cancel
Showing results for 
Search instead for 
Did you mean: 

Memtweakit under Windows 8 Release Preview

Thunderman
Level 9
Hello ROG community,

I'm only writing to share my experience with Memtweakit when running under Win8 x64 RP
on an Asus ROG Rampage 4 Formula board.

on Win 7 x64 and with the same default XMP profile for my Corsair Dominator GT memory i used to get around 45700 efficiency score.

Now when running Memtweakit on Win8 X64 with the same exact bios settings i'm only getting 4392 efficiency points
yet my actual DRAM performance in other benchmarks is even better under Windows 8 in addition to a large improvement in CPU performance (3930K).

This leads me to believe that Memtweakit needs some Tweaking to be Win8 compatible.

Thank you,

Thunder
14,162 Views
8 REPLIES 8

Thunderman
Level 9
EDIT : the latest version of MemTweakit can run under Win 8 RP x64 and provide score close to the Win7 section
yet i believe that Memtweakit through its DRAM efficiency does not really represent the actual increase in the memory performance achieved through the tweaking ie : on standard xmp profile i had an efficiency score of 45K approx and just but modifying 2 timings my score was up to 104K while the increase in benchmarks performance was barely noticeable few 1/10 of a percent while the efficiency score more than doubled.

Anyone, correct me if wrong, I think it's all because of MemTweakit scoring model, which seems to be almostly based on timings. At high freqs timings affect overall perf poorly. Also, ovc my RAM to 2200 from 2133 seemed to contribute very little to my score at MemTweakit, less then 100 pts I guess, while lessing Dram refresh interval affects result vividly.
All in all, its just another dimension of toughness (one's not just longer, but thicker as well)))

Pluribus Impar[/HR]

ElenHim wrote:
Anyone, correct me if wrong, I think it's all because of MemTweakit scoring model, which seems to be almostly based on timings. At high freqs timings affect overall perf poorly. Also, ovc my RAM to 2200 from 2133 seemed to contribute very little to my score at MemTweakit, less then 100 pts I guess, while lessing Dram refresh interval affects result vividly.
All in all, its just another dimension of toughness (one's not just longer, but thicker as well)))



Several of us helped with the equation. The original was based on real-world values broken down by timings. However, we wanted to make the scoring accessible by all and not too elite so it was toned down and made flexible enough to be fun rather than absolutely realistic. If we had kept it realist you would have had to buy the best kits out there and that would have put a lot of people out of the running.

Just look at it as a fun thing rather than accuracy.

Raja@ASUS wrote:
Several of us helped with the equation. The original was based on real-world values broken down by timings. However, we wanted to make the scoring accessible by all and not too elite so it was toned down and made flexible enough to be fun rather than absolutely realistic. If we had kept it realist you would have had to buy the best kits out there and that would have put a lot of people out of the running.

Just look at it as a fun thing rather than accuracy.


Defo agreed on the Fun rather then accuracy, Got Gskil Trident mem 32GB running at 2400Mhz 10,11,11,2 timings and get a score of 49K.

Modern CPUs cannot utilize 2400 RAM freq effectively. At present, everything that runs over 1866 (or even 1600) is PR-made. Actually, you only get +2-4% perf (and lots of to self-satisfaction) at doubled or even tripled cost. For me, it's more reasonable just to have more "more regular" RAM, and as far as it's done of the same chips as 2400 one, you still can ovc it and run benchs for sweet scores. As for MemTweakit, the real fun begins when Trident and other self-styled stuff are beaten with very modest RAM.(And money cannot buy that) And I believe accuracy is also out there))))

Pluribus Impar[/HR]

ElenHim wrote:
Modern CPUs cannot utilize 2400 RAM freq effectively. At present, everything that runs over 1866 (or even 1600) is PR-made. Actually, you only get +2-4% perf (and lots of to self-satisfaction) at doubled or even tripled cost. For me, it's more reasonable just to have more "more regular" RAM, and as far as it's done of the same chips as 2400 one, you still can ovc it and run benchs for sweet scores. As for MemTweakit, the real fun begins when Trident and other self-styled stuff are beaten with very modest RAM.(And money cannot buy that) And I believe accuracy is also out there))))


Not really fussed if the memory cost £150 more then other brand or make when my system cost 3K. As for memory not goona make much difference from 2000Mhz to 2400 Mhz a couple of FPS more in a game. As for memtweak, like raja said it aint accurate more geared to fun. Id leave memory benchmark to the likes of aida and or pass mark etc.
How you choose to take Memtewakit and its numbers is your opinion and choice. What i take is in realtime... in game,rendering,flaoting point etc.

X-ROG
Level 15
There's currently no support for Win8. That will come after retail launch later this year.

I'm not sure how it's calculated as it was done by the OC team, but it looks that way from the results Elen. I'll bring it up with them to see if there can be a middle ground between frequency and timings.

ClairHardesty
Level 7
There is a fundamental problem with the Mem TweakIt equations. The DRAM Refresh Interval belongs on the top, not the bottom of the formula. Currently a high interval results in a lower score and vice versa. Unfortunately, lowering the interval to get a higher score reduces memory throughput because the refresh cycles are happening more often. My highest score (126246) came at the cost of over half of my throughput (as measured by SiSoft Sandra), reducing it from ~52GB/s to ~24GB/s. The 52GB/s number was obtained with DRAM Refresh Interval set to 10,000.
ASUS RIVBE Win10 Pro 64 bit Preview Fast Ring
Intel Core i7-4930K @ 4.2GHz 1.2V
G.SKILL Ripjaws Z 4x8GB 2133MHz 9-10-10-24 2T 1.95V
2xEVGA GeForce GTX 980 SC 4GB, in SLI
2xCustom water loops (CPU-VRM-PCH-Mem & GPUs)
3xASUS VG278H 27" 120Hz 3D monitors
2xLogitech HD Pro Webcam C920
720GB total in 2 SSDs
SanDisk Extreme: 480GB Boot & 240GB Temp

44TB total in 9 HDs
WD Red: 4x3TB, WD Black: 2x8TB & 2x6TB, Seagate Barracuda: 1x4TB

Rosewill LIGHTNING 1300W
CyberPower PR2200SWRM2U 1500W UPS