G.SKILL Extreme Series 2GB (2 x 1GB) OC Report

[Thoth]

Still running these... running them @ 275mhz @ stock timings/voltage... had problems tweaking the ram @ 275mhz.. but i had them @ some nice timings @ 265... (10x265)

when working on 275 tonight... i was working through the settings and got through most of them and then when i got to idle cycle timer everything went haywire... even when i rebooted the comp and put the setting back (to auto) the comp refused to boot... it was passing memtest... but bsod'ing as soon as windows loaded... o well...

 

[bboyfobulous]

i finally got these babies stable !! woo hoo!!

2.65v 3-4-4-8 1T
cpu @ 286 x 10
ram @ 260
divider at 9:10

not exactly what i was goin for, but it's stable

 

[darksparkz]

i finally got these babies stable !! woo hoo!!

2.65v 3-4-4-8 1T
cpu @ 286 x 10
ram @ 260
divider at 9:10

not exactly what i was goin for, but it's stable

What did you set to get your divider at 9:10 on the A8N32? What value do you lower your memclock to, I tried setting it to 133 or 166 and just raising FSB, and it doesn't get high enough.

 

[bboyfobulous]

use above settings and use a 183 memclock limit

9 / 10 = .9
260 / 286 = .9

[edit] and i see you're runnin 0903 bios....i'm usin 1205 [/edit]

 

[CGR]

I have gotten mine stable at 275 with stock voltage. Have not been able to get them any higher. Even 1 mhz more and I get massive errors in memtest #8.

 

[Thoth]

i ALMOST had them stable @ 280 @ stock timings... lowered the "read preamble" and "max async latency" and the ram ran through 4-5 test 5's... but started generating errors on pass 6... (before someone asks... YES i do have a fan blowing over the ram... they are only *warm* to the touch)

would persue 280mhz more... but i have a lan party to go to tomorrow @ 9am... so tweaking will have to wait...

 

[darksparkz]

Mine isn't even Prime stable at 250, for a 1:1 ratio. How come many people get to 270 or higher? Maybe it's since I'm using an Opteron 175, the 11x multipler doesn't allow the ram to be clocked too high? Any ideas? I've tried loosening the timings, lowering and raising ddr voltages and everything. My previous OCZ Performance PC3200 could clock a bit higher then these G.Skills, while I thought the G.Skills were much better.

 

[dominick32]

Mine isn't even Prime stable at 250, for a 1:1 ratio. How come many people get to 270 or higher? Maybe it's since I'm using an Opteron 175, the 11x multipler doesn't allow the ram to be clocked too high? Any ideas? I've tried loosening the timings, lowering and raising ddr voltages and everything. My previous OCZ Performance PC3200 could clock a bit higher then these G.Skills, while I thought the G.Skills were much better.

Thats very strange. Something in your configuration must be wrong. Can you post up a pic of A64Tweaker and CPUZ for me. If you are sure all configuration settings are correct, try changing your voltage to 2.6 volts because I see you have 2.7volts listed in your sig. If all else fails I would definitely consider RMA'ing these sticks because failing Prime @ 250 would mean the sticks are not considered quality and may have passed right through Gskills QC department.

Good luck.

Dom

 

[kbtat2]

When I first recieved my set from newegg they weren't stable even at 250 and I rma'ed them and the second set is great. Have you tried memtesting them? I tested my first set and they spit out errors like mad.
Maybe you could just get another set?

 

[darksparkz]

Here's my screenshot, CPU-Z says 1.250v for vcore, but my motherboard is set to do a 0.2v overvolt, so the actual vcore is 1.450v. vdimm is set to 2.6v now as well in BIOS.

I never did an RMA before, I ordered these from newegg. What would my RMA reason be? Defective/Failure? Since these were PC4000, and they weren't even stable at 250mhz. Do I just put the sticks back into the box and ship it back with their RMA number on it? They would ship me a new set for an exchange right?

 

[dominick32]

Here's my screenshot, CPU-Z says 1.250v for vcore, but my motherboard is set to do a 0.2v overvolt, so the actual vcore is 1.450v. vdimm is set to 2.6v now as well in BIOS.

I never did an RMA before, I ordered these from newegg. What would my RMA reason be? Defective/Failure? Since these were PC4000, and they weren't even stable at 250mhz. Do I just put the sticks back into the box and ship it back with their RMA number on it? They would ship me a new set for an exchange right?

Yes, definitely call up Newegg customer service and tell them that you ordered memory from them that is defective. They will issue you an RMA number over the phone and e-mail you with details on how to return. Newegg happens to have a very fast RMA program as well. Once they receive your sticks, there is usually a 3 to 5 day turn around.

Any sticks that do not run at the rated frequency/timings: 250 mhz (3-4-4-8) are to be considered defective/failing.

Good luck man!

 

[darksparkz]

Thanks a lot for the information! I'm hoping the new sticks would let me hit 260x11 with 1:1.

Also: RMA is 30 days from invoice right?

 

[Fangs404]

I have to post here. This ram is totally ridiculous. I'm running 250x10 1:1 right now 3-4-4-10 1T in my ASRock mobo at 2.7V. One of the reasons I picked this ram was because of how little voltage it requires. I made the right decision. I stuck the ram in, tuned to 250, memtested, and bam, I'm home free. I'd recommend this stuff to anyone.

A+++++++++++++++++++ WOULD BUY AGAIN

 

[Thoth]

I have some new findings... as i was able to successfully reproduce the conditions in my last full scr shot (ram @ 296mhz) ... and post a a64 tweaker shot of the settings... (which btw except for the "main 5" were auto set)

Cpu @ 9.5 x 250 , ram @ ddr466 (233mhz) dividor

a thing or two i noticed about this that don't make a whole lotta sense...
1. this statement (from http://www.dfi-street.com/forum/showpost.php?p=203179&postcount=4 "Usually, this should be CAS latency + tRCD + 2 clock cycles. For example, if you set the CAS latency to 2 clock cycles and the tRCD to 3 clock cycles, the optimum tRAS value would be 7 clock cycles." ....so that would mean tcl (3) + trcd (4) + Delay (2) = 9 ... so how can this ram run BETTER at 8..
2. look at the "Max Async Latency" and "Read Preamble" .. stock settings for those are 7ns for "Max async Latency" and 5.5ns for "Read Preamble" everywhere i've read for those two settings... lower = faster... seeing that the motherboard automatically raised the "read preamble" didn't suprise me, but seeing that it LOWERED "Max Async Latency" by 1 did...

anyone else think this is just a tad weird...
also... correct me if i'm wrong... but even despite being a few hundred mhz slower than normal (cpu speed wise)... shouldn't 296mhz on the ram give me more bandwidth in sisoft (everest read lower #'s also but didnt put it in scr shot)

Comments/questions/suggestions welcome.

 

[Mustanley]

A few things...

First of all, on the DFI, dividers are computed using only whole multipliers. So when you are running a half multiplier (9.5 in your case) the board rounds up to the next whole number multi (10 in your case). Thus the actual frequency is not what you'd expect it to be, or even what is reported by cpu-z.

You actual memory frequency is more like 277MHz. 233/200=1.165 ratio.
So 250x9.5=2375MHz

2375(cpu freq)/10 (rounded up multi) = 237.5MHz x 1.165 (your ram ratio) = 277MHz

http://www.dfi-street.com/forum/showpost.php?p=203167&postcount=3

Next thing is that memory bandwidth is affected more by overall cpu speed than by memory speed on the A64 due to it's integrated memory controller. While increasing the memory bus speed does indeed help increase bandwidth, it still will not have as big of an impact as further overclocking the cpu.

For instance, leaving the FSB locked at 250MHz and changing the multi from 9 to 10 will have a huge impact on bandwidth.

 

[Fangs404]

anyone else think this is just a tad weird...
also... correct me if i'm wrong... but even despite being a few hundred mhz slower than normal (cpu speed wise)... shouldn't 296mhz on the ram give me more bandwidth in sisoft (everest read lower #'s also but didnt put it in scr shot)

Comments/questions/suggestions welcome.

I have a feeling it's because of the divider. The ram can finish with data faster than the CPU can get the data to the ram. I'm running 250x10 1:1, and my bandwidth is near 7000 (I need to rerun Sandra.....I forgot it exactly). My superpi is right around 34 or 35 also.

 

[Rickster]

Im getting errors on Memtest v1.65 on test 8, theres massive errors on it like 80 errors. Im running it 264Mhz and timings and voltages on stock. The RAM is quite hot to the touch. Can anyone give any suggestions?

Thanks

 

[sofarfrome]

first off I would try some active cooling to see if the errors decrease...after that perhaps you can bump vdimm a little but typically these sticks do not like more V. what mobo do you have?

 

[squads]

Im getting errors on Memtest v1.65 on test 8, theres massive errors on it like 80 errors. Im running it 264Mhz and timings and voltages on stock. The RAM is quite hot to the touch. Can anyone give any suggestions?

Thanks

80 errors is nothing. I was testing my RAM at 290MHz yesterday and I got over 1,000,000 errors on test 6 in the first run Turns out I was only memtest stable up to 265, which works out fine for me.

 

[Rickster]

What batch of RAM do you have? I have 0606. OK maybe 80 errors isn't as massive as compared to yours

I have an Abit KN8, I like the board. (Don't tell me to change it to DFI)

When I get back I'll let you know about the Memtest retest on test 8 again with and without cooling and on stock frequencies and voltages.