ADATA 16GB XPG DDR3-2600 CL11 Memory Review

We recently published the ADATA XPG DDR3-2800 memory review which consisted of a kit based on single-sided Hynix MFR IC. Overall performance was lower than expected, even though overclocking was solid and memory clock was high. Today, we take a look at another Hynix MFR IC based kit, the ADATA XPG DDR3-2600, but it’s double sided and gives us 16 GB total. What does ADATA have to offer with a 16 GB DDR3-2600 memory kit? We’ll find out the difference between single and double-sided modules and how it affects performance in this review.

Specification and Photo Gallery

As most available memory kits on the market, our ADATA XPG came also in blister-type package which protects the memory pretty well.

This slideshow requires JavaScript.

Inside the package we find two 8 GB ADATA DDR3-2600 memory sticks protected by well known Gaming series heatspreaders…

I think that some readers already noticed one thing about which I have to add a note:

This memory won’t be available on the market in the same form as our review sample. Since we have received an early version, you can expect exactly the same modules (so the same PCB and IC ) to be part of latest XPG V2 series that you can see here or in our last ADATA XPG DDR3-2800 review mentioned earlier. Simply, the only difference is new heatspreader design which will be yellow/gold or gray/silver.

Specification

• Product Number:  AX3U2600GW8G11-DG2
• Rated speed:  DDR3-2600 / PC3-20800
• Density:  2×8 GB ( Dual Channel )
• SPD Profile:  1600 9-9-9-24 1.50 V
• XMP Profile:  2600 11-13-13-35 1.65 V

ADATA programmed only one XMP profile which can be seen below.

This is how our testing motherboard, Gigabyte Z87X-OC, sees this memory running at XMP #1 profile. Worth noting is the Command Rate being set to 1T even though it’s high density memory which usually has this timing set to 2T. Memory voltage is 1.65V which is already standard for most DDR3-2400+ RAM.

As it was already mentioned, ADATA XPG DDR3-2600 16 GB is based on double-sided Hynix MFR IC. More precisely it’s H5TQ4G83MFR PBC. We can find detailed specification of this chip on the Hynix website. Probably the most interesting for overclockers will be the maximum safe voltage up to 1.975 V and normal operating temperature up to 85 °C. Simply way above regular values and gives us a chance at higher overclocking.

Let’s take a closer look at memory kit itself…

This slideshow requires JavaScript.

Stability Testing

As usual on Overclockers.com, we are starting with stability testing at the declared manufacturer settings. To perform these tests we are using UltraX R.S.T. Pro3:

R.S.T. Pro3 is the definitive memory diagnostic and validation tool designed for the rigorous testing needs of desktop & server memory manufacturers, system builders, design engineers, and service professionals.

R.S.T. Pro3 identifies memory defects that may have passed every usual manufacturing test, but which can still fail in normal use. A major advantage of R.S.T. Pro3 compared to stand alone testers, is the capability of testing and validating RAM within the system environment, testing for Behavioral failures that are sensitive to system idiosyncrasy. Quickly isolate intermittent failures that do not necessarily prevent a system from booting, but surface during extended testing sessions.

ADATA XPG 16 GB DDR3-2600 has passed this test flawlessly five times as we can see on the screenshot below. Since we are sure that memory is perfectly stable, we can continue with performance testing.

Performance Testing

Test System

• Intel Core i7 4770K @ 4.2 GHz
• Gigabyte Z87X-OC
• Corsair HX850, 850 W 80+ Silver PSU
• Crucial M4 64 GB AHCI
• Windows 7 Ultimate

Memory kits:

• Crucial Ballistix Sport XT 2×8 GB DDR3-1866 10-10-10-30
• Geil Veloce 2×4 GB DDR3-2133 10-11-11-30
• Kingston HyperX 10th Anniversary 4x4GB DDR3-2400 11-13-13-30
• G.Skill TridentX 2×4 GB DDR3-2400 10-12-12-31
• Team Group Xtreem 2×4 GB DDR3-2666 11-13-13-35
• ADATA XPG 2×8 GB DDR3-2600 11-13-13-35

All memory kits were set to XMP profile ( highest clocked one ) so settings declared by manufacturer.

Test Software

• HyperPi 0.99b
• AIDA64 Memory & Cache benchmark
• MaxxMem v1.99
• 7-Zip x64
• Cinebench 11.5
• CPU-Z 1.65
• ASUS MemTweakIT 2.00.01

HyperPi 0.99b

One of most popular benchmarks among overclockers in multi-threaded form. It’s a nice tool to test performance and also stability.

As we see on above graph, optimal memory clock is DDR3-2400 and tight memory timings. G.Skill TridentX is fastest in this test but our review sample of ADATA DDR3-2600 isn’t much slower. Considering that ADATA is higher density memory, the result is pretty good. Team Xtreem memory based on single-sided MFR is starting to show its weak points. Result of this memory is worse even than Crucial 1866 CL10.

MaxxMem v1.99

MaxxMem is a well-known memory benchmark which uses single-threaded operations to measure memory bandwidth. On the Haswell platform, we can see a big difference between single and multi-threading what is worth to mention comparing memory transfers.

Memory write bandwidth in this test is clearly limited by CPU speed and all memory kits are achieving similar results – about 24 GB/s. Looking at other results we can see that best overall performance is again G.Skill 2400 CL10 memory. ADATA 2600 CL11 results are not so bad, but aren’t best in any category. The Geil 2133 is interesting, being one of the lowest clocked memory kits in comparison achieved the best result in the memory copy test, which was more than 1 GB/s better than the second best result.

AIDA64 Cache & Memory Benchmark

Latest version of AIDA64 introduced new set of multi-threaded tests. As we see bandwidth is much higher than for MaxxMem or older AIDA64 versions so we can’t compare those results.

ADATA XPG memory looks pretty good in memory write bandwidth, where it’s the best kit in this comparison. Since writes are going up with CPU clock, we can expect much better results during higher CPU frequencies. Read transfers aren’t so good mainly because of higher density sticks which forces the loosening of some sub-timings by the manufacturer. Copy result could be better, but it doesn’t look bad.

7-Zip x64

7-Zip, as many probably know, is the benchmark included with the popular compression-decompression application. It better shows how memory speed affects daily work.

Similar to previous tests, the best is the G.Skill memory, but our ADATA XPG kit also has good results. Differences between all memory kits are higher than expected probably because of new Haswell instructions which makes it a slightly better benchmark than it was on Ivy Bridge.

Cinebench 11.5

Cinebench is popular benchmark based on rendering, so it’s also something that shows affects on daily work more than only synthetic tests.

In this benchmark, we have totally different and unexpected results. The best result is from the Crucial Ballistix Sport XT 1866 memory, lowest clocked kit in our comparison. I ran the test a few times to be sure about the results and every time it was about the same. ADATA XPG 2600 memory nearly has the best result, but it’s hard to compare real performance in this benchmark as all results are really close to each other.

Overclocking

It was mentioned couple of times already, but I will say that again. ADATA XPG DDR3-2600 is memory based on double-sided Hynix MFR, which means it won’t overclock as high as single-sided kits. In this case, we exchange the chance for higher clocks for higher performance. I can say it’s not that bad of a deal as performance isn’t scaling well at higher memory clocks because you have to set more relaxed timings.

Keeping standard voltage of 1.65 V and relaxing timings to 12-14-14-37 I was able to set stable 1400 MHz (DDR3-2800). Results at this clock are much better in AIDA64, but not so much in MaxxMem. Simple conclusion can be that single-threaded applications won’t be affected by high speed memory as much as multi-threaded apps.

I see no point in relaxing timings even more just because it’s only lowering performance, but if anyone is curious, below is validation from maximum clock. To be honest, I was expecting a higher clock, but the highest that could boot into Windows was 1466 MHz ( DDR3-2933 ) 12-15-15-37 1.775 V.

Final Thoughts and Conclusion

Recently, we’ve all heard a lot of noise about single-sided Hynix MFR kits breaking world records in DDR3 memory frequency. It didn’t take long for us to see that these kits are great for achieving high frequencies, but at the cost of overall performance. In this review, we see that real overall performance can be had by purchasing double-sided Hynix MFR, like ADATA XMP DDR3-2600.

All who saw bad comments about single-sided kits don’t have to worry in this case. ADATA prepared really fast memory which shows its full performance on Intel Haswell platform.

Haswell is showing optimal performance with memory at around DDR3-2400-2600 clocks. No matter if we use XMP profile or lower clocks to DDR3-2400, we can achieve similar performance and all depends on the programmed (or manually set) memory timings. I think that ADATA could work some more on XMP/SPD profiles as lower clocked Kingston HyperX DDR3-2400 11-13-13-30 was really close with most results. In this case, it’s probably because we tested an early version which may be slightly corrected in the future (or when it hit the stores).

In online stores like Newegg.com , the ADATA XPG V2 DDR3-2600 11-13-13-35 cost $269.99 which isn’t a low price but in comparing to similar kits it seems reasonable. Similar to DDR3-2800 kits, ADATA reduced retail price of XPG DDR3-2600 memory kit which was nearly $360! You can actually see price drop listed on Newegg under both ADATA XPG DDR3-2600 kits in yellow/gold and gray/silver heatspreaders. If you were already thinking about this or similar memory but price was an issue then $90 price drop will for sure help to decide.

Even though I saw some higher potential in ADATA XPG V2 , it’s pretty good memory. It’s a well-performing 16 GB kit with which I had no issues during the tests. I can recommend it to anyone who is looking for fast memory for the Intel Haswell platform. I doubt that anyone will be disappointed with its performance.

– Bartosz Waluk (Woomack)