Super Talent Chrome 2000CL8 : Review
Manufacturer: Super Talent
Kit: 3x2GB Kit
IC: MGH-E Hyper
It seems that as of late a lot of people are taking interest in the remaining kits of Super Talent memory that are available on ebay and ewiz. These kits get recommended pretty often around here.
A lot of people ask about these kits so I decided, as someone who owns a kit of WB200UX6G8 (2000 8-8-8-24, to adress some of the questions some people might have about these kits.
If you are debating whether you think finding a kit of these is worth it, then stay tuned, because I will cover the following:
Using what I have written in this guide, I hope those of you who have questions about this kit will be able to determine wether or not it is worth your time and money. If you are impressed by its benchmarks and overclockability, then you know what to do!
For the record, all the speeds/timings I use in this test will be tested to make sure they are stable, using a plethora of applications, such as memtest86+, prime95 blend, linx, superpi32m and a few more tests.
I don't want people thinking they can run certain speeds stably when that is not the case.
About the kit:
Super Talent Chrome 2000CL8
Available retailers: ewiz, amazon, ebay.
First off, the main draw of this ram kit is the fact that it uses mgh-e hypers, a particular type of IC that is end of line and is very hard to find. Super Talent sells various binnings of these chips in their ram kits, some of which advertise themselves as 2000 7-7-7-20 and 2133 8-8-8-24. I won't be looking at these kits in this review.
The reason these kits have been getting so much attention is because mgh-e hypers are able to pull of timings and speeds no other CURRENTLY MANUFACTURED IC/RAM KIT is able to do. 2000 8-9-8-24 is widely available in Corsair dominators, but aside from some older kits, Super Talent Chrome is presently the only ram kit you can get such timings in.
Its obvious why these kits would be ideal. Tighter timings and high speeds, this obviously will mean the ram will be faster.
The default voltage for 2000 8-8-8-24 on this parciular kit is 1.65v, standard for x58 memory, and the Command Rate is 2T.
So, if you are interested thus far, you are probably asking: How well do they perform? How does their performance compare to looser timings? What are some stable overclocks I can do with this ram? How well do they scale?
I intend on covering most of these things and I hope you will have atleast one question that gets answered.
Rampage III Extreme
Enermax Revolution85+ 1050
Megahalems + San Ace Fan
I will be using a Bloomfield Intel i7 950 for this review.
Uncore & CPU:
|Uncore is a term used by Intel to describe the part of a microprocessor that is not the core. The core contains the components of the processor involved in executing instructions, including the arithmetic logic unit, floating point unit, L1 and L2 cache. Uncore functions might include L3 cache, the on-die memory controller, and other bus controllers such as PCI Express.|
The i7 950 and Bloomfields in general require that the North Bridge Frequency of the IMC is atleast twice of your RAM Frequency. For the i7 950, for example, the CPU needs an uncore of atleast twice the ram speed to stabilize. For example, ram at 1600MHz requires at the very least 3200MHZ of Uncore / North Bridge Frequency to stabilize.
However, because the Uncore not only covers the L3 Cache but the Memory Controller as well, it is not uncommon to see an increase in uncore to result in higher bandwith / lower latencies on your ram overclock.
For example, here is two exact setups: one of which has an uncore of 3200MHz, and the other 3800Mhz. The Ram speeds and settings are the same, but notice the following:
It is clear that the increase in uncore helped out a little bit.
Because higher ram frequencies require the IMC to run faster to stabilize, you may find that your CPU is what could potentially limit your RAM overclock.
For example, in order for me to run 2000Mhz of ram, I need to increase my Uncore to 4ghz. This is far too much stress for my IMC, so I have to increase my QPI/VTT (IMC) voltage to allow for my chip to stabilize my ram oc.
This is the reality for many i7 chips: the uncore requirement for high frequency ram is just too damn much stress on the IMC.
That being said, its not impossible to stabilize, you just need to give more QPI/VTT voltage, something that is rather sensitive.
Because this varies between chips, I will not include my QPI/VTT required to stabilize a chip, because by and large its irrelevant unless someone has an exactly identical chip as mine. But just keep this in mind when thinking about buying high frequency ram: know what your CPU is capable of.
Anyway, because the Uncore frequency can have a play in the numbers that get scored in synthetic benchmarks, I will use the x2 rule for all ram speeds I test, so, for example:
1600mhz = 3200mhz Uncore
2000mhz = 4000mhz Uncore
I will however show results of how the ram performed by going above and beyond this x2 rule, but not beyond my CPU's limit.
For me personally, I run my QPi/VTT/IMC at 1.35v, a nice number. Going between 1.4-1.45 is fine, but I personally wouldn't recommend going over for 24/7 unless you are hardcore about your memoc.
Because QPI requirements vary, I will increase my QPI to accomodate for at the very least 4000Mhz of Uncore. Going over is not possible to stablize on my CPU without an unsafe amount of QPI, so unfortunately for the sake of my chip, this reviews scope is limited to 2000Mhz.
I do intend on trying out my ram on a stronger i7, so stay tuned, because it shouldn't be long before I add results for 2000mhz+.
But until then, this review is intended for moderate/intermediate overclockers, who don't wish to overvolt their ram or qpi but want to know what stable overclocks they can get with this kit of ram.
My CPU is a 4ghz stable chip. For consistensy reasons in our sysntehtic tests, I will be running my cpu at this speed throughout. However, because I want to test a wide range of frequencies, I will have to lower/increase my CPU clock every now and then.
For consistensy reasons I will keep my ram within a 200mhz range, 3.8-4.0ghz. This will prevent any results in synthetic benchmarks from being skewed.
Hopefully this OC range will allow me to test a wide variety of frequencies, as it should!
I intend on adding to this review as time goes on, but as I said, I am not a high voltage/extreme overclocker.
I am meerely testing this ram at 1.65VDIMM, to 1.7VDIMM, no more, no less.
If you want to see how mgh-e hypers scale with preposterous amounts of voltage, there are plenty of benchmarks and scores conducted by extreme overclockers / coolers that you can find on hwbot , xs, etc).
So for now, 1.65-1.7VDIMM is what all speeds will be tested using.