Kingston CL 3-3-3-10 memory tested

@ 2007/07/09
KHX6400D2ULK2 is a low latency memory module running at 800MHz, however it scores the same or better than 1066MHz memory. The manufacturer is Kingston, and it is a privilege to test this quality CL3 memory.

Comment from jmke @ 2007/07/10
minor, SupCom does love CPU power, so if you give CPU more bandwidth it will be able to work faster; But if you're asking me if swapping out mainboards for a few FPS increase is worth it... not quite
rule of thumb, don't upgrade unless it gives you 20+% boost overall. (20% is rather high for enthusiasts, who jump at 5% increase to upgrade less tech savvy users tend to have longer life cycle, and they also get larger rewards from their upgrade)

Went from Tbird 900 -> XP 1700+ -> P4 2.4C -> A64 3000+ -> C2D E6600. The last upgrade I noticed large increase in game speed. AXP to P4 noticed good speed increase in Windows (thanks HT)

Still using A64 3000+ for workstation, but even using Firefox with plugins, Outlook in background, and Winamp playing, I find that opening extra Word/Excel files can get "slow";
Comment from Kougar @ 2007/07/10
I see your point about modern games, that should be interesting to find out. So if I replace my P35-DQ6 with a 965P-DS3 I should see no difference with Supreme Commander @ 1920x1200 then?
Comment from jmke @ 2007/07/10
I know Anandtech article by heart by now that's why I'm certain in todays games DDR2vsDDR3 will give close 0% difference

Far Cry looks splendid, but on todays hardware it's not taxing enough, if they tried R6:Vegas at 1600x1200 the results would be 46fps DDR2, 46fps DDR3

yes their single game benchmark shows an improvement, but that 2-5% could have been larger if they ran Quake 3 or better yet, Quake 1. In the end, you have to wonder if you want a fast PC for bragging rights, if you want it because you have too much money, or if you want to get the most out of every $/€ you earn. Upgrading to DDR3 is definitely not a wise option if you're going for best bang for the buck
Comment from Kougar @ 2007/07/10
No, those 2-5% were specifically in games, and synthetics like SuperPI both. Link Considering the state of the P5K's non-mature BIOS in that test is likely a slight increase in those P5k results.

And again: Link2

Our target DDR-800 results showed a 4.6% to 5.2% improvement, so we conclude the P35 memory controller improves gaming performance roughly 5%
And I agree, 100FPS vs 105FPS, it doesn't matter in the end run, but the performance difference IS there however marginal it may or may not be.
Comment from jmke @ 2007/07/10
that 2-5% is NOT in applications; it's in synthetic benchmarks!

and in games where FPS is already >100FPS it doesn't matter; if you test a game which would give 45FPS on P965 it will be 45FPS on P35 too!
simply put: memory subsystem hasn't been the bottleneck of the PCs for quite some time now, thus increasing the bandwidth does not yield the performance increase we're used to seeing with AXP/P4 "C"
Comment from Kougar @ 2007/07/09
Originally Posted by lazyman View Post
Could the difference be with AMD onboard controller?
I am fairly sure that is part of the story. I recall reading the K8 memory controller cannot make use of timings that tight. This is also reportedly something Barcelona/Kuma/Agena is supposed to fix. I believe this info is from Ars Technica, but I'm not sure.

Intel benchmarks using similar RAM situations with games have shown higher FPS differences. I am not sure what game settings Anandtech tested though? Link

Originally Posted by Jmke
AMD has similar story, real world games/applications don't really get performance boost because of higher speed memory, the CPU has "enough" to do the job with lower rated ram.
If this was true, then why does the P35 memory controller offer 2-5% higher performance numbers over the 965P chipset, if the CPU already has enough to do? I'd agree the boost is not worth the cost of high perhformance RAM, but there is a measurable couple extra FPS.
Comment from jmke @ 2007/07/09

AMD has similar story, real world games/applications don't really get performance boost because of higher speed memory, the CPU has "enough" to do the job with lower rated ram. Increasing the Memory bandwidth without increasing the CPU FSB doesn't yield performance increase.

That's why you now see Core 2 with 1333FSB, more room, and here speedier memory does pay off (a very little bit) more.
Comment from Sidney @ 2007/07/09
Could the difference be with AMD onboard controller?
Comment from jmke @ 2007/07/09
Those last 2 game tests tell the whole story:

Don't you just feel all warm and fuzzy inside knowing that high speed memory doesn't matter at all?