• Some users have recently had their accounts hijacked. It seems that the now defunct EVGA forums might have compromised your password there and seems many are using the same PW here. We would suggest you UPDATE YOUR PASSWORD and TURN ON 2FA for your account here to further secure it. None of the compromised accounts had 2FA turned on.
    Once you have enabled 2FA, your account will be updated soon to show a badge, letting other members know that you use 2FA to protect your account. This should be beneficial for everyone that uses FSFT.

SR-2 Optimization Thread

I got a P2686 (R0, C0, G31) on my SR-2 (L5640s @ 3.6GHz), and with NUMA disabled I have a new high: 12:46 TPF, 135,086 PPD

It's interesting, but I'm pretty sure that P2692s were faster than P2686s when I had the X5650s installed. With the L5640s, it's the reverse.

I think I'm going to buy another SR-2, but I'm going to wait a few weeks to make sure I still have a job. :eek: I'm going to set it up as a SQL Server box 2008 with 24GB of RAM (6 x 4GB).
 
Since I was still having some issues running -bigsdv units and I'm visiting the land of the big Microcenter this weekend, I picked up 12 gigs of brand new DDR3 for my SR-2 motherboard. I was using Corsair 9-9-9-24 without much luck on the SR-2, even though it would run -bigadv units all day long on all of my Gigabyte P55 and X58 boards.

Anyway, since this OCZ DDR3 8-8-8-24 is on the recommended compatibility list for the SR-2, I'm hoping to have better luck with it this time around. Anyone else using this RAM and can confirm successful -bigadv runs?

On a side note, I also picked up a i7-950 and a Swiftech H20-220 Ultima XT setup for one of my X58A-UD3R setups. :D This will be my first full water install instead of using H50's so I'm kind of pumped about it. :cool:
 
Axdrenalin said:
Anyway, since this OCZ DDR3 8-8-8-24 is on the recommended compatibility list for the SR-2, I'm hoping to have better luck with it this time around. Anyone else using this RAM and can confirm successful -bigadv runs?
Click to expand...

I am running this memory in one of mine at 8-8-8-20 if I am not mistaken. It is at 198 bclk/1564 MHz. It will go higher, but the machine is actually slower over 198. I actually had it up to 208 blck/1664 MHz when I had Gainstown Xeons in the board.
 
NOOOO!!! OCZ OCZ3G1600LV6GK PC3-12800 Gold 3x2GB?

Cost me weeks of hell across 4 kits. On the list, but impossible to get stable at 2:8 at stock/ any speed.

Stay away if your life depended on it.
 
Vaulter: Good stuff, not much more money than the 3x1GB kits I was looking at.
 
musky said:
I am running this memory in one of mine at 8-8-8-20 if I am not mistaken. It is at 198 bclk/1564 MHz. It will go higher, but the machine is actually slower over 198. I actually had it up to 208 blck/1664 MHz when I had Gainstown Xeons in the board.
Click to expand...

may i be worthy said:
NOOOO!!! OCZ OCZ3G1600LV6GK PC3-12800 Gold 3x2GB?

Cost me weeks of hell across 4 kits. On the list, but impossible to get stable at 2:8 at stock/ any speed.

Stay away if your life depended on it.
Click to expand...

Wow....Not what I was hoping to hear. Musky, any way you can confirm your chips to verify? I read through the reviews on Amazon about the OCZ RAM, and they were a bit less than I was hoping for. I really need to make a quick decision, cause I'm headed back tomorrow and wanted some *good* recommended RAM so I can get these -bigadv units running on my SR-2. :rolleyes: My Corsair passed all the MemTest and LinX tests, but just doesn't seem to be doing it without causing errors. Scratch that - it processes *everything* perfectly except -bigadv units.

<sigh> Frustrating...I guess I can take this back to the store tomorrow before I head back. Just hate to leave without a decent set that I know will work.
 
Axdrenalin said:
Wow....Not what I was hoping to hear. Musky, any way you can confirm your chips to verify? I read through the reviews on Amazon about the OCZ RAM, and they were a bit less than I was hoping for. I really need to make a quick decision, cause I'm headed back tomorrow and wanted some *good* recommended RAM so I can get these -bigadv units running on my SR-2. :rolleyes: My Corsair passed all the MemTest and LinX tests, but just doesn't seem to be doing it without causing errors. Scratch that - it processes *everything* perfectly except -bigadv units.

<sigh> Frustrating...I guess I can take this back to the store tomorrow before I head back. Just hate to leave without a decent set that I know will work.
Click to expand...

I know it sucks, but better waste a day than weeks of mucking around with phantom errors. The OCZ was a nightmare because it passes memtest86+, but fails HCI-Memtest when running a dozen threads. So it might be on the recommended list simply because people don't test their mem hard enough.

I can recommend my Kingston KHX1600C7D3K3/6GX - having bought 6 kits across 2 machines. Not perhaps for 24GB, but it does run flawlessly at 1850DDR in 12GB.

SazanEyes said:
I have checked with a DMM, and the voltage is about .03 higher than set in BIOS which I why I run a bit lower. A few hundredths either way probably aren't a big deal, but I wouldn't go much higher than 1.65V (or the max specified by your RAM, whichever is lower).
Click to expand...

Thanks for that - I was thinking of buying a multimeter at my local electronics shop, but I don't know enough about it, and assumed a cheapy one might be more inaccurate than it is worth. But it sounds like the eleet reported voltages are close to the mark.

Tobit said:
MIBW: Did you ever get a decent overclock working when running 12 sticks of memory? Looking for ways to get to 24GB without spending a premium on 6x4GB DDR3-1600 sticks or going the cheap hex route and buying 6x4GB DDR3-1333.
Click to expand...

Yes I have - my x5650 system does ok at 4.2Ghz and 2:8 @ 1512 ddr.... but it cannot go to 2:10 like I have on my x5660 sytem This is partly because the multiplier is 1 step harder - needing 1910DDR vs 1850DDR - and also because of the 12 dimms. So I would probably look at getting a 6 x 4GB DIMM solution next time.

SazanEyes said:
I got a P2686 (R0, C0, G31) on my SR-2 (L5640s @ 3.6GHz), and with NUMA disabled I have a new high: 12:46 TPF, 135,086 PPD. It's interesting, but I'm pretty sure that P2692s were faster than P2686s when I had the X5650s installed. With the L5640s, it's the reverse.
Click to expand...

Interesting. I just got my first ever P2692, and it is a smidgen faster than the P2686 benchmark I have been using. (11:13 compared to 11:16 with NUMA disabled) And yes, you bet I captured the mofo, and ran a few frames on my other SR2.:cool:

NO NUMA NUGGET NUMBER #1 ;)

This leads into another thing I have noticed. I haven't had a chance to bench properly since the disable NUMA trick came out, but already I have the gut feeling that it evens the field for memory performance - making it less critical on bigadv. I am not seeing a large difference between my SR2s than my previous benchmarks would suggest. (minor clock diff, major memory difference) With NUMA on, there was a 3% gain (normalised for clock speed) to the system with the faster memory. Now it looks to be closer to 0.8% to 1.2%.

This could make a lot of sense if data is not being shunted around as much, it would not be stressing the memory as much = performance depends more on clock. Oh good, I just went through a month of spreadsheets, choosing x5660 over x5650, all to maximise memory performance.

Oh well, who cares, if you told me a month ago I would be benching 163,942 PPD on a P2692 I would not have thought it possible.:D
 
Last edited:
may i be worthy said:
I can recommend my Kingston KHX1600C7D3K3/6GX - having bought 6 kits across 2 machines. Not perhaps for 24GB, but it does run flawlessly at 1850DDR in 12GB.
Click to expand...

Thanks for this...I did see at least one kit of Kingston 6Gb there, so hopefully the numbers will be the same as your reference there. I'd rather come home with a good working set than something that is questionable at best.

Would anyone else please post the make and model numbers of DDR3 being used successfully with there SR-2's processing -bigadv units? I'll check back in the morning and hope that one of these two Microcenter stores nearby has something in stock.

Thanks in advance!
 
Corsair xms3 1600 8-8-8 here. Bought both 6gb kits at microcenter actually.
Posted via [H] Mobile Device
 
 
Axdrenalin said:
Wow....Not what I was hoping to hear. Musky, any way you can confirm your chips to verify? I read through the reviews on Amazon about the OCZ RAM, and they were a bit less than I was hoping for. I really need to make a quick decision, cause I'm headed back tomorrow and wanted some *good* recommended RAM so I can get these -bigadv units running on my SR-2. :rolleyes: My Corsair passed all the MemTest and LinX tests, but just doesn't seem to be doing it without causing errors. Scratch that - it processes *everything* perfectly except -bigadv units.

<sigh> Frustrating...I guess I can take this back to the store tomorrow before I head back. Just hate to leave without a decent set that I know will work.
Click to expand...

Mine show up in HWiNFO32 as OCZ3G1600LV2G. They are actually pieced together from 2 x 6Gb kits and 1 x 4 Gb kit. I had 2 bad modules from the original 2 x 6Gb kits (I believe they were both from the same kit), so I grabbed a 4Gb kit and everything works fine. They are at 8-8-8-20-2T.

As far as a recommendation...*shrug*...I have now had a bad 6Gb OCZ Gold kit and a bad Corsair XMS kit. Memory just seems to be a crap shoot. I can confirm compatibility of the OCZ Gold and the Kingston HyperX PC1600 8-8-8-20.
 
Thanks for the info guys - I appreciate the insight on whats working for you. I'll take these back this morning and try to get me some 8-8-8-24 Corsair if I can find it. I've always had a preference for Corsair, but since my 9-9-9-24 sticks weren't too well I thought I'd try a different brand. Guess we'll see once I get them installed.
 
Granted I've never bought OCZ ram before, I do like their tech support staff and how visible they are in forums...however, I've seen so many complaints and problems with OCZ ram now and in the past, I'll never buy their ram. If it boils down to money, I'd buy G.Skill or A-Data to save some cash otherwise I'd go Corsair. None of those ram brands has given me problems in the past (dating back....10 years now?)
 
zero2dash said:
Granted I've never bought OCZ ram before, I do like their tech support staff and how visible they are in forums...however, I've seen so many complaints and problems with OCZ ram now and in the past, I'll never buy their ram. If it boils down to money, I'd buy G.Skill or A-Data to save some cash otherwise I'd go Corsair. None of those ram brands has given me problems in the past (dating back....10 years now?)
Click to expand...
yeah, my gut tells me there probably is a higher percentage of OCZ issues, but that also might be due to a higher volume of sticks being in the market due to genrally lower prices. I've personally had good luck, having owned at least half a dozen kits.

I don't know about everyone else, but I rarely hear about Corsair incompatibility.
 
I usually buy Corsair, but I do have a set of OCZ DDR3 and I think two sets of OCZ DDR2 and have had no issues.
 
Vaulter98c said:
http://www.newegg.com/Product/Product.aspx?Item=N82E16820231225

Thats the kit I linked earlier, got 2 sets of that kit, ran fine on bigadv on my 5530's @ 208 bclk, mid divider; Havnt gotten a bigadv yet on my hexes, but I'm back to 195 bclk without tweaking anything and mem is doing just fine

EDIT
Here's the same thing in a 3x1 kit, save 45$, but I can only vouch for the 3x2 kits, 2 of them to be exact
http://www.newegg.com/Product/Product.aspx?Item=N82E16820231224
Click to expand...

This is the kit I am looking at. It does save nearly $100 and since this is a dedicated folder, every penny saved goes towards an upgrade of my sig rig.
 
Came home with 2x 6GB kits of Corsair 7-8-7-20 DDR3 instead. I'm waiting until my SR-2 finishes the A3 it's crunching now and will insert this RAM then. I'm just going to set the initial 7-8-7-20 parameters and then leave the rest of the memory settings on AUTO for the time being, just to see what I come up with. Of course, I'll set the -bigadv flag as well just to see if I can't get something to check this out tonight.

...should be interesting, but hey, this will also free up the 9-9-9-24 RAM I was using to go back in my i7-860 boxen, so I should be able to bring at least one of those back online to fold with. Hmmm, I guess I'm going to be short a power supply now. :rolleyes:
 
Axdrenalin said:
Came home with 2x 6GB kits of Corsair 7-8-7-20 DDR3 instead. I'm waiting until my SR-2 finishes the A3 it's crunching now and will insert this RAM then. I'm just going to set the initial 7-8-7-20 parameters and then leave the rest of the memory settings on AUTO for the time being, just to see what I come up with. Of course, I'll set the -bigadv flag as well just to see if I can't get something to check this out tonight.

...should be interesting, but hey, this will also free up the 9-9-9-24 RAM I was using to go back in my i7-860 boxen, so I should be able to bring at least one of those back online to fold with. Hmmm, I guess I'm going to be short a power supply now. :rolleyes:
Click to expand...
I can give you a good deal on a pc p&c 620w
Posted via [H] Mobile Device
 
If the fan in my 750TX didnt just die I'd hook you up
 
Ok nevermind my neverminded brain fart. :D
I'm going to share my findings.

Started having some more ram problems. Random ram disappearing on some boots...8 or 10gb instead of 12. (NUMA enabled) Dug around for 1600 12gb ram kits, noticed a few NewEgg reviews stating on the nicer Asus boards (the blue one with SATA 3...P5XD whatever) they had randomly disappearing ram on some boots, they dug around the Asus forums and were told to adjust their "Back-to-Back CAS Delay" from 0 (auto) to 8 or 10 to fix the problem. I thought - hey, maybe that's why I can't get all of my ram to work with NUMA disabled.

I start farting around with B2B CAS Delay in the SR-2 bios, fail a few boots (depending on the setting), come to a point where I think that B2B CAS Delay is disabled when NUMA is disabled, so I figure that's not it and post the above edit (nvm brain fart).

Wake up this morning, my client has another "client core communication error" on (of all things) a 2684 around the 88% mark. Sheist!!!!!!!!!!!!!!!!!!!!!!!!!!!!!~~! :mad: I am now at a point where I'm about 99.9% sure I'm going to be buying a new kit of ram because either a) the SR-2 does not like these Corsairs, b) they do not like the oc to 1600, c) I need to loosen the timings, or d) a combo of the above.

I went ahead and changed my timings to auto instead of putting in 9-9-9-24, and they booted at 11-11-11-29 (which according to MIBW is loose as a goose :)). I then started farting around (again) with B2B CAS Delay this morning before coming to work, and I found out using a tool called CPU-Tweaker that whatever you put in the bios, the board actually adds a 1. I initially put in 11 (since the CAS's are set at 11) and it booted and showed a B2B of 12 in CPU-Tweaker. I went back into the bios, changed it to 10, and it then booted and showed a B2B of 11 in CPU-Tweaker. I'm hoping that loosening my timings and setting the B2B CAS will keep me from losing ram randomly (or having any more of these friggin' client core communications error crash and flushes), but we'll see.
 
Last edited:
Thanks for the info.

I'll try that tonight and see if I get anything to report.

I'm finding my dual x5650s are EXTREMELY sensitive to RAM timings, more than any other Core i7 based CPU I've used.

Case in Point: at 4.11 Ghz (187x22) and RAM set to 2:8 divider for 1496 I had to change my timings to 11-11-11-25 1T to get things working, with no disappearing RAM with a VTT of 1.425 volts and full stability.

At 9-9-9-24 I'd get anywhere from 8 to 10GB of my 12 showing up at POST or in Windows 7 x64. Even at 10-10-10-25, 2T still disappearing RAM.

But, at 195x20 for 3.9Ghz with the same VTT, and all other settings the same, except a much lower VCore, RAM shows up at 9-9-9-24-1T without issue......

What's strange is that even though the chip's core is running slower, the BCLK is a fair bit higher and RAM timings can be tightened. This sorta flies in the face of logic for the "uncore" where a lower BCLK should stress the chip(s) less than higher... ?

This lends me to think the IMCs on these chips are extremely timing sensitive.
 
are the people having RAM issues running their RAM at their rated specs, or is your RAM overclocked or the timings tighter than normal?
 
Part of me is glad to see it is not just me that went through hell to get memory working well, but I would not wish that on anyone. Well not anyone in [H] :eek:

What MCH are you on Mr 10e? I hope you might be able to reduce vtt if you can find a good memory combo. 1.425 running 24/7 makes me bite my fingernails a bit.
 
jebo_4jc said:
are the people having RAM issues running their RAM at their rated specs, or is your RAM overclocked or the timings tighter than normal?
Click to expand...

In my case it's 1333 oc'd to 1600.
Granted I've had 0 issues with the ram in the past, but then again that was on a UD3R, not the SR-2. So in some capacity the SR-2 is at fault, perhaps just for splitting the ram up and maybe that is what's causing the problems. Or it's more picky (or the ram is more picky). Or something. I believe it's more or less that the SR-2 magnifies any potential problems due to the nature of the board, whereas regular x58's are a little more lax. I mean let's face it, the SR-2 is a workstation/server class board at it's core....and my little experience with servers is that they are picky and either work fine or they just crap all over themselves. :eek:

may i be worthy said:
What MCH are you on Mr 10e? I hope you might be able to reduce vtt if you can find a good memory combo. 1.425 running 24/7 makes me bite my fingernails a bit.
Click to expand...

Seconded, and I wonder if it's due to running a higher Uncore. IME lowering my uncore to 16x (which technically is what I should run since I'm using 2:8 ram divider) let me lower my Vtt by nearly .1v on my i7's.
 
zero2dash said:
I believe it's more or less that the SR-2 magnifies any potential problems due to the nature of the board, whereas regular x58's are a little more lax.
Click to expand...
right, this is what I'm wondering too.

It's possible the server/workstation class hardware is going to expose weaknesses much more quickly than x58 would.
 
jebo_4jc said:
right, this is what I'm wondering too.

It's possible the server/workstation class hardware is going to expose weaknesses much more quickly than x58 would.
Click to expand...

Yes, two QPI, multiple bus, and a different chipset (5520 vs x58) all have some effect on OC-ability from what I can see. Also, it's very difficult to isolate one CPU over another in a dual-CPU board. One CPU may be able to use different speeds for different volts, but it's a bit tougher to figure out which one :)

In terms of the other question, my RAM is rated for DDR3-1600, and has worked without issue at these voltages at 1633mhz at 8-8-8-24-1T timings in both dual and triple channel configs.

may i be worthy said:
Part of me is glad to see it is not just me that went through hell to get memory working well, but I would not wish that on anyone. Well not anyone in [H] :eek:

What MCH are you on Mr 10e? I hope you might be able to reduce vtt if you can find a good memory combo. 1.425 running 24/7 makes me bite my fingernails a bit.
Click to expand...

I've generally had the best results with MCH at 1600mhz, and the uncore at x12 (1600mhz). Anything lower on the uncore results in no boot, anything higher manifests in almost proportionately higher temps without performance change.

Having tried VTT at lower amounts with looser timings I can only go down to 1.40V if I set the RAM at 10-10-10-30-2T at BCLK of 195.

BTW I did add the extra 6-pin PCI-E plugs from my 850W PSU.

I'll see if that B2B CAS delay can help in my instance. Generally I've read that as long as VTT is lower than V-DIMM and within .5 volts the CPUs won't experience reverse bias conditions and the IMCs and rest of uncore should be safe. :should be" :)

But again, what's strange is if I set turbo to on and use the 22 multiplier, it does not allow me to relax VTT voltages. First time I've seen that.

While I'm not super overclocker I do have both my Core I7 970 and Core I7 980x on other boxes at 4.1 and 4.212 Ghz respectively at 100% stability. But again, this seems to be an almost entirely different animal.

Right now since I'm still trying to pump out max pointage I am letting her go. P2692 at 11:58 for nearly 150K PPD still keeps me [h]appy :)

I may try other kits that I have, including even a couple of 3GB kits all rated at 1600mhz.
 
10e said:
Also, it's very difficult to isolate one CPU over another in a dual-CPU board. One CPU may be able to use different speeds for different volts, but it's a bit tougher to figure out which one :)
Click to expand...

That's why there are jumpers to disable individual CPUs. ;) I haven't played with them while overclocking, but it was very helpful in isolating the bad DIMM slot on my first board.

I'm interested in your results, 10e. I've just run the Corsair Dominator 1600 at the specified 8-8-8-24 1.65V and haven't played around too much with loose timings. It works fine with the L5640s but the X5650s seem to be extra sensitive.

I'm also theorizing that the X5650s may just be low-binned hexes and therefore generally crappy overclockers. It seems to be incredibly painful to try to get the X5650s to 4GHz, but MIBW seemed to have much better results with his X5660s. If someone is shelling out $1K per Xeon anyway, it could be worth the extra cash for the X5660s. (I will gladly spend someone else's money to test this theory. :p )
 
Does anyone else have problems with fluctuating multipliers on their ebay special L5640s? Above 195ish bclck, the multiplier fluctuates between 15x and 25x under heavy load, which eventually causes a the system to crash. I have turbo/speedstep on and cstates off. However, turning off turbo/speedstep does not stop the multiplier from changing.

I've tried CPU PLL up to 1.8v, but increasing the PLL voltage seems to, if anything, make the fluctuations worse.
My settings are:
CPU Vc: 1.2v
CPU Vtt: 1.3v
IOH: 1.4v
RAM: 1.65v
CPU PLL: 1.65v

I've tried bumping the Vcore as high as 1.35v, which does seem to reduce the severity of the multiplier changes, but not enough to stabilize the system. I've also tried Vtt as high as 1.4v, which did not appear to help at all.
 
Last edited:
SazanEyes said:
That's why there are jumpers to disable individual CPUs. ;) I haven't played with them while overclocking, but it was very helpful in isolating the bad DIMM slot on my first board.

I'm interested in your results, 10e. I've just run the Corsair Dominator 1600 at the specified 8-8-8-24 1.65V and haven't played around too much with loose timings. It works fine with the L5640s but the X5650s seem to be extra sensitive.

I'm also theorizing that the X5650s may just be low-binned hexes and therefore generally crappy overclockers. It seems to be incredibly painful to try to get the X5650s to 4GHz, but MIBW seemed to have much better results with his X5660s. If someone is shelling out $1K per Xeon anyway, it could be worth the extra cash for the X5660s. (I will gladly spend someone else's money to test this theory. :p )
Click to expand...

I thought about using the jumpers as well, but I haven't yet. It seems like chip 2 works better with slightly higher VCore and VTT.

I've been able to lower VCore to 1.34375 for chip 1, and 1.3500 for chip 2 and VTT on chip 1 is now 1.4v. It works here by me simply lowering the BCLK by a whole 1 mhz to 194. Strangely enough GFlops in LinX went UP by 2 - 3 Gflops / sec.

Weird.

But I would hasten to agree with your assertion, that the x5650s are lower binned Westmeres thus meaning lower OC-ability.

Too bad all those special L5640s have dried up, but I'm sure with their lower multi, would be just as much of a challenge to get up to this speed.

I'm going to switch out of LinX and starting up FAH again and seeing what happens. If performance goes up or stays the same, I think this will be the sweet spot (3860 - 3880 mhz) for the chips.

thefreeaccount said:
Does anyone else have problems with fluctuating multipliers on their ebay special L5640s? Above 195ish bclck, the multiplier fluctuates between 15x and 25x under heavy load, which eventually causes a the system to crash. I have turbo/speedstep on and cstates off. However, turning off turbo/speedstep does not stop the multiplier from changing.

I've tried CPU PLL up to 1.8v, but increasing the PLL voltage seems to, if anything, make the fluctuations worse.
My settings are:
CPU Vc: 1.2v
CPU Vtt: 1.3v
IOH: 1.4v
RAM: 1.65v
CPU PLL: 1.65v

I've tried bumping the Vcore as high as 1.35v, which does seem to reduce the severity of the multiplier changes, but not enough to stabilize the system. I've also tried Vtt as high as 1.4v, which did not appear to help at all.
Click to expand...

I've actually lowered my PLL to 1.725 without any ill-effects.

I thought the turbo was going to go on depending on thermals? Ie. lower than a certain TDP and turbo will kick in properly?

What are your temps like?

EDIT: Have you also set Windows to be at full power/High Performance in the power control panel? This can cause that behavior too.
 
Last edited:
What are your temps like?
Click to expand...
They are fine - if anything, fairly chilly. Less than 50C across the board, except for CPU0 VRM, which gets as high as 55C. I'm using magahalems with some 3000rpm deltas.

I thought the turbo was going to go on depending on thermals?
Click to expand...
Turbo for my L5640s is 18x, which is where they stay most of the time. If temps were the problem, I would not expect to see the multiplier increase from 18x to 24x.
 
Picked up a 2684 early this morning, it went to 24% before it crashed out with an UNSTABLE_MACHINE error. Once I get thome tonight I'll probably bump up the core voltage a little bit to see if that helps (set at 1.325 right now). Anyone else have any quick suggestions that might help?

At least now I'm not getting that big memory error that I was seeing before, and it's actually processing *something* in the -bigadv arena. :)

Ax
 
So I checked out this Back-to-back CAS delay setting.

On the folding front the news is good.

I set the b2b CAS delay setting to a level of 5 and without changing VCore, VTT, or other memory timings I was able to bring BCLK up by 3 mhz from 195 to 198 and retain full stability.

I tried 3 but it didn't really help, so I incremented it up by 1 until I hit stability at 198 and this happened with the 5 setting.

With my 20 multi this brought the chips up to 3.96Ghz from 3.9 which is the fastest I've run them with a 20 multiplier without raising the VCore to 1.41 (needed for 4Ghz).

Alternatively LinX showed a slight decline in Gflops from 78 - 79 to 75 - 76, but that doesn't affect folding generally. This just indicates a slight decline in memory bandwidth. That's fine.

After continuing my P2692 from before I'm seeing a slight decrease in TPF by another few seconds to around 11:55 which is better than 3.9Ghz (11:58 average) so the slight decrease in memory bandwidth was offset by the slight increase in CPU clocks.

Thanks for the info Zero2Dash.
 
Tobit said:
MIBW: Did you ever get a decent overclock working when running 12 sticks of memory? Looking for ways to get to 24GB without spending a premium on 6x4GB DDR3-1600 sticks or going the cheap hex route and buying 6x4GB DDR3-1333.
Click to expand...
tobit, you still looking for 6x4gb?

Newegg has a 48 hour sale on single modules for $80 each. Which, by my math, would save you $50 vs the kit.

Limit 5, but if you want another one, I will buy one and ship it to you.

RAM in question
Sale link
 
jebo_4jc said:
tobit, you still looking for 6x4gb?
Click to expand...
Good stuff man, thanks. However, I wasn't planning on buying memory til the 1st of the month. Will have to see what kind of deals are around come then.
 
Also have Gskill 1333 mem w/ 2x L5640s ES in SR-2 using A49 Bios
... BUT ...
Max BCLK=170 (using EXACTLY same settings)

Using 12 x 4 GB DIMMS (48 GB)
Perhaps this is reason for difference?

Can you please let us know what size DIMMs you are using to get BCLK=200 with this Gskill memory?
 
You must log in or register to reply here.
Back
Top