AOA Forums

AOA Forums (http://www.aoaforums.com/forum/)
-   General Hardware Discussion (http://www.aoaforums.com/forum/7-general-hardware-discussion/)
-   -   Bad G.SKILL-F2-8500CL5D-4GBPK RAM? (http://www.aoaforums.com/forum/general-hardware-discussion/49769-bad-g-skill-f2-8500cl5d-4gbpk.html)

MUff1N 28th April, 2011 03:48 PM

Bad G.SKILL-F2-8500CL5D-4GBPK RAM?
 
2 Attachment(s)
Hey guys, when I upgraded my system I replace my 4x1Gb modules of Corsair XMS2 TWIN2X2048-6400C4 Ram with a G.SKILL-F2-8500CL5D-4GBPK set.

Every performance test I ran looked low to me for the speeds & timings they were running, so I did a test & swapped out the G.Skill with my old Corsair Ram with the exact same timings & speed & tested again.

Much to my amazement the Corsair scores soared over the G.Skill BIG-TIME with the exact same settings!!! :eek:
So this makes me think there's something wrong with the set I got.
What's weird though is when I run diagnostics on them they report back fine with no errors of any kind. :confused:

Here's the pics of Everest Cache-Memory Benchmark running the G.Skill & the Corsair.
Tell me what you think about this & get back to me as to why this is...weird that my Corsair DDR2-6400 wipes out the G.Skill DDR2-1066...
Is this normal performance for G.Skill 1066 Ram???
I don't thinks so... :confused:

Anyways, until I can get a replacement set (2x2Gbs instead of 4x1Gbs modules) I'm going to keep running my old Ram for the performance.
Seems stupid to have the rest of the system in high gear & the Ram in 1rst gear! :banghead:

G.SKILL-F2-8500CL5D-4GBPK- Left Picture Benchmark results.

Corsair XMS2 TWIN2X2048-6400C4- Right Picture Benchmark results.

ThunderRd 28th April, 2011 06:08 PM

Muff1N;

You should be able to get that RAM to run faster, but I believe that using 4x1G sticks is a big mistake. Most of these current mainboards are notoriously difficult to run highly overclocked with all slots populated.

When you get the 2x2G sticks, try playing with the dividers. I have the exact same P5Q Pro Turbo board and 1066 Kingston Hyper-X. On the 333 strap, FSB is 475X8=3.8G with memory running at DDR2-1521, 5-5-5-13 (5:8 divider).

The machine doesn't run Windows, so I can't do the same benchmarks you can, but I can assure you it is 100% stable at full load 24/7 on a Q6600 Quad (a 1.2G overclock), which is significantly more challenging to get to than on a Core Duo. So it can be done, but I'm sure you'll be better off with two sticks.

I can only guess that the Corsair stuff has a lower latency than the G.Skill stuff, and that is why it performs better. It isn't DDR2-6400, BTW. It's PC-6400, which translates to DDR2-800 vs. the G.Skill at DDR2-1066, but I've been able to run Corsair XMS DDR2-800 at 1066 easily in the past.

You could be right, though, maybe there is a bad stick in there. Did you try only two sticks of the G.Skill ram just for giggles? It might tell you something.

MUff1N 28th April, 2011 09:17 PM

1 Attachment(s)
Quote:

Originally Posted by ThunderRd (Post 511716)
Muff1N;
Did you try only two sticks of the G.Skill ram just for giggles? It might tell you something.

That is the test result you are looking at...2x2Gb modules of G.Skill PC-1066.
Btw, the G.Skill modules won't OC at all past 1066 no matter how much voltage you give them!

I can also run the Corsair to about 1100MHz but I have to use 2.20v & only use the match set that uses the Micron D9 chips.
The test with the Corsair is with 2Gbs (1x1Gbs-matched pair & D9 chips) at 1000MHz with the same exact timing as the G.Skill.
The other set I got has who know what in them as they don't OC very well at all & can't even do 1000 @ 5-5-5-15 CR2 like the ones in the test comparison.

Sandra reports that the G.Skill is running at only 32% efficiency at 1066!
Whereas the Corsair at 800MHz has just about the same scores at 5-5-5-15 timings with 200MHz less speed, so somethings not right.

I left a message on the G.Skill support site to see if these scores are average for the set I got, but I don't see how. :confused:

Anyways I'm able to run all 4 sticks of the Corsair modules at 980MHz which still out scores the G.Skill 1066MHz... :rolleyes:

ThunderRd 29th April, 2011 05:05 AM

I have found that the key setting for memory on this board is getting the right strap, not more voltage. Certain memory speeds at 400, for instance, would not even boot on my board, while others ran fine. I had to do a lot of experimentation with the straps and speeds before I found the one my memory liked, and when I finally found the right one, things fell into place. And I used only a slight voltage bump to do it.

Now, I can't guarantee that you have the same situation, but before giving up try the 266, 333, and 400 straps at all speeds (this means you will try all the different dividers. On ASUS boards I've been most successful with the 333 strap, but others say differently.) I'm assuming that you know how to calculate the dividers. That's very helpful as well, because some dividers are offered on more than one strap, so you can compare how the machine overclocks using the same divider and different straps. Knowing the divider values takes out a lot of the guesswork, and lets you approach the overclocking in an organized way. Unfortunately, ASUS doesn't provide the actual divider values in the BIOS, so you have to work them out.

Just in case you're not familiar with how to calculate them, you could calculate my divider/speed setting like this:

475FSB x 2(DDR) x 8/5 = 1520 (5FSB:8MEM divider)

With a 3:4 divider it would be like this:

475 x 2 x 4/3 = 1267

MUff1N 30th April, 2011 05:33 PM

Testing Results-Good!
 
1 Attachment(s)
I found something better & easier as well TR...which was combining the Corsair with the G.Skill at the same time. ;)
After I tested & found out that the Corsair (the ones with the D9 chips) would run at 1000MHz @ 5-5-5-15 with 2.10v I just
added in the G.Skill too to see what would happen.

Somehow because of the Corsair's SPD, the mobo reads increased the throughput a ton on the G.Skill just having the Corsair installed! :eek:
So now I got 6 Gigs installed running all the same together with 2.10v which is what they're designed to do with much better
memory throughput! :D

Check out the results with the 6Gbs installed! :thumbsup:

ccperf721p 2nd May, 2011 07:23 AM

CL, tRCD, tRP and tRAS are significant, but don't forget about all of the other timings associated with memory. They are also set by the board based on information from the dimm.


All times are GMT +1. The time now is 03:23 PM.


Copyright ©2001 - 2010, AOA Forums


Search Engine Friendly URLs by vBSEO 3.3.0