Corsair Dominator GTX2 “Hyper” vs G.Skill Flare
The Hyper theory :
The highly acclaimed and ultra-vulnerable Hyper rams are a benchers dream. Though this is all at a big cost : A gaping hole in your wallet. Though how much is the performance difference between e.g. a Corsair GTX 2 kit and this 4gb G.Skill Flare kit. Before we continue the GTX2's cost about 200 euros per 2GB dimm (yes per dimm), the 2000C7 G.Skill Flare kit retails at a price of 180 euro's. On top of that the Hypers are getting very rare, making the prices go through the roof. I also wonder if one of our precious Hyper dimms bites the dust, what the RMA will bring.
Let's compare the G.Skill (6-9-6-24) now against the Corsair Hyper rams at Cas6-6-6-18 and Cas7-7-7-21. We took the fastest timings for the Flares, to give the Hyper kit a good run for its money. This time only 1600mhz and 1800mhz are compared as it was impossible to get SuperPi 32M running stable on my setup.
At 1600mhz the, Hypers are barely faster than the Gskill Powerchips based Flare kit. The bandwidth provided to the CPU's memory controller can't take advantage of the slightly tighter timings.
Running Cas 6 at 1800mhz makes the Hyper kit scream. A 14 sec gap, is huge in my book. Even when running the Hyper kit at Cas 7 it's still topping the Flare kit at Cas6. But don't forget the price tag…
Everest shows us a little better performance for the Hyper kit. Though the difference is minimal here. I only adjusted the basic timings of the rams. Maybe further sub-timing tweaking could increase performance for both involved parties.
1800mhz still in favor of the Hyper kit. Though looking at these numbers, I can hardly explain the mega SuperPi advantage. The Latency (not displayed in the graphs) is superior with the Hyper kit. Let's move onto the High Definition Encoding test.
The Hyper kit still beats the Flare kit, though differences are again minimal. Price wise it's another story. At 1800mhz the Cas 6 tests reveal again the superiority of the Hyper IC's.
When benching the games (via their included benchmark tool) the Hypers are faster, though again it's not an impressive win. Both ram kits are pushed near their limits here at the stock rated Vdimm. For daily usage I can't recommend the Hyper rams. This due to the skyhigh retail price. The powerchip based G.Skill kits provides a nice alternative at an affordable price. Serious benchers again have no real option. If they can afford it, it's Hypers all the way. This when being on the hunt for that extra 10th of a second or the last extra 3dMarks.
The game tests at 1800mhz show the small benefit, it's a mere one FPS that you gain. For gamers again, no need to cash out and settle with something far cheaper to power your rig. The speed advantage will just in-between your ears and not noticeably displayed on your monitor.
As a final OCing note on this page : Eat your heart out Hypers, 2200CL7 action on 1.65Vdimm on air !
(click to zoom)
Question : Is it the board, RAM or the CPU's IMC that caps out ? I've been gaming on these particular speeds for a few weeks now, without any issues. Though the stress of Prime95 or LinX brings the system on its knees...
Now comes a small but part. I've been doing some chit chat with mista Chew* (one of the best AMD tweakers to be found on Xtreme Systems forum) He referred to some issues with stability testing. Most of the daily overclockers (read not the LN2 guys) still use prime 95 or an Intel Burn Test program as stability test. I got a Prime 95 Blend screenshot just before it BSOD on me. This with the CPU at 4ghz and the rams at the rated 2000Mhz settings. When I dropped the ram clocks, there seemed to be no issue. Contradicting the prime95 issues, testing with Folding at Home and Memtest in windows didn't yield any errors.
(click to zoom)
So time to go a little deeper and after some fiddling around a small bump in Vcore and NB voltage got it LinX stable. It used to crash after a few seconds without the above alterations. I included my voltage settings via the TurboEVO tool, hoping it will help your.
And a quick spin at 2100Mhz LinX
I'm really not a fan of these stress testing tools. As it required me to feed my daily stable rig with more Vcore, NB and HT voltage. This just for a screenshot… I’d rather be gaming then :)