R4 Extreme Ram 2400Mhz slower than 2133Mhz

[BRUNEZ]

Hi all,
I have a problem with a R4 EXTREME, I start telling you that operative system Windows 10 is clean and the 3930K has always worked at 4,5 MHZ with 1,31 of Vcore.
At the moment I fit 16 gb. (4x4) of Corsair Dominator-GT of 2133 mhz with latency 9-11-11-27 and CR2, filling all 4 red slot of Quad Channel, using the profile XMP.
With this configuration at the moment Linx it gives me back at each turn around 156 Gflops., on the contrary the test Ram of AIDA 60.000 Mbs. , both in writing than reading.
For information, I tell you the BCLK is 100 AND CPU STRAP is set at 100, on the contrary , the connection memory of DRAM:FSB is 32:3 .
It is since few day I replaced the kit 2133 with a kit 2400 , having latency 11- 13-13-31 CR2

The Ram in object are Dominator Platinum with code CMD16 GX3 M4 A2400 C11, also them in kit of 4x4 fill all 4 red slot of MOBO.

Activating profile XMP setting manually the latency and putting in mode 1 LINXS gives me 148 Gflops and AIDA 58.000 MBS in writing and reading.
Unfortunatley it is clear thet even if memories have 267 MHZ more, is obtaining results less performanced of previously fit Ram
The strange thing is that if I set the 2400 Dominator manually at 2133 facing the same speed of previously momories , the result of LINX AND AIDA does not change, I always obtain 148Gflops and 58.000 mbs.

I used , to do these test, both profile XMP than manually setting RAM values, (Voltage, latency, ecc.. )

What I ask is why I have a performace failure using Ram that should give me higher values.

Thank you

[Arne_Saknussemm]

Frequency is not the only metric you have to look at for performance. You have to look at timings too. In this case CAS 9 at 2133 will be faster than CAS11 at 2400

And all the subtimings can play an important role too.

As a crude guide you can use (CAS/frequency) x 2000 gives you a number in nanoseconds for comparing speeds.

[BRUNEZ]

Arne Saknussemm wrote:
Frequency is not the only metric you have to look at for performance. You have to look at timings too. In this case CAS 9 at 2133 will be faster than CAS11 at 2400

And all the subtimings can play an important role too.

As a crude guide you can use (CAS/frequency) x 2000 gives you a number in nanoseconds for comparing speeds.

Yes, i forgot to write that i tried 2400Mhz. memory to 10-11-11-27 ( at 9 the system don' t boot ) and LINKS give me 149 Gflops… only 1 more, i don' t think it' s only a timing\subtimings issue.

Platinum 2400 11-13-13-31 148Gflops

Platinum 2133 (downclocked) 11-13-13-31 148Gflops --- This is very strange: Less Mhz, same timings but same result thna 2400 XMP

Platinum 2400 10-11-11-27 149Gflops --- Gain only 1 Gflops

GT 2133 9-11-11-27 156 Gflops --- XMP default

[BRUNEZ]

Arne Saknussemm wrote:
Frequency is not the only metric you have to look at for performance. You have to look at timings too. In this case CAS 9 at 2133 will be faster than CAS11 at 2400

And all the subtimings can play an important role too.

As a crude guide you can use (CAS/frequency) x 2000 gives you a number in nanoseconds for comparing speeds.

Ok, thank you for the information, i' ll try to do some test.

[Arne_Saknussemm]

Well as I said...it is not only primary timings...secondary and tertiary timings, 1T vs 2T etc. etc.... they all have an effect.

One thing to point out is that benchmarks may not reflect real word performance...