Corsair Breaks World Record for Dual-Channel Memory on AMD Phenom II

[btarunr]

Corsair, a worldwide supplier of high-performance computer and flash memory products, today announced that the Corsair Dominator GTX4 ultra-high-performance memory module has achieved a new world record for dual-channel memory frequency on AMD processors.

Using an AMD Phenom II X6 Black Edition CPU, Corsair Hydro Series H50 CPU cooler, and a pair of Corsair Dominator GTX4 memory DIMMs with Airflow memory fan, Corsair Labs was able to achieve the world-record memory frequency of 2287.6MHz. Independent validation of this achievement can be found here.



"The new Phenom II X6 CPUs offer a quantum leap in overclockability for the AMD platform," stated Jim Carlton, VP of Marketing at Corsair. "The combination of the new CPU core and Corsair's most aggressively sorted DIMMs results in some truly amazing memory performance."

Key components used in setting this record include the following:

• AMD Phenom II X6 1095T Black Edition CPU
• Asus Crosshair IV Formula Motherboard
• 4GB Corsair Dominator GTX4 memory (2 DIMMs)
• Corsair Hydro Series H50 CPU Cooler
• Corsair Nova Series V64 SSD
• NVIDIA GeForce 6600GT GPU
• Corsair Professional Series 850HX PSU

Complete details regarding the test setup, methods, and results can be found here. For more information on Corsair Dominator GTX memory, please visit this page.

View at TechPowerUp Main Site

 

[DigitalUK]

excellent work corsair, i do want a 1095T but think i hold out until bulldozer,but why the 6600gt seems like buying a ferrari and putting caravan wheels on it.

 

[Delta6326]

nice over clock

but is it just me or is that pi a little slow? i do mine in like 17sec. and i got slower stuff or am i looking at it wrong

 

[Initialised]

20 LinX and 24 Hours Prime Blend and 12 Hours MemTest or it doesn't count.

excellent work corsair, i do want a 1095T but think i hold out until bulldozer,but why the 6600gt seems like buying a ferrari and putting caravan wheels on it.

1095T is what my ES 1055T shows up as.

 

[cadaveca]

Northbridge speed is almost MORE important than raw mhz. 9-12-9 sucks. I'd take 900mhz CAS6 first.

After spending several hours of testing timings, sub timings, voltages, multiple processors and various frequencies, a new World Record for AMD Memory frequency was achieved! The maximum speed achieved was 2287.6 MHz at Cas 9. The two other processors were able to produce similar memory clocks at 2271.6 MHz. These results, across several processors, would suggest that it is running into a limitation of the memory controller on the new AMD Phenom II X6 processors.

But, after reading this, I am very happy. Not every cpu can do this(but at least 3 others can), so they are really highlighting how useless this mem is for AM3...


Pretty bold to claim, to have found the IMC limitation...that sounds like a challenge, to me!

 

[Mussels]

nice over clock

but is it just me or is that pi a little slow? i do mine in like 17sec. and i got slower stuff or am i looking at it wrong

intel was always faster in Pi than AMD.

it wasnt because AMD were slower chips, they were just slower at Pi for whatever reason.

 

[Kitkat]

2500 = new ceiling, not 18XX its not just the chip.

 

[DigitalUK]

Super pi is optimized for Intel and means very little kind of like 3dmark vantage and ati, its very good for testing memory stability tho.

 

[btarunr]

intel was always faster in Pi than AMD.

it wasnt because AMD were slower chips, they were just slower at Pi for whatever reason.

Not just Pi. AMD chips are known to be slower in most popular math benchmarks, be it from SANDRA, Geekbench, Fritz Chess, or Everest. It's up to the mark only in Sciencemark.

 

[Mussels]

superpi is very good for comparing stock vs OC'd and so on, just not for comparing between CPU types.

 

[btarunr]

superpi is very good for comparing stock vs OC'd and so on, just not for comparing between CPU types.

SuperPi is a math test, just like Prime, Sin-Julia, and so on. Results can be used to compare across CPU types as well, like most CPU reviewers do.

 

[Mussels]

SuperPi is a math test, just like Prime, Sin-Julia, and so on. Results can be used to compare across CPU types as well, like most CPU reviewers do.

yeah but if one architecture is superior at that test, its an unfair comaprison. Its fine as part of a range of tests, but should not be considered purely on its own.

 

[btarunr]

yeah but if one architecture is superior at that test, its an unfair comaprison. Its fine as part of a range of tests, but should not be considered purely on its own.

No, that's flawed logic. If one architecture is superior in a broad range of math tests, it's just superior at math performance. As long as both architectures are able to run the tests, the grounds are even.

 

[Kitkat]

agree now we are just designing chips to beat a test. no thanx. benching software isnt scrutinized enough either its taken for granted.

 

[Mussels]

No, that's flawed logic. If one architecture is superior in a broad range of math tests, it's just superior at math performance.

well yes, but dont take it as OVERALL performance. dont assume intel is better for gaming, just because it whups AMD in calculating pi to over 9000 digits.

the question asked was 'why does that score look low?' the answer was cause its faster on intel.

 

[btarunr]

well yes, but dont take it as OVERALL performance. dont assume intel is better for gaming, just because it whups AMD in calculating pi to over 9000 digits.

the question asked was 'why does that score look low?' the answer was cause its faster on intel.

I never referred to overall performance. My argument is limited to math performance. AMD processors are traditionally known to have weaker floating point units than Intel.

 

[Mussels]

I never referred to overall performance. My argument is limited to math performance. AMD processors are traditionally known to have weaker floating point units than Intel.

and i was never talking purely about FPU/math performance.

case closed.

 

[btarunr]

and i was never talking purely about FPU/math performance.

case closed.

Let me refresh your memory, you were talking about SuperPi, and how different brands of processors perform in it, and how performance at SuperPi cannot be compared between different brands of processors. That notion is incorrect.

Case closed.

 

[Mussels]

it can be COMPARED, certainly. but its apples vs oranges.

anyway, staff bickering should be avoided on the forums lets get it back on topic.

 

[Kitkat]

u 2 are funny. I like turtles.

 

[btarunr]

it can be COMPARED, certainly. but its apples vs oranges.

anyway, staff bickering should be avoided on the forums lets get it back on topic.

Can game frame-rates be compared between systems with Intel and AMD processors? Yes (game tests in CPU reviews); then yes you can compare math performance (like most CPU reviewers do).

 

[demonkevy666]

Northbridge speed is almost MORE important than raw mhz. 9-12-9 sucks. I'd take 900mhz CAS6 first.


But, after reading this, I am very happy. Not every cpu can do this(but at least 3 others can), so they are really highlighting how useless this mem is for AM3...


Pretty bold to claim, to have found the IMC limitation...that sounds like a challenge, to me!

at those rams speeds you would need about 6.4ghz cores just o make use of the 4.5ghz Nbs speed for that ram speed of 2.2ghz.

 

[cadaveca]

Exactly, and even under LN2, you'll barely need that. 6.4ghz and 4.5NB isn't exactly an easy thing. I mean..it's CAS9, FFS. They are running 2800mhz NB. 3600mhz NB and 900MHZ CAS6ram will probably perform better. 1150mhz ram is cool and all, but only those going for WR might be interested in buying(and few of those, even). They probably know that though.

 

[arnoo1]

Super pi is optimized for Intel and means very little kind of like 3dmark vantage and ati, its very good for testing memory stability tho.

also it works with L2/L3 cache
correct me if i'm wrong
I heard that once

 

[Assimilator]

also it works with L2/L3 cache
correct me if i'm wrong
I heard that once

Nice overclock, but those memory timings are horrific; this RAM isn't much use unless you're going to be pushing it to 2500MHz every day.

As for Super PI - my Yorkfield is a generation behind but manages to do 1M in under 13 sec (50% faster than a 6-core Phenom, and with 50% fewer cores) - my CPU has triple the L2 cache of the Phenom as well! I've never understood why AMD skimps on cache, it seems such a waste...