need help w/mve & 2133 cl8 hypers

[funsoul]

Hi again folks

Finally (slowly) getting the mve/2600k setup dialed in for some 3d benching. Hoping for a little boost in scores, I picked up a pair of hyperx 2133 cl8 sticks (KHX2133C8D3T1K2/4GX).

Gave them a shot last night but was unable to get the sticks running stock (8-8-8-24 1T @ 2133 w/1.65v) or 7-7-7-21 1T @ 2203 w/1.8v like this guy: http://www.xtremesystems.org/forums/showthread.php?251835-Kingston-HyperX-CL8-8-8-24-(DDR3-2133

Did get 8-8-8-24 1T, 2133 but at 1.7v. Also got 8-8-8-24, 2203 @ 1.8v working (not hugely stable, though). Neither 7-8-8-24 nor 8-7-8-24 were possible. At 2203, it did help improve my 2600k pifast sub from 13.03 to 12.95 (and that was at a slightly slower cpu speed) but am sure more's possible.

My current bios settings are here: http://www.overclockers.com/forums/showpost.php?p=7620697&postcount=44

Any ideas?

Thanks once again (jeeezz....I'm asking so many questions lately!)

 

[Woomack]

Hypers are running much better on older platforms and guy in the link was testing it on X58 board.
On the new boards you have to set different sub timings and usually higher voltages to make it run stable.
On some boards hypers are just not clocking good. I have HyperX 2000 9-9-9 kit based on hypers and it's not running above 2200 on any new board up to 1.85V.
Usually sub timings for this memory are more relaxed than new IC that's why without tweaking, results in benchmarks are much worse than new RAM.

Also you can check DRAM vrm frequency and VCCIO/SA switching frequency settings. Sometimes it's helping ... and it's already hitting nearly max memory clock for this platform.
Check if there are hypers profiles in BIOS , it will be easier to compare what to change. Also check higher VCCSA/IO ( 1.35/1.30V maybe )

 

[Johan45]

You have to remember that the guy in the link is using a different system and they all react differently. You can try raisng the RTC and refresh time a bit , I'll start with 2T command as well to see where I can get and then try 1T later. One ting that woomack suggetsed to me when I was going for more speed wast to use 9-9-9 at 1.65v and got me to 2400 9-10-9-24 with better performance than 7-9-7 at 2200.

 

[funsoul]

Hmmmm...thanks guys. Will give it a try in the next couple days and report back. Trying to run in the 1950-2020MHz range.

Also have other sticks to play with so welcome your thoughts if you think one of them would be better than the hypers on this setup:
- Geil Evo 2 2000- 6-9-6 @ 1.65v
- Corsair Dominator GT 2133 9-11-10-27 2T @ 1.5v (been running these 8-11-10-27 1T at 2203 w/1.7v)
- GSkill Tridentx 2400 9-11-11-31 2T @ 1.65v (tried underclocking these down to 2203-ish but was not able to tighten timings even at 1.75v)
- GSkill Tridentx 2666 11-13-13-35 2T @ 1.65v (these seem irrelevant for the current setup?)

Guess another, alternate question would be....what would be the best possible ram for my setup (and speeds) I'm running?

 

[Woomack]

Geil will be probably best. Something like 6-10-6 2250 is possible. G.Skill 2400 9-11-11 can probably make 2200+ 8-10-10 and tight subs. Other 2 are hynix so will be slightly slower at these clocks.

 

[funsoul]

Thanks Woomack Excluding than the ones I've already got, what do you think would be the best possible sticks for this setup/speed?

 

[Woomack]

waste of money to get anything else for this setup

 

[funsoul]

Have the system back up and running. Am using a 2700k and the evo2 sticks. Been working on the cpu overclock....will likely end up ~5525GHz. Once I get that stable it's onward to the memory....

The evo2's are running 6-9-6-24-2T @ ~1000MHz (bclk overclocking). The system refuses to post with the sticks set at anything other than the 1866 strap. Would like to be able to run them from 2133 to around 2200MHz (dual channel) and the tightest timings possible.

My (more or less) current settings are in the first post.

Ideas? Recommended voltages and/or other settings to try?

tia as always...really appreciate the help!

 

[funsoul]

Hi again :/

OK...been trying out the evo2's and hypers. The timings on the evo2's are tighter but the hypers overclock better.

Here's where I've gotten so far...

In the high multi/low bclk scenario (54x102.4,5529MHz):
- hypers- 8-8-8-24 1T @2184, 1.75v
- evo2's- 6-9-6-24 1T @1910, 1.65v, 5-8-6-24 1T with 1.75v, could not post at 2184 with 1.75v @ 6-9-6-24 2T

In the low multi/high bclk scenario (53x104.3,5528MHz):
- hypers- 8-8-8-24 1T @1946, 1.65v, could not post at 2224 with 1.75v @ 8-8-8-24 2T
- evo2's- 6-9-6-24 1T @1946, 1.65v, could not post at 2224 with 1.75v @ 6-9-6-24 2T

Testing using wprime, oddly results were all more or less the same using any of the above configurations/settings. Also, in the high multi scenario, going from 6-9-6 to 5-8-6 didn't seem to make any difference in my wprime results.

So...what do ya'll think? Would like your thoughts on other timings/voltages to try. Am trying to figure out the best sticks to use for the high/low multi scenarios as well as the best possible speeds/timings for each case.

Thanks!

 

[Woomack]

Hypers look much better here. You can try 7-7-7, 7-8-7, CL6-7-6, 6-7-7 on Hypers and how high they will go up to 1.85V. Also check higher tRFC.
I remember that some IMC didn't like CL6 above ~2000 so maybe something like 7-9-7 on PSC is going to work @2200+.

Wprime is almost not using memory. You can even use 1 stick and it won't make any special difference.

I doubt that any tighter timings will change much in anything else than Spi32m but you can try.

 

[funsoul]

1.85v?? Just have a 140mm fan blowing over the ram...is that sufficient? Would really like to avoid killing any sticks.

Hoping the get the ram dialed in tonight so tomorrow night can actually try a little benchiebenchie action. Think I'll test using spi this time :/

 

[Woomack]

I'm not making you to use higher voltages but if you wan't to improve something then for these older kits you simply need higher voltage. Just don't blame me if you kill memory
But really all my kits were running at least up to 1.85V on air without bigger issues. One BBSE kit was overheating above ~1.85V so I was keeping it at 1.8V for tests.
PSC are generally scalling on air till ~1.85V. Some kits don't like much above 1.8V but some can run above 1.9V. At least my Dominators GT don't like more than 1.82V while G.Skills were running up to 1.92V on MVG.
Earlier Hypers could randomly die at 1.8V+ when temps were high but most newer batches could run above 1.8V without issues.
My HyperX 2000 8-8-8 were running up to 1.85V on air without issues but lately I'm using them only on H55 board as they easily run @1800 7-7-7 1.65V.

 

[funsoul]

Awesome...thanks again Woomack. Will check 'em out and get back.

 

[freeagent]

I'm curious how they do for cas 6 and 7 operation.. My chromes are rated at 2000 8-8-8

 

[funsoul]

Testing last night was a struggle :/ Wasn't able to improve speeds or timings. Did get the hypers to boot at 2224 w/1.85v @9-9-9-24 1T but the system would lock up if I tried to actually start any apps. :/

Now, 10 hours later, it finally occurred to me that I should try it in manual instead of XMP mode :/ (yes...I can be a little slow sometimes).

Will try again tonight.

 

[Johan45]

Yes memory can be frustrating at times. Oh man what was I talking about??

 

[funsoul]

While waiting for the new haswell build, re-loaded the mve/2700k and figured I'd make another go of getting my ram to run as tight as possible at 2184MHz and 2224MHz (dual channel mode).

Did some searching around for reviews and such. I figure I'll try some of the timings/voltages found.

Here's what I found on-line:
1) Kingston HyperX 2133 (khx2133c8d3t1k2/4gx), stock 1.65v 2133 8-8-8-24
Results found on-line:
2184 7-7-7-21 1T @1.85v
2208 8-8-8-27 1T @ ?
2256 8-8-8-27 1T @ ?

My current results:
2184 8-8-8-24 1T @ 1.75v
2224 9-9-9-24 1T @ 1.85v- very unstable with frequent lock-ups

2) Geil Evo 2 (get34gb2000c6dc), stock 1.65v 2000 6-9-6-24
Results found on-line:
2200 6-10-7-26 1T @ 1.86v
2224 7-10-7-? @ ?
2256 6-11-7-24 1T @ ?

My current results:
-none good-

At 2184MHz or 2224, it looks like I should focus on the hypers (as Woomack already pointed out) and see if I can get to the 2184@ 7-7-7-21 using 1.85v and 8-8-8-27 @ 2224. Trouble is that I've not been able to nearly hit those timings at 2224. Is it correct to assume that the Evo2's at 7-10-7-24 @ 2224 would be more or less the same speed as the hyperx at 9-9-9-27?

Argghh...SO frustrating!!!

 

[Woomack]

Hypers @9-9-9 will be about the same as PSC @7-10-7 and I think 2200+ clocks can be limited by memory controller so depends from main and sub timings you may see drop in stability even though memory should run even higher.
I don't think that difference in benchmarks is worth all this time spent on tweaking memory on SB. Maybe try some in sub timings to check if it helps to gain some performance but I doubt it will be much better.

 

[freeagent]

You should quit messing with those old hypers, and just sell them to me.

 

[Woomack]

I think it would be better for all