Value DDR3 For Intel's P55: Six 4GB Kits Rounded Up

Test Results: Latency, Frequency, And Bandwidth

The official signal voltage limit of Intel’s integrated memory controller is 1.65V, which is well below the rated voltage of Patriot’s DDR3-1600. Yet we were able to reach Patriot’s full rated speed in spite of our lower voltage limit, and were even able to push the memory timings down significantly below its ratings.

Best Timings at 1.65V

DDR3-1600
DDR3-1333
DDR3-1066
Patriot DDR3-1600 CAS 9
PVS34G1600ELK
8-7-7-136-6-6-125-5-5-11
Crucial DDR3-1333 CAS9
CT2KIT25664BA1339
8-7-7-186-6-6-155-5-5-12
PNY DDR3-1333 CAS 9
MD4096KD3-1333
8-7-8-176-6-7-155-5-6-12
A-Data DDR3-1333 CAS 7
HY63I1B16K
8-8-7-136-7-6-125-6-5-10
G.Skill DDR3-1333 CAS 8
F3-10600CL8D-4GBHK
N/A6-6-7-125-5-5-10
OCZ DDR3-1066 CAS7
OCZ3G10664GK
N/AN/A5-5-4-11


Patriot takes the latency win at both DDR3-1600 and DDR3-1333, with Crucial behind only in tRAS. OCZ surprised us in two ways, first by reaching amazing CAS 5-5-4-11 latencies at its rated speed, and then by not being able to reach the next test speed at any latency or voltage settings.

Crucial edges out Patriot in overclocking, while similarities between the two in both overclocking and under-latency capability cause us to question whether Patriot uses the same Micron-brand ICs as its competitor. Unfortunately, we found it impossible to remove Patriot’s heat spreader without permanently damaging the modules themselves.

Crucial edges out Patriot with both kits set to their lowest stable DDR3-1600 settings. But differences this small are meaningless.

We see a repeat performance at DDR3-1333.

OCZ’s lower tRP latency puts it ahead at DDR3-1066.

Create a new thread in the US Reviews comments forum about this subject
This thread is closed for comments
60 comments
    Your comment
  • Onyx2291
    Very nice how Crucial edges out so slightly and is cheap.
    2
  • El_Capitan
    I've been advocating that Crucial memory kit since time began.
    0
  • presidenteody
    should i wait for the price to be cut in half in 6 months? I want it now for L4d2!
    0
  • dirtmountain
    Would running these same kits work any differently in an AM3 system with it's ability to handle higher voltages?
    4
  • NeatOman
    Why would you guys use memtest86+ v1.7? if "v4" or is out?
    1
  • MAD4AMD
    Could you please provide information what memory chips are used in all RAMs. Thank you!
    1
  • micky_lund
    looks good, especially for CAS9. spose it makes it better for OCing.
    pity it ain't available in Australia :S
    0
  • jrharbort
    I've always insisted that Crucial DDR3 kits are an excellent value, and recommended them to friends for their amazing stability, overclockability, and price.

    These kits totally deserved the Tom's award for best hardware of '09.
    0
  • mlcloud
    I've been buying OCZ for all my DDR2 needs (they were very cheap after rebates), but I think I know who to go to for DDR3 (and awesome power supplies)
    0
  • coolkev99
    This is good info, but I wish they would develop some 3gb x 2 kits out there.
    0
  • Crashman
    coolkev99This is good info, but I wish they would develop some 3gb x 2 kits out there.


    3GB modules don't make a lot of sense due to the way memory chips are organized. In order to do it "right" (same physical organization), you'd have to use different chips on each side of the module, such as 1gb chips on one side and 2gb chips on the other, and I'm not even sure such a module would be compatible with current memory controllers.
    0
  • enzo matrix
    What about Mushkin ram? I'd like to see them included in future articles.
    0
  • jodpel
    I had an idea in my mind that they look at the test results from every shipment of memory they get from the fab... the stuff that tests poorly gets sold as value and the stuff that exceeds specs gets heat sinks and sold as performance RAM.

    I was always leery of buying value RAM because I figured it failed a spec somewhere along the line and got thrown in the "sell it to Mikey because he'll buy anything" bin at the virtual store.
    0
  • Ryun
    "OCZ’s maximum data rate of only 1,240 MT/s came from our second attempt to test the memory. Both modules in our first kit threw errors even at a modest 1,066 MHz data rate."

    Glad you found this as well. I, and others on newegg who ordered that set of sticks, got DOA memory. Gave me errors and couldn't OC at all. It's crap RAM.
    0
  • Crashman
    jodpelI had an idea in my mind that they look at the test results from every shipment of memory they get from the fab... the stuff that tests poorly gets sold as value and the stuff that exceeds specs gets heat sinks and sold as performance RAM.I was always leery of buying value RAM because I figured it failed a spec somewhere along the line and got thrown in the "sell it to Mikey because he'll buy anything" bin at the virtual store.


    From the samples Tom's purchased it appears that there are two grades of memory: "Everything Else" and OCZ Gold.
    2
  • void_pointer
    Providing results for Memtest86+ and SiSoftware Sandra is important (IMHO), but these results really don't tell us much about real-world performance of apps/games/etc. The Crucial kit gets a win based on max data rate (and price), but there is no guarantee that you'll see a performance increase when you crank up the data rate and run your favourite app/game/etc., because the way it accesses memory with the cranked-up latencies actually cause a reduction in performance.

    If we could get a small sample of benchmarks with real apps (e.g., one or two games, an encoder or two, a file compressor or two) I think readers would have a far better understanding of what "Super-Value" means for them (and their apps/games/etc.)
    1
  • notty22
    I think its the end user, the home builder that is being cheated out of quality and value in the current ram market. With all these supposed ram modules available at different speeds/timings then multiply that by meaningless things such as color and non functional heat spreaders the end result is story after story of ram NOT WORKING at default bios settings for a given processor/memory configuration. All I know is it seems like its russian roulette building a system and expecting your ram to work with default bios settings.
    0
  • tpb
    I have found that memtest86+ simply isn't good at finding problems from overclocking.
    I have an AM3 motherboard with 8 gig of ram and I could run memtest86+ for hours and not have any problems, boot into windows and have unpredictable crashes from minutes to hours later.
    If I boot into windows and run the memory tester in the link below, the problem shows up within seconds or minutes.
    I have no idea why this memory tester appears to be better than memtest86+.
    After you get the system working reliably, run one of the CPU stress programs at the same time as the memory test to heat up the CPU and ram and see if the system is still reliable.

    http://hcidesign.com/memtest/download.html
    0
  • rockyjohn
    I used Crucial in my last build and am very happy with it.

    I agree with those who stated it would have been informative to see some applications used in the test. Also, it would have been more informative if they had included one or two of the performance RAM for comparison.
    0
  • philosofool
    That crucial memory kit isn't compatible with many Gigabyte boards. I went through RMA hell over it. It's not crucial's fault, but if you're a system builder, you should know this.
    0