidden.

Because of these limitations, the only way I was able to test these was to tighten up the latencies and drop the frequencies to see if there was any hint of performance scaling.  The ADATA sticks ran fine at DDR3-1333 CL7 settings with 1.65 V and I dropped them to DDR3-1066 to compare them to a set of Kingston sticks I had to hand.

The Tests

System Specs

  • Motherboard: Mvix Minix 890GX mini-ITX
  • RAM (Review Kit): ADATA Supreme 2×2 GB DDR3-1333 CL9
  • RAM (Alternative Set): Kingston 2×1 GB DDR3-1066 CL7
  • Processor: AMD Phenom II X2 560 BE
  • Power Supply: Silverstone NightJar ST40NF 400 W
  • Operating System: Windows 7 Ultimate x64
  • Hard Drive: Western Digital WD800JD-75JNAO 80 GB SATA

Results (Lower Scores are Better)

Adata Stock 1333 MHz, CL9 Adata 1333 MHz, CL7 Adata 1066 MHz, CL7 Kingston 1066 MHz, CL7
Pifast 34.07 33.51 33.71 33.87
SuperPi 1M 21.06 21.123 21.06 21.263
SuperPi 32M 1306.487 1285.489 1315.815 1323.069
wPrime 32M 23.865 23.974 23.946 23.995
wPrime 1024M 761.082 760.316 740.895 760.456
Relative performance graph

Relative performance graph

These results certainly look erratic and don’t represent what you’d expect.  They suggest that in Windows there are other factors at play which decide RAM performance.  I don’t have any theories as to what is going on but all of the results are within 3% of each other which doesn’t separate out a true winner in any test.

Phoronix Test Suite

Phoronix Test Suite is a collection of cross-platform tests that can be used to compare systems of any kind.  They can be run as single tests or in a bulk mode.  I chose to run the Memory Suite under Ubuntu 10.10 to get another perspective of how these memory sticks performed.

Adata Stock 1333 MHz, CL9 Adata 1333 MHz, CL7 Adata 1066 MHz, CL7 Kingston 1066 MHz, CL7
CPU: AMD Phenom II X2 560 @ 3.30GHz (Cores: 2) CPU: AMD Phenom II X2 560 @ 3.30GHz (Cores: 2) CPU: AMD Phenom II X2 560 @ 3.30GHz (Cores: 2) CPU: AMD Phenom II X2 560 @ 3.30GHz (Cores: 2)
Motherboard: MINIX 890GX-USB3 Motherboard: MINIX 890GX-USB3 Motherboard: MINIX 890GX-USB3 Motherboard: MINIX 890GX-USB3
Chipset: AMD RS780 Alternate Chipset: AMD RS780 Alternate Chipset: AMD RS780 Alternate Chipset: AMD RS780 Alternate
Memory: 3710MB Memory: 3710MB Memory: 3710MB Memory: 1754MB
Disk: 1500GB Seagate ST31500341AS Disk: 1500GB Seagate ST31500341AS Disk: 1500GB Seagate ST31500341AS Disk: 1500GB Seagate ST31500341AS
Graphics: ATI RS880 [Radeon HD 4290] (700/667MHz) Graphics: ATI RS880 [Radeon HD 4290] (700/667MHz) Graphics: ATI RS880 [Radeon HD 4290] (700/533MHz) Graphics: ATI RS880 [Radeon HD 4290] (700/533MHz)
Audio: Realtek ALC892 Audio: Realtek ALC892 Audio: Realtek ALC892 Audio: Realtek ALC892
Monitor: SyncMaster Monitor: SyncMaster Monitor: SyncMaster Monitor: SyncMaster
OS: Ubuntu 10.10 OS: Ubuntu 10.10 OS: Ubuntu 10.10 OS: Ubuntu 10.10
Kernel: 2.6.35-22-generic (x86_64) Kernel: 2.6.35-22-generic (x86_64) Kernel: 2.6.35-22-generic (x86_64) Kernel: 2.6.35-22-generic (x86_64)
Desktop: GNOME 2.32.0 Desktop: GNOME 2.32.0 Desktop: GNOME 2.32.0 Desktop: GNOME 2.32.0
Display Server: X.Org Server 1.9.0 Display Server: X.Org Server 1.9.0 Display Server: X.Org Server 1.9.0 Display Server: X.Org Server 1.9.0
Display Driver: radeon 6.13.1 Display Driver: radeon 6.13.1 Display Driver: radeon 6.13.1 Display Driver: radeon 6.13.1
OpenGL: 2.1 Mesa 7.9-devel OpenGL: 2.1 Mesa 7.9-devel OpenGL: 2.1 Mesa 7.9-devel OpenGL: 2.1 Mesa 7.9-devel
Compiler: GCC 4.4.5 Compiler: GCC 4.4.5 Compiler: GCC 4.4.5 Compiler: GCC 4.4.5
File-System: ext4 File-System: ext4 File-System: ext4 File-System: ext4
Screen Resolution: 1680×1050 Screen Resolution: 1680×1050 Screen Resolution: 1680×1050 Screen Resolution: 1680×1050
Administrator: overclockers Administrator: overclockers Administrator: overclockers Administrator: overclockers
Date: 2010-11-21 00:09:23 Date: 2010-11-21 00:42:01 Date: 2010-11-21 01:01:53 Date: 2010-11-21 06:33:22
PTS Version: 2.8.0 PTS Version: 2.8.0 PTS Version: 2.8.0 PTS Version: 2.8.0
Test Notes: Compiz was running on this system. AMD CnQ was enabled. Test Notes: Compiz was running on this system. Test Notes: Compiz was running on this system. Test Notes: Compiz was running on this system.

Results (Higher Scores are Better)

Adata Stock 1333 MHz, CL9 Adata 1333 MHz, CL7 Adata 1066 MHz, CL7 Kingston 1066 MHz, CL7
RAMSpeed Integer Add 6179.4 6503.74 5971.04 5740.21
RAMSpeed Integer Copy 5611.73 5936.38 5444.52 5156.1
RAMSpeed Integer Scale 5453.99 5781.88 5268.11 5055.94
RAMSpeed Floating-Point Add 6262.3 6575.44 6006.99 5810.91
Stream Add 7964.48 8343.09 7398.67 6965.54
CacheBench Read Cache 1886.6 1872.18 1881.59 1875.43
CacheBench Write Cache 8148.23 8156.45 8153.19 8138.11
Relative Performance Graph

Relative Performance Graph

The ADATA ram set to 1333 MHz and CAS Latency 7 wins in every test except for one, which is odd.  I cannot explain why the ADATA ram running 1333 MHz but CL9 wins in the CacheBench Read Cache test, or why the same speed but tighter latency is slowest overall.  What is very interesting, though, is that the ADATA ram set to 1066 MHz and CL7 is faster than the Kingston ram in all tests at the same settings.  This could possibly be explained as the difference between 2×2 GB (ADATA) sticks and 2×1 GB sticks (Kingston).  In reality, the CacheBench results are all within 1% of each other so it is difficult to declare a true winner.

These results are on Phoronix Global, too, so you can test you own system and compare it directly to these numbers.

Conclusion

In Linux, the memory performance is much more predictable and a little easier to declare a winner.  The ADATA Supreme DDR3-1333 sticks really surprised me that they ran so smoothly at CL7.  This coupled with the information from the Hynix website suggests to me that the ADATA Supremes are binned very well and are certainly worth tweaking to get the most performance out of them.  What’s even more interesting is that the Supremes beat the Kingstons at DDR3-1066 in all of the RAMspeed and Stream tests in Phoronix Test Suite.  For these reasons, I am marking these Overclockers Approved and wouldn’t hesitate to recommend them if you are looking for RAM for you laptop or HTPC.

– splat

Leave a Reply

Your email address will not be published. Required fields are marked *