Corsair Vengeance LPX DDR4-2666 Review

Overclocking, Tuning, Bandwidth And Latency

This CPU doesn’t work beyond a 132MHz base clock, so stepping past DDR4-3168 (132 x 24) requires me to increase the DRAM ratio and “underclock” the 100 x 1.25x BCLK. Because the Vengeance LPX wouldn't reach DDR4-3168, that step wasn’t necessary.

That’s not to say the Vengeance LPX DDR4-2666 is a bad overclocker though; its maximum-stable DDR4-3119 easily beats the DDR4-3071 of Kingston’s DDR4-3000.

Best Stable Timings

Loading...

Corsair Vengeance LPX DDR4-2666 also edges out the best timings of Kingston’s DDR4-3000. That should help it in a few benchmarks, though there are secondary and tertiary timings left to “automatic” mode that could have some effect.

Reconfiguring DDR4-2666 from this motherboard’s 21.33 x 125MHz custom overclock to a more broadly-compatible 26.66 x 100MHz required several adjustments, which, through the magic of Photoshop, are combined in a single screen shot.

Dropping the BCLK ratio (CPU Strap) from 1.25x to 1.00x (125 to 100MHz) automatically causes the actual BCLK setting to fall to its stock 100MHz. Though the motherboard automatically lowered the CPU cache ratio to 24x when setting the 1.25x strap, it did not automatically restore its 30x default when forced back to the 1.00x strap. After resetting CPU Cache to 30x, I changed the CPU multiplier from 32 (32 x 125 = 4000) to 40 (40 x 100 = 4000). Since the motherboard’s 21.33 x 125MHz data rate (DDR4-2666) was now at 2133MT/s, I also had to manually set the memory to 2666MT/s.

In spite of motherboard readings, my voltmeter shows that the board’s 1.330V setting produces an actual 1.35V to 1.36V.

Seen in the Vengeance LPX “Rated Data Rate” value, its DDR4-2666 bandwidth suffers slightly from loose 15-17-17-35 timings. Capable of far better, its manually-optimized DDR4-2400 timings match Kingston’s in both setting and bandwidth.

Loose timings likewise hurt Corsair’s rated DDR4-2666 settings in Sandra's latency measurement, though manually-optimized timings are again roughly tied with Kingston's DDR4-3000. Remember, lower is better for latency.

Create a new thread in the US Reviews comments forum about this subject
This thread is closed for comments
22 comments
Comment from the forums
    Your comment
    Top Comments
  • blackmagnum
    I have to ask: Do these aluminum clip-on heat spreaders help with lowering the temperature or they just like huge spoilers on Honda Civics (I look fast)?
    17
  • COMONGUISE
    I have to ask - When will Tom's show a comparison with DDR3? I know its on a different chipset, but we would really like to know.
    17
  • Other Comments
  • blackmagnum
    I have to ask: Do these aluminum clip-on heat spreaders help with lowering the temperature or they just like huge spoilers on Honda Civics (I look fast)?
    17
  • COMONGUISE
    I have to ask - When will Tom's show a comparison with DDR3? I know its on a different chipset, but we would really like to know.
    17
  • tom10167
    They won't do that because then nobody would buy any DDR4 ram
    4
  • SuperVeloce
    @tom10167: Well yeah, 1600 ram is enough for 99% home usage scenarios, no matter how much faster a DDR4 can be. But now you can already get 2666 cl15 ram (latency something like ddr3 1600mhz cl7, cl8), and if that's at 1,2V, even better. First ddr3 ram sticks for first amd ddr3 chipsets and core2 intels were terrible and with high voltage for faster models (I even saw 1,7V+). So in a way this time around latency drops faster and it's not even available yet in mainstream platforms...
    1
  • Crashman
    Anonymous said:
    I have to ask - When will Tom's show a comparison with DDR3? I know its on a different chipset, but we would really like to know.
    Different processors make an inaccurate comparison, but you can compare two articles if you like. In fact we used the same benchmarks so that you could.

    On the other hand, comparing two articles to prove that DDR4 currently performs slightly worse than DDR3 is a futile task. Anyone who buys LGA 2011-v3 does so for its added PCIe lane OR added CPU cores. They have no choice other than DDR4.

    Anonymous said:
    They won't do that because then nobody would buy any DDR4 ram
    Right, go ahead and pair DDR3 with your Core i7-5930k. Tell us how it goes. Thanks.
    5
  • codo
    everything on my machine is blitzing as it should be but just cause I know this exists I want it.
    -1
  • CaedenV
    Quote:
    I have to ask: Do these aluminum clip-on heat spreaders help with lowering the temperature or they just like huge spoilers on Honda Civics (I look fast)?

    To say that it did not help at all would be a lie... but it is mostly for looks.

    Quote:
    I have to ask - When will Tom's show a comparison with DDR3? I know its on a different chipset, but we would really like to know.

    When Skylake comes out we should see the first chips that can support DDR3 and DDR4 so we can see some true head-to-head comparisons. I suspect that DDR4 is (at least at the moment) no better than DDR3 in performance and only has an advantage in a performance per watt perspective. But like Crashman said: The first DDR3 modules (and controllers) were horrible and no better than DDR2 at the time. At least DDR4 is improving faster, and dropping in price sooner, than DDR3 did.
    3
  • kardinin
    @blackmagnum Love your comment. Spoilers on a front-wheel-drive car. Hurk hurk hurk.
    4
  • Andy Chow
    DDR4 is a bad joke right now. Kits are expensive, slow and low capacity. Where are the 16GB sticks? Timings of 15-15-15-35-2T?
    5
  • CaptainTom
    Damn these prices need to come down. I would maybe pay half the asking price.
    2
  • Niva
    My phenom 2 with ddr2 is starting to look obsolete... nah, I can get another decade out of it!
    5
  • cd000
    At what point does latency overrule frequency? I'm looking at my Corsair Vengence DDR3 1600 Cas 9 kit, and then looking at the CPUz screenshot showing CAS 15 @ 1066 (2133, yes?) and wonder if my kit might not be faster.

    Secondly, out of curiosity, would it be possible to adjust the timings down to typical Cas 9 settings, and see how fast you could get the DDR4 to go?
    2
  • SinxarKnights
    At what point does overclocking lose any real value? I think this is a prime example of it. You are overclocking this stuff for a increase in microseconds (and nanoseconds... really?) in benchmarks. Who cares? Too much effort for no gain IMO.

    Unless it was the point of the article to demonstrate there is so little difference in DDR4 speed vs performance gain that you might as well go for the cheapest possible.
    4
  • Drejeck
    Quote:
    I have to ask - When will Tom's show a comparison with DDR3? I know its on a different chipset, but we would really like to know.

    Just look at the bandwidth. It doesn't represent real world performance gain though. Ram performance always depend on the CPU controller.
    2
  • Crashman
    Anonymous said:
    At what point does latency overrule frequency? I'm looking at my Corsair Vengence DDR3 1600 Cas 9 kit, and then looking at the CPUz screenshot showing CAS 15 @ 1066 (2133, yes?) and wonder if my kit might not be faster.

    Secondly, out of curiosity, would it be possible to adjust the timings down to typical Cas 9 settings, and see how fast you could get the DDR4 to go?

    You know the math really isn't that hard. DDR3-1066 CAS 5 has the same latency TIME as DDR3-2133 CAS 10. Just divide and multiply...or remember that cycle time is the inverse of frequency (easier to divide and multiply) :)

    DDR3-1600 CAS 9 was our previous "standard". That would be the same latency as DDR3-2133 CAS 12. Not too tough right? That means we need JEDEC to move forward on a DDR4-2133 CAS 12 standard as well.

    But they probably won't. JEDEC approved DDR3-1600 CAS 9 a couple years after DDR3-1600 CAS 11. Immediately following that, a couple companies started making memory that defaulted to DDR3-1600 CAS 9. And then, those companies stopped using the newer standard a reverted to the old.

    Now we have DDR4-2133 CAS 15 and, having learned from the DDR3-1600 CAS 9 fiasco, they'll probably never update it to lower latency. Performance seekers will be stuck with XMP forever.
    1
  • Sabishii Hito
    I'm not surprised the reviewer hasn't been able to find anything to outclass the G.Skill Ripjaws 3000C15 sticks, early batches of those could overclock like nobody's business. The newer batches, not so much.
    1
  • Sabishii Hito
    I do find it odd that most of the recent reviews of this kit actually have been of older batches (2014 earlier than week 40 as evidenced by the first four digits of the serials) and are Ver5.29 (Hynix) while most of the stuff out in the wild now is 1440 or newer and Ver4.23 (Samsung) and is much weaker at running tight timings. Corsair may be cherry picking the kits it sends out for review.
    0
  • Xythras
    I would like Tom to TAKE OFF those aluminum clip-on heat spreaders and see whether is Kingston or Samsung written underneath. My money is on Kingston.
    0
  • Crashman
    Anonymous said:
    I would like Tom to TAKE OFF those aluminum clip-on heat spreaders and see whether is Kingston or Samsung written underneath. My money is on Kingston.
    To be fair, I quit doing the easy ones after opening some of the harder ones caused chips to be ripped-off the PCB :)
    0
  • Sabishii Hito
    Kingston doesn't manufacture memory ICs. DDR4 at this stage is going to be either Samsung, Hynix or Micron, no matter the brand.
    0