Today's test will be based on a Corsair Vengeance LPX 16GB DDR4-2666 CL15 1.2V memory kit designed for X99 platform.
I don't remember when I was testing last Corsair kit and DDR4 availability is really limited so I picked one of the Vengeance series. There is also higher - Dominator Platinum, but it cost much more.
Let's focus on the memory itself.
Corsair Vengeance LPX 4x4GB DDR4-2666 CL15 is based on a Hynix IC as we can compare version number - 5.29 to the Corsair memory IC list that we can find on the forums. I don't think it was changed comparing to DDR3 list.
Four 4GB modules of this kit are single sided so we can expect that kit based on 8GB modules will be double sided and in the future we will see higher density DDR4.
I had a chance to pick memory in black or red heatsinks. I've decided on a black version. Some photos of the package and memory itself are below. Not the best quality but I bet that most interested readers will check manufacturer's site and some reviews available in the web ( also on Overclockers ).
Test platform will be based on i7 5820K CPU and so far the cheapest ( but not the worst ) X99 motherboard - MSI X99S SLI Plus with the latest BIOS/UEFI 1.10. 6 core 5820K CPU will let me compare DDR4 performance directly to DDR3 based on 4930K CPU.
What I wasn't expecting are the XMP profiles ( in 2.0 version ). Vengeance LPX has two XMP profiles:
XMP1: 2666 15-17-17-32 1.20V
XMP2: 2800 15-17-17-32 1.35V
Yes, memory has 2666 clock on the box/label but we can also use 2800 profile at slightly higher voltage. That's quite nice.
Below we can see performance at the XMP1 profile and also readings from AIDA64 and CPU-Z. Both programs are not perfect for X99 platform but in this case CPU-Z has less problems to read profiles and actual memory settings. AIDA64 is not prepared for X99 platform and new CPUs so we have to wait for an update to fix some problems.
and XMP2 at 2800 15-17-17-32 1.35V
Overclocking results
3000 16-16-16-38 2N 1.35V
Performance at 3000 is actually lower than at 2666/2800 even though I've raised CL by 1 but lowered all other timings. On the other hand we see that this Vengeance memory has some potential.
So far I couldn't set much higher clock than 3000 and I think it can be related to the motherboard. However I will for sure try some more later and now will be some lower clocked results at tighter timings.
2800 CL15-15-15-32 1N 1.35V
Looks like at 2800 we can still tighten the timings. I was able to boot and run AIDA64 or Maxxmem at 2800 14-14-14 but it wasn't stable and sometimes board was seeing memory in triple channel. Maybe I will back to CL14 later. Better DDR3 series were running at 2800 12-14-14 so I think that similar settings on DDR4 would be already good.
2666 CL13-13-13-32 1N 1.35V
2666 CL13 is probably best performing setting. Timings are also much tighter than on declared settings - 15-17-17. I will try to tighten timings some more later. 2666 is also the highest setting possible at 100MHz bclk. For some reason this platform has maximum memory ratio of x26.66. I thought it will be higher with declared clocks of 3200+. XMP2 setting is already running at 127MHz bclk which is also in XMP profile.
2400 12-12-12-32 1N 1.35V
2400 seem already low clock but as we see performance isn't that bad comparing to higher clocks. Actually I thought it will be worse.
At the end SPD setting to compare all above results to the standard profile of 2133 15-15-15 1.20V.
This setting is also suprisingly good. AIDA64 bandwidth is low but other benchmarks are not looking as bad as expected.
I will try to add some more results later when I play some more with sub timings.
Also as always feel free to comment this thread.
If you can't see what is on the screenshots then try to open pictures in a new window. Maybe someone will fix that as on the old forums it was working good.
I don't remember when I was testing last Corsair kit and DDR4 availability is really limited so I picked one of the Vengeance series. There is also higher - Dominator Platinum, but it cost much more.
Let's focus on the memory itself.
Corsair Vengeance LPX 4x4GB DDR4-2666 CL15 is based on a Hynix IC as we can compare version number - 5.29 to the Corsair memory IC list that we can find on the forums. I don't think it was changed comparing to DDR3 list.
Four 4GB modules of this kit are single sided so we can expect that kit based on 8GB modules will be double sided and in the future we will see higher density DDR4.
I had a chance to pick memory in black or red heatsinks. I've decided on a black version. Some photos of the package and memory itself are below. Not the best quality but I bet that most interested readers will check manufacturer's site and some reviews available in the web ( also on Overclockers ).
Test platform will be based on i7 5820K CPU and so far the cheapest ( but not the worst ) X99 motherboard - MSI X99S SLI Plus with the latest BIOS/UEFI 1.10. 6 core 5820K CPU will let me compare DDR4 performance directly to DDR3 based on 4930K CPU.
What I wasn't expecting are the XMP profiles ( in 2.0 version ). Vengeance LPX has two XMP profiles:
XMP1: 2666 15-17-17-32 1.20V
XMP2: 2800 15-17-17-32 1.35V
Yes, memory has 2666 clock on the box/label but we can also use 2800 profile at slightly higher voltage. That's quite nice.
Below we can see performance at the XMP1 profile and also readings from AIDA64 and CPU-Z. Both programs are not perfect for X99 platform but in this case CPU-Z has less problems to read profiles and actual memory settings. AIDA64 is not prepared for X99 platform and new CPUs so we have to wait for an update to fix some problems.
and XMP2 at 2800 15-17-17-32 1.35V
Overclocking results
3000 16-16-16-38 2N 1.35V
Performance at 3000 is actually lower than at 2666/2800 even though I've raised CL by 1 but lowered all other timings. On the other hand we see that this Vengeance memory has some potential.
So far I couldn't set much higher clock than 3000 and I think it can be related to the motherboard. However I will for sure try some more later and now will be some lower clocked results at tighter timings.
2800 CL15-15-15-32 1N 1.35V
Looks like at 2800 we can still tighten the timings. I was able to boot and run AIDA64 or Maxxmem at 2800 14-14-14 but it wasn't stable and sometimes board was seeing memory in triple channel. Maybe I will back to CL14 later. Better DDR3 series were running at 2800 12-14-14 so I think that similar settings on DDR4 would be already good.
2666 CL13-13-13-32 1N 1.35V
2666 CL13 is probably best performing setting. Timings are also much tighter than on declared settings - 15-17-17. I will try to tighten timings some more later. 2666 is also the highest setting possible at 100MHz bclk. For some reason this platform has maximum memory ratio of x26.66. I thought it will be higher with declared clocks of 3200+. XMP2 setting is already running at 127MHz bclk which is also in XMP profile.
2400 12-12-12-32 1N 1.35V
2400 seem already low clock but as we see performance isn't that bad comparing to higher clocks. Actually I thought it will be worse.
At the end SPD setting to compare all above results to the standard profile of 2133 15-15-15 1.20V.
This setting is also suprisingly good. AIDA64 bandwidth is low but other benchmarks are not looking as bad as expected.
I will try to add some more results later when I play some more with sub timings.
Also as always feel free to comment this thread.
If you can't see what is on the screenshots then try to open pictures in a new window. Maybe someone will fix that as on the old forums it was working good.
Last edited:
As i have seen and found myself was 2133 with tighten timings and also you could only go so far with the Bclk @ 100. Would the X99 be around the same even though you can go higher up to 3000 etc. The CAS at the moment does not look to great i hope they can lower it down to Cas 9 etc. 
