Here's a little pretest redline eye candy,........

[fareastgq]

http://pg.photos.yahoo.com/ph/neeb_vwj/detail?.dir=f279&.dnm=b5a9.jpg&.src=ph

unfortunately, I have to go to work, I'll be anxious all damn day at work now........ test results to come. Direct comparison for volts/highest htt, between the 3 sets, should give u a pretty good idea of an average for these sticks :) (which is actually the real purpose of my thread)

[JNav89GT]

droolsssssssssssssssssss :O

[THunDA]

droolsssssssssssssssssss :O


me too.. :slobber:

I feel like I have a ram fetish myself lately.. :p:

[justwOo]

Thats just cause its so easy to have a ram fetish these days :(
bh-utt, ch-utt, bh5, bh6, tccd, tcc5 OH THE MADNESS

[charlie]

:thumbsup:

[[XC] moddolicous]

Can’t wait for results. Put them in one by one, burn them in the same, and find the two that clock the highest in single channel, then put them together and you'll get a really high clock (hopefully).

[fareastgq]

to clarify on how I'll be testing, since I haven't really gave it any thought yet ( think about this stuff at work), hehe, anyway
volts/htt tested single sticks
then they'll be matched by htt/volts results
then highest 2 will be put together for some final tests
I'll be using sandra and 3dmark2001 for final results for each stick in single and dual channel. benching only the highest htt/volts results, Maybe I can draw up a nice chart for everything.

[metro.cl]

nice, are those 2x256mb or 2512mb packs??

[HiJon89]

I don't think Redline comes in 2x256MB, nor any CH-UTT AFAIK.

[TheMeatFrog]

droolsssssssssssssssssss :O
no doubt :slobber:

[fareastgq]

Ok here are the results, this got to be a bit more in depth than I originally planned due to some quirks I cannot explain.
The test Bed :

Sandy 3700 05-17 UPMW BIOS 2-17!!!
DFI Ultra D running the 5 volt line
(quantity of 3) 2x512 Redline 4000 dual channel kits
Gigabyte 6600 PCI Express (stock settings)
PC&C 510 deluxe (20 pin psu with 24 pin adapter)

First off, I need to explain a few things. I am running the stock HS to cool the cpu, the ram and 5volt sink on the board is being cooled by a small walmart twin fan running on LOW. the ram/5volt HS during all the testing never got hotter than lukewarm. For max HTT bootable with NO bootup errors, Every HTT max was rebooted 3 times to make sure it was not a fluke, as well as the next HTT up to see if it did boot 3 times consistently. I pulled the power cord everytime I changed the ram to let the power cycle out because I didn't want to end up with dead sticks like some of u guys, hehe, they all survived :)

ALL ram timings are set to auto with the exception of the primary 2 2 2 6, I can take a pic of the bios settings if u like. chipset volts and LDT volts are at max settings. All other pics with 3dmark scores/sandra benches/Cen.Brain ID/cpu-z is below. NOTHING in the bios was changed while testing all sticks except ram volts. all ram was using the ORANGE slots

Each set of dual channel is labeled 1,2,3, ie: 1A and 1B are a set.... and so on, this is how they came in the package. All sticks ran at rated 2 2 2 6 spds at 3.3 volts/250 HTT NO PROBLEMS. I did not bother to test below 3.3 as I am not interested in that, fyi. I ran 3.3-3.5 volts because this is within mushkin warranty limits and this is what I will run 24/7.

I am running bios 2-17 because this is the ONLY bios in my experience that DOES NOT corrupt my OS install/HD no matter how much I oc the htt/cpu. The rest of them, when I use it them I always get dumb errors, ie: the most annoying one is "this computer has recovered from a serious error" right after the desktop loads. I have tried all the bios's in hopes of getting a better bios and NOT corrupting my OS due to oc'ing, none of the newer ones have worked. I thought 5-10-2 was good until that happened also while I was trying to hit high HTT's. (I am running 274x10 @ 3.5 volts 2 2 2 6 while I am typing this)

If you have any questions/comments, freel free to ask :)

what am I trying to answer here?
1. Are those reviews for redline 4000 cherry picked or not?
2. What are my chances of getting good sticks?
3. what is the average HTT of a stick I will get?
4. is 1t really better than 2t, and show me the benches?
5. How does the ram really affect my gameplay?
6. Are these really better than the cheap UTT $30/40/50 sticks of 512?
7. Are you getting paid for this? (I wish)
8. What are you going to do with the other 4 sticks? (they'll be thrown in the for sale forum)

Any other questions u might think of while reading this might be answered also

SINGLE CHANNEL RESULTS:

Lets start off with Max HTT per stick. Ran at 9x multi, 3.5 volts 2 2 2 6 timings.

1A = 277HTT 2A = 267HTT 3A = 276HTT
1B =269HTT 2B = 275HTT 3B = 269HTT

AVG HTT of a stick = 272.16 HTT :)

As u can see, even the worse stick has a quite nice HTT of 19 over the rated specs, while the best one is 27 over :woot: . One funny thing, each set had a really really good stick, and a low HTT stick.... guess maybe it's just chance, heh. Seems like from the results, u will either get a high HTT stick, or a low one.

DUAL CHANNEL RESULTS:

These are ran at 3.5 volts, 2 2 2 6 timings, x9 multi

1A(277)+3A(276) = 276HTT max ---> Sandra = 6947/6874 (1t)
------------------------------------------->= 6308/6329 (2t)

8% bandwidth efficiency loss from 1t to 2t according to sandra 2004 pro, this was consistent for every HTT ran at 1t vs 2t, 8% was the low, 9% was the high.

Best 3dmark stable I could get was 9x274 = 16872
***remember, this is a plain old stock 6600 :)***

Now, lets say I had only bought 1 gig, what would I have ended up with?? Out of the box :

1A(277)+1B(269) = 273HTT @ 3.5 volts
SANDRA = 6681/6812
Best 3dmark stable I could get was 9x273 = 16744

*** this is one of the unexplainable things, I don't know why it was 273 when the slower stick was only 269, guess they like each other, this is the only time I got 2 sticks to run FASTER than the slowest one***

What did the 2 worse sticks do together??

2A(267)+3B(269) = 263HTT
SANDRA = 6616/6537

And last but not least: Some random benches and suicide screenies I couldn't resist doing.

edit : I had to bump it down to 11x269, I was using 11x270 and it froze up on me, heh. guess 10 mhz makes a bit of diff.

2970mhz @ 11x270 1.55+1.10% volts in bios (which was roughly 1.7 volts) cpu-z read it as 1.64 or close.
Sandra = 7632/7605 (1t)
**note, I could not run 2t, I tried to adjust to 2t with a64 tweaker and kept getting reboots when trying to run the mem bench no matter what I did**
**Another quirk, when I ran sandra the first time I got 7732/7605 (1t), I unexplainably lost this bandwith the next 3 times I ran the bench again, resulting in what i posted above, if someone knows why or has any insight, please share, I tried the disable bank 1 and 2 thingy, result was the same, still did not gain my bandwidth back, I took the first test as an error**Sadly, I couldn't do 3dmarks at that volt because the stock HS is crap. I was at 51c in bios alrdy, who knows what it really was. 3dmark kept freezing up, indicating that heat was not my freind. I'll be updating this thread when my g5/iwaki md20rtz/weapon rad and shroud get put together (and get here) :) My ultimate goal is to run an xtreme 24/7 oc on water :)

Also, I'd like to note that, I did ZERO tweaks to the OS or ram in bios, I still have to do ram tweaking to see how tight I can get these to go, I will update with that also. That's just more time consuming and I don't have enuff patience this weekend, heh.

In conclusion, these sticks simply rock, the slowest HTT spd that resulted is still nothing to sneeze at because its a gig in 1t timings we are talking about here, the fastest was just a nice surprise, deff worth the money either way. Compared to the 8 gigs of twinmos UTT I've tested, avg HTT here is 271, avg HTT for those sticks was 255HTT at 3.5 volts. Hope you enjoyed this as much as I did :) Here are the pics :

(I accidentally deleted one of the pics, so I re ran the 3dmark test, but this time with my card oc'd to 499/650, hence the big jump in 3dmark score.)
http://pg.photos.yahoo.com/ph/neeb_vwj/detail?.dir=/f279&.dnm=7199.jpg
http://pg.photos.yahoo.com/ph/neeb_vwj/detail?.dir=/f279&.dnm=e3bc.jpg&.src=ph
http://pg.photos.yahoo.com/ph/neeb_vwj/detail?.dir=/f279&.dnm=2089.jpg&.src=ph
http://pg.photos.yahoo.com/ph/neeb_vwj/detail?.dir=/f279&.dnm=12d5.jpg&.src=ph
http://pg.photos.yahoo.com/ph/neeb_vwj/detail?.dir=/f279&.dnm=96e9.jpg&.src=ph
http://pg.photos.yahoo.com/ph/neeb_vwj/detail?.dir=/f279&.dnm=7061.jpg&.src=ph
http://pg.photos.yahoo.com/ph/neeb_vwj/detail?.dir=/f279&.dnm=3869.jpg&.src=ph

update : Here's what I have for ram timings so far, I will be tweaking more later and updating this post with 24/7 stable htt.

EDIT : Tweaking is done, THESE ARE MY OPTIMAL TIMINGS, they MAY help u, they may not, to explain, these are timings that A: boot into windows with no errors, B: gave me the least memtest errors while not relaxing tight timings. I did do cas 1.5, I found 2 was better stability and actually less laggy booting up windows. C: felt fastest booting into windows

tcl -2
trcd -2
tras -10
trp -2
trc -9
trfc -17
trrd -2
twr -2
twrt -2
trwt -2
tref -2064 ( I have to note, I picked the 1st 2064 listed)
twcl -auto
bank int Enabled
skew control 255+ (increase)
drive str - 7
data str 4
max async - 8ns
read preamble - 5ns
idle cycle -064
dyn counter -enabled
r/w bypass -16x
bypass max -7x
32bit gran - disabled

consequently, it seems on my system, idle cycle tightening and dynamic counter enabled and Tref made the most gains for bandwidth at a certian HTT, I don't know why, if u do, feel free to explain :)

[TheMeatFrog]

Well done and a very nice read man!

[Major]

No doubt Mushkin Red is the way to go right now with UTT and low latency ! ;)

Good results and thanks for sharing the info with everyone ! :toast:

[boostdfd3s]

Very nice writeup man! :toast: One question though, which slots were used for the tests? Orange or yellow?

-Zach

[fareastgq]

hahaha, sorry, I was gonna note that too, I forgot, I used the ORANGE slots. Updated in thread now..

[TheMeatFrog]

The DUAL CHANNEL RESULTS: are a little hard to read on my screen could you change the color on the by any chance? :D

Edit: and check your PM :)

[mouawad]

good stuff there dood ....no doubt mushkins cash register will be getting even more action now.

agree with you on the twinmoss utt ...my 1a4t sticks were dissapointing to say the least - like the 1 stick which refused to work in dual channel no matter how many pulses i put through it :(

the redline stuff is great and i am happy to pay a premium for some pre-testing knowing i will get some good performance out of them without having to put through 3.6v.

[bateslabel]

I still don't see what the hoopla is about? No way 15 mhz is worth upwards of $100 more. You have got to be kidding right? What was the price of the Twinmos sticks, $60-70 less? I am not going to argue that these are some nice sticks, everyone who has reviewed them has given them good grades, but they aren't in reality going to change the overall speed of a system where it justifys the price. Thats really the bottom line.

[bateslabel]

Well I guess I put my foot in my mouth, they seem to have dropped the price quite a bit so they are reasonable.

[[XC] moddolicous]

very nice. This might be a good thing for overclockers to do. Buy about 3 sets, check each stick, keep the best ones, and sell the others here. Was there any burnin?

[fareastgq]

zero burn in, nothing but sandra benches and 3dmarks, I am burining in the sticks I kept as we speak, I expect 5-10 mhz from them. That is about what the twinmos sticks gained with burn in. I need to beat charlie, going for xtreme 24/7 fsb here, hehe.

[ben805]

running superpi 32m is going to be a challenge....

[fareastgq]

I'll run it for ya when I get my water all setup (be about 2 weeks hopefully), I don't wanna do super pi on a lame 2.75 gig oc and no ram burn in. also, I know nothing about super pi, what version do I get?

Also, I did ram tweaking all day today, updated thread with what was stable enuff to run 1 loop of 3dmark (and hopefully more, hehe).

[ben805]

I'll run it for ya when I get my water all setup (be about 2 weeks hopefully), I don't wanna do super pi on a lame 2.75 gig oc and no ram burn in. also, I know nothing about super pi, what version do I get?

Also, I did ram tweaking all day today, updated thread with what was stable enuff to run 1 loop of 3dmark (and hopefully more, hehe).

cool.

The regular version should be fine, I think you still have lots of headroom left....the timing table is very time consuming to play with but once you get everything dial in these little red devil will scream, I'm going to post some of my result soon.

[[XC] moddolicous]

I'll run it for ya when I get my water all setup (be about 2 weeks hopefully), I don't wanna do super pi on a lame 2.75 gig oc and no ram burn in. also, I know nothing about super pi, what version do I get?

Also, I did ram tweaking all day today, updated thread with what was stable enuff to run 1 loop of 3dmark (and hopefully more, hehe).
Get the millisecond version (in the superpi subforum). As far as I know, there is no way for it to cheat. Also, if u could run SPI32mb @ 275 or 280, that would be insane.

[HousERaT]

running superpi 32m is going to be a challenge....

If those sticks are stable enough to pass 3DMark and perhaps some clean testing in memtest86 then they should be within 1-3Mhz of a full 32mb superpi.

[g0dM@n]

I noticed that you ran it at 2-2-10-2... does that lower bandwidth by much to lower the tRAS from 6 to 10?

[fareastgq]

I don't know, sandra was being dumb, it varied 30 points every time I ran the test 3 times at a certain tras, I was messing around with 2-3-4-5-6-7-8-9-10, I just happen to play with 2 and then 10 to hopefully see if there was actually a difference, far as I can tell, 2 and 10 are almost the same., also, it doesn't like 0, hehe, I'm temped to not use sandra at all. I think sandra cannot bench correctly on my system, I lost like 200 points of bandwidth, then gained it back, then 2 hours later lost it again, I can only explain this by saying that sandra simply doesn't work corrrectly, so I started running 3dmark benches instead, heh. I'll just run the tightest timings possible. my theroy is 10 is prob more stable though. since I can't rely on sandra, I'll have to do some memtest benches too now, so I'll add that on too. I hate it when you can't interpet information or don't get a good picture to hypothesize what's going on.

Anyone know where I can get the 64 bit version? this would prob help me out.

[fareastgq]

I've updated the thread with my optimal timings for my system. try at your own risk, hehe.

I have also come to the conclusion I cannot use my version of sandra, I'll have to use the most updated one, I have studied everyone elses screen shots of sandra benches, they are not all messed up like mine. ie : the info at the top matches the info at the bottom.

[fareastgq]

well I just did some surgery on my cpu, no it didn't help my htt, but I can super pi all damn day now at 2.7 gigs with only 1.4 in bios now, also 10c cooler :) Can't wait for water...........

On a side note, I'll be using bios 510-2 now and see if I gain anything, I'll post the results in the main post also just so ppl can see a good comparison. I did gain like 5 htt going from 2-17 to 510-2 using my gskill LE. I'm wondering if it will help my redline also, hopefully that dumb "windows has recovered from a serious error" boot up error doesn't creep up on me, I found a fix for it anyway I think. Also, looks like ram burn in is helping me slowly but surely. I have gained
1 fsb only max htt, but the good news is, my stable HTT has gone up 2 HTT. All I have been doing is running memtest # 8 when I go to sleep and go to work @ 3.5 volts in bios fyi. I was a bit afraid there was no more headroom in these sticks, looks like there is, if I gain a few more HTT, I'll be satisfied as I am not interested in suicide HTT shots much.

[=[PULSAR]=]

Is it me or does Mushkin always seem to sell 1 slow stick and 1 fast stick in a pair :confused: Looks that way from your results fareastgq :(

[ben805]

=']Is it me or does Mushkin always seem to sell 1 slow stick and 1 fast stick in a pair :confused: Looks that way from your results fareastgq :(


Not only mushkin, but Gskill, OCZ, and many other also packed the pair like that....I suppose they're trying to save some money and even out the overall performance instead of down-rated the slower sticks from say pc3500 to pc3200. The gap and differences may vary but most of them are between 1~5mhz. :D :D

[fareastgq]

yeah, it sucks, it's like they are doing it on purpose so ppl buy more sticks. We are victims! Ben, I JUST ran a superpi 16m, my time was crap though I think 13m 54.87 for a 16m superpi at 2.7 gigs, 31.531 for the 1m superpi sucks I think, of course, I don't know, first time i have ran it, heh. good news is, it's only 1.4 volts in bios (1.36 in cpu-z). Looks like my cpu really did benefit from a face lift. I ran it multiple times just to make sure I wasn't kidding myself, heh. Unforutnately my max 0c didn't go up, I hope I didn't hit a wall at 2970 at 1.7 volts....

[ettis]

Can you run 4 sticks together and compare overclockability and performance with only 2 sticks?

[ben805]

yeah, it sucks, it's like they are doing it on purpose so ppl buy more sticks. We are victims! Ben, I JUST ran a superpi 16m, my time was crap though I think 13m 54.87 for a 16m superpi at 2.7 gigs, 31.531 for the 1m superpi sucks I think, of course, I don't know, first time i have ran it, heh. good news is, it's only 1.4 volts in bios (1.36 in cpu-z). Looks like my cpu really did benefit from a face lift. I ran it multiple times just to make sure I wasn't kidding myself, heh. Unforutnately my max 0c didn't go up, I hope I didn't hit a wall at 2970 at 1.7 volts....

superpi depend heavily on memory timing, you'll need to tighten up some of those settings for better score.

[EMC2]

Not only mushkin, but Gskill, OCZ, and many other also packed the pair like that....I suppose they're trying to save some money and even out the overall performance instead of down-rated the slower sticks from say pc3500 to pc3200. The gap and differences may vary but most of them are between 1~5mhz. :D :D

Only 1->5Mhz?! Considering they run at over their rated speed still, there would hardly be any need to down-bin them. ;)

yeah, it sucks, it's like they are doing it on purpose so ppl buy more sticks. We are victims!
:rofl: Hardly victims... do you know how hard it would be to match sticks at OC speeds within 1Mhz? The cost would likely have to double.... or more. No way I would wish to pay the $ for such matching. Let's be realistic... if they both do rated speeds we got what we paid for, anything above that is gravy ;) Would be like crying "victim" or "plot" because some CPUs clock higher than others...

Peace :toast:

[ben805]

Only 1->5Mhz?! Considering they run at over their rated speed still, there would hardly be any need to down-bin them. ;)



agreed, and I found only 1mhz difference with my Redline..... :D :D :D

[fareastgq]

I was kidding when I said we were victims, heh, also, when I flshed to bios 5-10-2 my sticks appear to be dead.... hrm. hopefully, I can revive them, nothing appears to have burnt up.

[fareastgq]

As I suspsected, the ram was ok, flashed back to bios 2-17, (with a stick of plain pc2100), powered off comp, changed jumper back to 3.3v, cleared cmos, put 1 stick of redline in, booted up fine, powered down, took it out, put the other one in, booted fine, put them both in, they are fine. Fired it right back up again, hehe. I'll have to try bios 5-10-2 again. Hopefully, this doesn't happen again. Also guys, I have seen dead sticks of UTT before (my own), you may want to check the gold fingers on the ram for damage, I could actually see burnt fingers, I suspsect THIS is what actually happens when a stick dies, if you do not see this on your sticks, don't be so quick to rma them yet. My sticks of redline, MB beeped like both sticks were dead. I'm sure many ppl have experienced the same thing, but it's hard for me to believe that the stick actually died after inspecting the gold fingers. ( would have inspected the traces too if I could have seen them) because in my experience, it's hard as hell to kill ram unless u put 3.6+ volts into it. Not saying it isn't possible, just saying make sure u check every other possibility first. :) Back to your super pi dream ben :)
Ettis, I would run it if I could, I only have 3 sticks of redline now, the rest are someone else's now, 2t performance with 4 sticks would have been nice to see, I would have tested it for u too, check out threads of ppl that have ran 4 dimms :)

edit : I had this problem happen with bios 2-17 too, only diff is, with 2-17 if I raise the chipset volts and ldt volts, I CAN boot with either stick, on bios 5-10-2, I cannot boot with 1 stick, can boot with the other stick, then sometimes cannot boot with the stick that I just booted with, what does this imply? this is all conjecture, but the bios is flakey imo, it's not doing somthing right in reguards to voltage, I was able to reproduce this 3 times. I don't claim to know what's happening, I just claim to know what I have experienced. bios 2-17 is still the best bios imo. My other suspicsion is the voltage on the board, prob a combination of these 2 things, 2-17 seems to know how to handle the volts better though reguardless of what is in the bios.

[JNav89GT]

anyone want to see A64 X2 and 4x512mb Redline performance?
Ask me nicely and I might comply ehhe!
Actually, I'll do it anyway b/c I'm curious.
I think I drank some bad OJ this AM :O.

[uscfan]

Is the redline BH-5 or CH-5?

[fareastgq]

anyone want to see A64 X2 and 4x512mb Redline performance?
Ask me nicely and I might comply ehhe!
Actually, I'll do it anyway b/c I'm curious.
I think I drank some bad OJ this AM :O.

yep post it, hehe, I'm almost tempted to sell this chip and get some dual core 2 MB action going, just for the hell of it, hehe. I believe it's ch-5, mushkin's site should be able to tell u.

[JNav89GT]

I don't think Mushkin is going to name IC b/c then if they have to switch everybody will get their panties in a bundle

X2 lovin is the cure though baby! For whatever itch you need to scratch that is ;)

[fareastgq]

running superpi 32m is going to be a challenge....

Sorry Ben, u can bug Vapor now, Muahahahhahaa!!!!!

[dnottis]

Is the redline BH-5 or CH-5?

Redline CH - confirmed with Mushkin