Latency Results
Because Intel’s new architecture doesn’t support good base clock overclocking options, today’s tests are limited to multiples of 266.6 MHz. Even then, the highest multiplier that works provides a data rate of only DDR3-2133. For most builders, going beyond basic DDR3-2133 performance requires tighter timings rather than higher frequencies, and that’s where we placed our focus.
| Best Timings at 1.60 Volts | |||
|---|---|---|---|
| DDR3-2133 | DDR3-1866 | DDR3-1600 | |
| Geil Evo Two GET38GB2200C9ADC | 9-11-9-12 | 8-9-8-10 | 7-8-7-8 |
| G.Skill Ripjaws X F3-12800CL7D-8GBXH | 9-11-9-13 | 8-9-8-12 | 7-8-7-10 |
| PQI Immortality Turbo MFAFR602SA7001 | 10-11-10-11 | 9-10-9-10 | 7-9-7-9 |
| Patriot Viper Xtreme PXD38G1866ELK | 10-11-10-11 | 9-10-9-10 | 8-9-8-9 |
| Kingston HyperX T1 KHX1600C9D3T1K2/8GX | 10-11-10-12 | 8-9-8-10 | 7-8-7-8 |
| Crucial Ballistix BL2KIT51264FN2001 | 10-12-10-12 | 9-10-9-10 | 7-9-7-10 |
| Corsair Vengeance CMZ8GX3M2A1600C9 | Unstable | 9-10-9-10 | 7-9-7-10 |
| PNY Optima MD8192KD3-1333 | Incapable | Incapable | 10-8-8-24 |
Geil’s DDR3-2200-rated Evo Two tops the stability charts at DDR3-2133, while Kingston’s modestly-named DDR3-1600 matches it at both DDR3-1600 and DDR3-1866. That’s extra impressive for a set of Kingston modules that are only rated at DDR3-1600 CAS 9.
The two biggest surprises in today’s tests are the G.Skill and Kingston DDR3-1600 modules that clock up to an impressive DDR3-2133. G.Skill even places second in DDR3-2133 timings, though it also starts off with a better DDR3-1600 CAS 7 rating.
By now, some readers are likely screaming that these must be hand-picked samples. And yet, our recent System Builder Marathon machine also reached DDR3-1866 CAS 8 using retail samples of G.Skill’s cheaper DDR3-1600 CAS 8 DIMMs.
Corsair also reaches DDR3-2133, but an error on one memory IC at that frequency prevents it from remaining stable for more than 20 minutes—regardless of timings. Rather than view the other samples as having questionably good luck, it appeas Corsair simply had worse-than-average luck on a single DIMM. Remembering that these are also rated at “only” DDR3-1600 CAS 9, an overclock to DDR3-1866 CAS 9 is more than acceptable.
If we modify our expectations a bit and embrace acceptability, PNY’s low-cost DDR3-1333 kit reached DDR3-1600. It did so at a relatively-poor CAS 10, but with a value-oriented price that helps keep the firm competitive.
now watch as that kingston ram price suddenly jumps up
i miss the days when you could get 4gb of ram for 30 bucks *sigh*
What happened to "get DDR3-1333 with tight timing: You'll never be able to appreciate `faster' stuff"?
[Brand X] stuff rated 7-7-7-18 ment something, I thought …
what cpu cooler did you use? because some cpu coolers don't alow big memory ram instal
MMk sorry goose more ram is almost always and i mean 99.99999% better then faster ram
i never saw any differences in ram speed turning off t1 and t2 timings slowing it down to a crawl turning off dual channel.. NEVER gave me an effective visual result. But adding more ram always gave me an instant result. I had this same argument with someone last week.. Nice to now have proof he was wrong about faster ram meaning something 
My ram is rated at 1600 but i have it only set on 1333 
My DDR3 2000 CL9 runs @ 1915MHz CL6 just fine. And I had some DDR2 800 @ 1010MHz & DDR2 533 @ 727MHz. I think most RAM just OC's nicely. I've also had a few weaker sticks (DDR2 667 that can't go over 727MHz), but it all exceeds posted specs.
I'm expecting to get my 2*4GB + 2*4GB DDR2 800MHz this week
BUT
If I could do it over again I'd get that overly expensive DDR3 motherboard and just 1GB of RAM then later add more RAM sticks
Whenever DDR4 comes I'll jump in with small sticks and upgrade to more RAM when it gets cheaper (due to 20nm->15nm shrink)
Well, that's when Windows 9 arrives and 16 cores is the mainstream (2017?) I hope I have enough money for 3D projector at QuadHD, 4feet by 8feet white wall...
I would like to have seen something like the G.Skill Value series tested, but it really looks like RAM just doesn't make all that much difference for games.
I know this wasn't the primary focus of the article, though it is a good example of how sending more juice to your RAM and upping the speed on it has no noticeable effect on gaming performance.
I know this wasn't the primary focus of the article, though it is a good example of how sending more juice to your RAM and upping the speed on it has no noticeable effect on gaming performance.
Agreed,
I would like to see a cheaper stick thrown in there like my Kinston Standard 512M X 64 Non-ECC 1333MHz 240-pin Unbuffered DIMM (DDR3, 1.5V, CL9, FBGA, Gold)
These RAMs with 19" Alloy wheels dont really seem to be worth their pricetags.
I think it is safe to say it is better to spend money on a better graphics card or CPU, perhaps a PSU.
In the performance tests I notice that the PNY at 1600 performs just as well as the G.Skill at 2133. So from a gaming price/performance point of view cheaper most definitely means better!
Overclockers no longer NEED fast RAM. High speed RAM has simply become a bragger's product, since there is little performance increase. K brand 1155 CPUs can easily be overclocked with cheapo value RAM. Stability is the new name of the RAM game.
so basically get the cheapest fastest RAM you can possible buy
Hasn't that always been the case? Even more so now, when most ram comes with lifetime warranty.
Overclockers no longer NEED fast RAM. High speed RAM has simply become a bragger's product, since there is little performance increase. K brand 1155 CPUs can easily be overclocked with cheapo value RAM. Stability is the new name of the RAM game.
There are some serious bottleneck issues at extremely high performance scenarios, as demonstrated by previous system builder articles on TH in the past. I suggest going back to read those.
Overall, I think this is a great article. But I think where it does fail is the lack of reproducing and demonstrating a ram bottleneck scenario, to show how and why faster ram can be important in overall system performance.
Had it first started with that info, with clear visual proof that bottlenecks can and do occur, this article would have been ace.
Ive never seen any article show that faster ram = any significant change in gaming FPS. Every article completely ignores this, and I've complained multiple times before. The range of prices between RAM in this article is nearly double. Will this net you anywhere near double FPS? not even 25%, I suspect not even 5% difference. I can only imagine buying high end RAM if I just didnt know what to do with all my money, or I were buying a $2000 system.
I don't feel too bad about getting 16GB of the PNY now for $100
Just bought 8GB of Corsair Vengeance on a Newegg Shellshocker for $75.
Nice, now what about triple channel RAM?