Specs, Photos, and Thoughts on Memory Speed
| CAS Latency | 11 cycles |
| Row Cycle Time (tRCmin) | 42.125ns (min) |
| Refresh to Active/Refresh Command Time (tRFCmin) | 260ns (min) |
| Row Active Time (tRASmin) | 35ns (min) |
| Maximum Operating Power | 2.160W (per module) |
| Operating Temperature | 0º C to 85º C |
| JEDEC SPD Programming | DDR3-1600 CL11-11-11 @ 1.50V |
| X.M.P. Profile #1 | DDR3-2666 CL11-13-13 @ 1.65V |
| X.M.P. Profile #2 | DDR3-2400 CL11-13-13 @ 1.65V |
Memory Speed
Before continuing, I would like to briefly discuss memory speeds. For some time now, DDR3-1600 has been the standard on which virtually all desktop RAM is based, whether stock or overclocked. JEDEC, the council responsible for memory specifications, lists 1600MHz at a nominal voltage of 1.50V, and the most common “binning” of this speed has a CAS latency of 11. I mention this because some variant of this standard is present in virtually all shipping desktop systems, and all overclocked memory sold.
The HyperX kit running at a full 1333MHz (DDR3-2666)
On first boot the memory will always default to the JEDEC programmed speed (far left)
Yes, whether it’s 1866MHz or 3000MHz, your memory is programmed at the lowest level to operate at much lower speeds. This often causes confusion and frustration for end users, who are expecting the memory to perform at the rated speed right out of the box.
To make use of higher memory speeds, CPU’s need to have support in built in to their memory controllers. AMD has listed official 1866MHz memory support with their APUs and FX series CPUs since 2012, and Intel, though not supporting 1866MHz DDR3 officially for desktop parts, does support the faster standard on the server/workstation side. So for desktop users, the new Ivy Bridge-E powered Core i7 processors for the X79 chipset are Intel’s only chips officially supporting the faster 1866MHz memory – without overclocking, of course.
(In 2013, JEDEC revised the specification of DDR3 to include standards for 1866MHz and 2133MHz speeds, though memory manufacturers have yet to make use of these standards for desktop users, and have relied on overclocked profiles and custom timings to this point.)
Standardized or not, most of the retail desktop motherboards sold today offer memory overclocking support. To make it easier to exceed the JEDEC specs programmed into a module’s ‘SPD’ chip, there is Intel’s “extreme memory profile”(X.M.P.), and to a much lesser degree AMD’s “AMP” profile. These overclocked memory profiles allow manufacturers to eliminate the need to manually adjust speeds, voltage, and timings – when everything goes to plan, of course.
One quick note about latency, and it is simple: the lower, the better. It would take a separate (and probably very boring) study to really break down the differences between timings at various speeds, and I won’t get into any of that here. Higher speeds do make up for a higher latency number in most cases, but between two otherwise identical DIMMs, lower latency will provide higher bandwidth.
(The upcoming DDR4 spec promises to improve efficiencies, so we might see some nice performance gains with the next generation of processors.)
Without further exposition on the state of DDR3 in the industry, let’s put this kit on the test bench!
why not bench it against the
why not bench it against the IGP instead of cpu bound tasks? or better yet try to get it stable on an APU.
would it be possible to run
would it be possible to run some gaming benchmarks on amd apu’s as you suggested? Sure its unlikely that someone would buy ram this pricy and use it on a apu but it would be very interesting to see how it would affect a kaveri or richland part, just for shits and giggles?
pricey? its on sale for $100,
pricey? its on sale for $100, not too expensive for very highspeed ram.
well yea if I was buying ram
well yea if I was buying ram today, but the msrp is $300, this is just a great sale on amazon
Please as suggested test with
Please as suggested test with an AMD APU. If it is only $10 more than regular memory then many, many people will go for this and an AMD APU. It makes little sense to test this with a fast i7 and dedicated GPU as we know what the resuls will be already. Test where faster memory will make the biggest difference and that will be with an APU. With this and A10-7850K you could get some decent entry level performance, overclock the integrated GPU and 1080p gaming at low / medium details are a very real option.
Oh, see here for an
Oh, see here for an example:
http://www.overclockers.com/amd-a107850k-kaveri-apu-review/
I agree with the comments
I agree with the comments about APU testing, and trust me, I want to see how it would improve the graphics performance myself! I mentioned in the article that the memory would be worth revisiting with an APU, I just didn’t have the necessary components.
I’m going to put together a test system with a 7850K and see if I can get this memory stable at 2666MHz. I don’t know specifically when I’ll have this done, but I’ve already ordered the parts!
Sounds like a great idea
Sounds like a great idea Sebastian! It might be interesting to look at some GPU overclocking for the 7850K. On overclockers they got some really decent performance around the 1 GHZ mark.I know that is kind of out of the scope of the article but it shows what a ‘budget’ system can do with the right memory and GPU / CPU config. Just a thought anyway.