On a faaast lane --> Corsair TWIN2X2048-6400C3 OCing Report on Socket AM2!

Printable View

06-30-2006, 10:57 AM
bachus_anonym

On a faaast lane --> Corsair TWIN2X2048-6400C3 OCing Report on Socket AM2!

Heyloo guys! :toast:

Yep, you right... We're gonna have another OCing Report and this time I'm focusing on Corsair's latest-greatest 2GB DDR2 set of low latency TWIN2X2048-6400C3.
This is the first ever PC2-6400 memory kit to be rated at CAS3 and also the one to feature EPP (Enhanced Performance Profiles), new memory standard that Corsair and NVIDIA came up with. The kit is rated at DDR2-800 9-3-4-3.0 @ 2.2v and built with highly screened for low-latency performace, Micron's new D9xxx ICs. After Micron's legendary D9 "Fatbody" was discontinued, memory manufacturers went searching for the next best thing. They sampled few other of Micron's ICs, like D9GKX, D9GHM, D9DCN... Some in-house testing, screening and sorting later... bang!... Le Roi est mort. Vive le Roi! :)

Anyway, for those of you that don't know yet what EPP is about, here a short explanation:

Quote:

Jointly developed by Corsair and NVIDIA® as a new open memory standard, Enhanced Performance Profiles elevate memory performance by taking full advantage of the additional memory parameters (such as Command rate, memory voltage) that are now added to the unused portion of the JEDEC standard Serial Presence Detect (SPD). Memory modules from Corsair support two types of performance profiles that are based upon a high frequency or low latency design. Enhanced Performance Profiles brings simplified overclocking success for entry enthusiasts, and access to more advanced memory specificationsand a higher recovery point to hard-core overclockers.

Unfortunately, M2N-SLI Deluxe BIOS doesn't support this feature yet and I'm honestly not sure if it ever will. From what I gathered, only nForce590 SLI chipset supports it, so no EPP for me... I would most likely not use it anyway, but I'm sure there are quite a few beginners that are just starting their adventure with overclocking. With help of EPP and compatible motherboards, they would have a good start and a convenient way to get their overclocking adventure going.

Let's take a look at modules, shall we?

http://www.clockmehigh.com/cooler/sc...00C3_front.jpg

http://www.clockmehigh.com/cooler/sc...400C3_back.jpg

http://www.clockmehigh.com/cooler/sc...00C3_cross.jpg

http://www.clockmehigh.com/cooler/sc...00C3_label.jpg

This OCing Report will focus at two aspects of memory overclocking: MAX all-stable and MAX 1M/Sandra only DDR2 clocks. The platform I've been testing this set on, delivers up to 2.5vdimm and this is how far I went to get the best out of this RAM. I will be soon volt-modding this board to allow 2.7+ vdimm and then try to hit even higher speeds ;)

Configuration of my platform is as follows:

Testbed Configuration
AMD AM2 3200+ "Orleans"
ASUS M2N-SLI Deluxe (nForce570 SLI)
Corsair TWIN2X2048-6400C3 2x1024MB
Single EVGA 7800GTX KO @ 490+40/1390, ForceWare 84.21, Modified driver settings (High Performance)

Let's get going with the first part of my stress-testing:

Part1: High-stress testing

MAX fully stable DDR2 clocks @ 8-3-3-CAS3.0 ---> DDR2-882 @ 2.5v

SPi32M -------------------------------------------------- SP2004 LargeFFT 1024MB used
http://www.clockmehigh.com/cooler/sc..._32M_1_4x3.jpg http://www.clockmehigh.com/cooler/sc...2004_1_4x3.jpg

3DMark2001 & 3DMark03 -------------------------------- 3DMark06
http://www.clockmehigh.com/cooler/sc...DM03_1_4x3.jpg http://www.clockmehigh.com/cooler/sc...DM06_1_4x3.jpg

6400C3 reached very impressive, fully stable DDR2-880. Past that, I was not able to pass SPi32M, nor required 2hrs of SP2004. Nevertheless, very good result at 8-3-3-3.0!

MAX fully stable DDR2 clocks @ 8-3-4-CAS3.0 ---> DDR2-940 @ 2.5v

SPi32M -------------------------------------------------- SP2004 LargeFFT 1024MB used
http://www.clockmehigh.com/cooler/sc..._32M_1_4x3.jpg http://www.clockmehigh.com/cooler/sc...2004_1_4x3.jpg

3DMark2001 & 3DMark03 -------------------------------- 3DMark06
http://www.clockmehigh.com/cooler/sc...DM03_1_4x3.jpg http://www.clockmehigh.com/cooler/sc...DM06_1_4x3.jpg

After relaxing to TRCD=4 I managed to climb a bit higher, up to DDR2-940 and still at CAS3. This is 17.5% overclock from stock DDR2-800, which is quite sweet, isn't it? :)

MAX fully stable DDR2 clocks @ 8-4-4-CAS4.0 ---> DDR2-1100 @ 2.5v

SPi32M -------------------------------------------------- SP2004 LargeFFT 1024MB used
http://www.clockmehigh.com/cooler/sc...ight_1_4x3.jpg http://www.clockmehigh.com/cooler/sc...2004_1_4x3.jpg

3DMark2001 & 3DMark03 -------------------------------- 3DMark06
http://www.clockmehigh.com/cooler/sc...DM03_1_4x3.jpg http://www.clockmehigh.com/cooler/sc...DM06_1_4x3.jpg

Yep, DDR2-1100 has been broken with this kit and 2.5vdimm. Good for the eyes and very good for performance as well :)

Part2: Low-stress testing

MAX benchmark stable DDR2 clocks @ 8-3-3-CAS3.0 ---> DDR2-920 @ 2.5v

SPi1M --------------------------------------------------- Sandra 2007 Memory Bandwidth
http://www.clockmehigh.com/cooler/sc...0_1M_1_4x3.jpg http://www.clockmehigh.com/cooler/sc...aMem_1_4x3.jpg

Yeah, DDR2-920 8-3-3-3.0 for you 1M'ers :up:

MAX benchmark stable DDR2 clocks @ 8-3-4-CAS3.0 ---> DDR2-956 @ 2.5v

SPi1M --------------------------------------------------- Sandra 2007 Memory Bandwidth
http://www.clockmehigh.com/cooler/sc...6_1M_1_4x3.jpg http://www.clockmehigh.com/cooler/sc...aMem_1_4x3.jpg

How about DDR2-956, with slightly relaxed TRCD=4 and still CAS3.0 ?

MAX benchmark stable DDR2 clocks @ 8-4-4-CAS4.0 ---> DDR2-1112 @ 2.5v

SPi1M --------------------------------------------------- Sandra 2007 Memory Bandwidth
http://www.clockmehigh.com/cooler/sc...2_1M_1_4x3.jpg http://www.clockmehigh.com/cooler/sc...aMem_1_4x3.jpg

And some more in 8-4-4-CAS4.0 compartment...

Conclusion

So what can I say? TWIN2X2048-6400C3 has some serious OCing potential and will sure please a lot of hardcore OCers out there. I could not think of any disadvantage of this set, except for the price... Extensive screening and sorting of already expensive Micron D9 dies that this set is made of, contributes to it's premium price. It sure isn't cheap but If one wants some of the greatest then one often needs to pay for it. It's a kickass performer, not only at low latency but high frequency overclocks as well! You name it, 6400C3 delivers! :)

July 1st UPDATE!

MAX SPi32M stable DDR2 clocks @ 8-3-3-CAS3.0 ---> DDR2-900 @ 2.5v

SPi32M
http://www.clockmehigh.com/cooler/sc..._32M_1_4x3.jpg

MAX SPi32M stable DDR2 clocks @ 8-3-4-CAS3.0 ---> DDR2-950 @ 2.5v

SPi32M -------------------------------------------------- 3DMark2001 & 3DMark03
http://www.clockmehigh.com/cooler/sc..._32M_1_4x3.jpg http://www.clockmehigh.com/cooler/sc...DM03_1_4x3.jpg

Any questions or requests are welcomed,

So on :peace:
06-30-2006, 11:05 AM
G H Z

bacus,

Excellent write up with detailed testing, good job :toast:
06-30-2006, 11:08 AM
bogdannovaceanu

XMS 6400C3s: DDR2-1150 3-3-3-1!
http://www.xtremesystems.org/forums/...d.php?t=103629
06-30-2006, 11:24 AM
bachus_anonym

Thanks G H Z :)

Quote:

Originally Posted by bogdannovaceanu

XMS 6400C3s: DDR2-1150 3-3-3-1!
http://www.xtremesystems.org/forums/...d.php?t=103629

I said that before in your thread and few others and I will say it again... No offense, Bogdan, but it's IMPOSSIBLE. Try it on a 1066fsb strap or a different mainboard. Not BadAxe, please.
06-30-2006, 01:40 PM
Jaco

kickass memory.
any idea what kind of D9 flavor is used here?
I hope we will figure out all D9 variants and their oc results
06-30-2006, 04:40 PM
Exedy

2.5v on bios?

ASUS M2N-E Socket AM2 NVIDIA nForce 570 Ultra can hit that too? I hear that voltage is up to 1.95v only =(

good results :)
06-30-2006, 05:04 PM
hipogtr

Great writeup. Looks like some quality D9. Thanks.
06-30-2006, 07:09 PM
^don.k's^

Nice sticks & tests bachus, i'd like to test ones like that :p: But they are "$%"·!· expensive lol

Regards! :toast:
06-30-2006, 07:17 PM
nn_step

damn I would love to see some latency numbers on those sticks
07-01-2006, 07:03 PM
bachus_anonym

Thanks guys! :toast:
So, I've been messing around a little more today to see If I can get anything more 32M stable and still stay at CAS3.0... I managed to squeeze out what's left at this voltage :)

I have updated my 1st post under July 1st UPDATE! with new results:

MAX SPi32M stable DDR2 clocks @ 8-3-3-CAS3.0 ---> DDR2-900 @ 2.5v

MAX SPi32M stable DDR2 clocks @ 8-3-4-CAS3.0 ---> DDR2-950 @ 2.5v

Next up, vdimm mod on Monday and we'll see what 2.7-2.8v can do to those sticks :)

Quote:

Originally Posted by ocfx

2.5v on bios? ASUS M2N-E Socket AM2 NVIDIA nForce 570 Ultra can hit that too? I hear that voltage is up to 1.95v only =(

Yes, up to 2.5v in BIOS for M2N-SLI and M2N32-SLI. You can vdimm mod your M2N-E. Another member here, ^don.k's^ already did it and is flying high around 3.0v...
07-01-2006, 07:47 PM
[XC] 4X4N

Bachus, your results are kick ass as usual. :toast: Are you going to stay with am2 or will you go for conroe? Just curious as I am on the fence, and will probably wait for rev g, but conroe is hard to resist.
07-01-2006, 07:50 PM
Exedy

Bachus, did you try max HTT for ASUS M2N-SLI Deluxe?

thx
07-01-2006, 07:56 PM
bachus_anonym

Quote:

Originally Posted by 4X4N

Bachus, your results are kick ass as usual. :toast: Are you going to stay with am2 or will you go for conroe? Just curious as I am on the fence, and will probably wait for rev g, but conroe is hard to resist.

I'm actually going both :) It's going to be fun, for benching and some DDR2 testing as well...

Quote:

Originally Posted by ocfx

Bachus, did you try max HTT for ASUS M2N-SLI Deluxe?thx

Yes, Sir! Nailed 380-400HTT - http://www.xtremesystems.org/forums/...d.php?t=102595
07-01-2006, 08:29 PM
Exedy

Quote:

Originally Posted by bachus_anonym

Yes, Sir! Nailed 380-400HTT - http://www.xtremesystems.org/forums/...d.php?t=102595

omg! :slobber:

I guess K9N SLI is nothing for me now :D
07-02-2006, 07:57 AM
dsl

Nice work bachus, I always enjoy your excellent mem ocing reports :)
07-02-2006, 12:58 PM
Deathspawner

Quote:

Originally Posted by bachus_anonym

This is the first ever PC2-6400 memory kit to be rated at CAS3 and also the one to feature EPP (Enhanced Performance Profiles), new memory standard that Corsair and NVIDIA came up with.

Just because Corsair was the only one NVIDIA mentions, they were definitely not the -only- ones to help in building the profiles.

That aside... great results as always :D
07-02-2006, 08:15 PM
ZX7891

finally ddr2 is starting to outpace the best bh5
07-02-2006, 09:07 PM
3oh6

excellent write up BA, as usual from what i have seen in the short time i have been lurking around here. these 6400 C3s seem to be the toast of the town right now. i can hardly wait to see what OCZs new rainbow warriors can do compared to these bad boys.

i have a 2gb set of the OCZ 7200 EPPs and they seem to perform as good as these at DDR1100 4-4-3-x and higher but can't touch DDR900 C3...that is just ridiculas.
07-02-2006, 09:16 PM
bachus_anonym

I'm glad you guys are enjoying this test and results with 6400C3. Thanks! :toast:

Quote:

Originally Posted by Deathspawner

Just because Corsair was the only one NVIDIA mentions, they were definitely not the -only- ones to help in building the profiles.

You're right... I just quoted what I found online :) But I suspect, since OCZ has EPP-compatible memory coming they also co-operated with NVIDIA on development of this new standard.
07-02-2006, 09:26 PM
dinos22

can you not set CMD=1T for AM2?
07-04-2006, 09:52 AM
bachus_anonym

Quote:

Originally Posted by dinos22

can you not set CMD=1T for AM2?

Well, not really. I got only up to DDR2-830 1T and OPP, k|ngp|n and afireinside got a bit at more at 1T out of their setups, I think. Then ^don.k's^ got his "fatbody" to DDR2-850 1T @ 2.7v. But that's about it as far as command Rate 1T goes. In order to hit anywhere higher, you need 2T. But that's not really much of a performance hit on AM2, from what it seems. So, I'd not worry about that :)
07-05-2006, 08:22 AM
^don.k's^

Quote:

Originally Posted by bachus_anonym

Well, not really. I got only up to DDR2-830 1T and OPP, k|ngp|n and afireinside got a bit at more at 1T out of their setups, I think. Then ^don.k's^ got his "fatbody" to DDR2-850 1T @ 2.7v. But that's about it as far as command Rate 1T goes. In order to hit anywhere higher, you need 2T. But that's not really much of a performance hit on AM2, from what it seems. So, I'd not worry about that :)

For ddr2-850 1T i didnt use the fatty's, i used the ballistix. I used fatty's for ddr2 1152 4-3-3-5 2T :D
07-05-2006, 09:43 AM
informal

bachus_anonym ,really nice effort man.It must have been time consuming to try out all the combinations,and find and measure the benefits.Thanks:)
07-05-2006, 11:25 AM
dz_jad

I enjoy reading these :) I'm gonna see how my ram on the way overclocks, I've got 2 sticks of ddr2 1066 on the way (I think, I've read so much I've forgotten what I've ordered). Its CAS-5 at stock. It'll be my first time, so I've gotta do some studying!
Thanks :)
07-08-2006, 06:32 PM
akijikan

BA FTW! Awesome thorough job!

I have the C4 version...but I haven't been able to push it because the Biostar Tforce 550 seems to lack stability about 290 HTT and the VDimm only goes to 2.2 if you use the jumper! lol
10-07-2006, 10:32 AM
Clones123

Corsair IC Inquiry Thread

I found the information below interesting. What's with the variety of DIMM's all using the same IC's ("ProMOS 667MHZ, CL5, 0.09u 64Mx8 ETT") ?? Why pay extra for the 6400C4 when the cheaper standard 6400 uses the same IC's - is there any difference?

---------------------------

Corsair IC Inquiry Thread
http://www.houseofhelp.com/forums/s...ead.php?t=44365

Value Select Modules
Value Select modules are built with RAMs that are procured at the wafer level and are packaged and tested to our specifications. As such, they are not available as individual ICs, only as modules. This approach allows us to ship product with very good performance characteristics, low product cost, and a very low return rate.

DDR2 - XMS
XMS, or eXtreme Memory Speed, is the process that Corsair uses to take ICs rated at one speed and verify/guarantee their operation at another speed. Since specifications have not been generated to accurately specify operating parameters, Corsair gives these parts an XMS rating. For example, since PC2400 does not exist according to any valid standards institution, Corsair parts are called XMS2400.

5400C4
CM2X1024-5400C4 XMS5402 v2.10 0628134-2 = Micron, 64Mx8, DDR2-533, CL4, Die Rev D
CM2X1024-5400C4 XMS5402 v5.2 0633071-3 = PCS DDR2, 675MHz, CL 4, 64x8 90nm ETT
CM2X1024-5400C4 XMS5402 v5.2 0633071-5 = PCS DDR2, 675MHz, CL 4, 64x8 90nm ETT
CM2X1024-5400C4 XMS5402 v7.1 0629230-11 = ProMOS 0.09u 64Mx8 ETT Rev E

6400
CM2X1024-6400 XMS6405 v1.4 0606096-2 = Micron, 64Mx8, DDR2-667, CL5, Die Rev D
CM2X1024-6400 XMS6405 v4.1 0609102-1 = Infineon AENEON, Die Rev B, 64M X 8 Rev B
CM2X1024-6400 XMS6405 v4.1 0626413-3 = Infineon AENEON, Die Rev B, 64M X 8 Rev B
CM2X1024-6400 XMS6405 v5.1 0634287-2 = ProMOS 667MHZ, CL5, 0.09u 64Mx8 ETT
CM2X1024-6400 XMS6405 v5.1 0631252-2 = ProMOS 667MHZ, CL5, 0.09u 64Mx8 ETT
CM2X1024-6400 XMS6405 v5.1 0631252-9 = ProMOS 667MHZ, CL5, 0.09u 64Mx8 ETT
CM2X1024-6400 XMS6405 v5.1 0633171-6 = ProMOS 667MHZ, CL5, 0.09u 64Mx8 ETT
CM2X1024-6400 XMS6405 v5.1 0629230-7 = ProMOS 667MHZ, CL5, 0.09u 64Mx8 ETT

6400C3
CM2X1024-6400C3 XMS6403v1.1 0619028-4 = Micron, 64Mx8, 2-667 CL5, Die Rev D

6400C4
CM2X1024-6400C4 XMS6404 v1.1 0626099-3 = Micron, 64Mx8, 2-667 CL5, Die Rev D
CM2X1024-6400C4 XMS6404 v1.2 0620097-1 = Micron, 64Mx8, 2-667 CL5, Die Rev D
CM2X1024-6400C4 XMS6404 v2.1 0635221-22 = ProMOS 667MHZ, CL5, 0.09u 64Mx8 ETT

6400PRO
CM2X1024-6400PRO XMS6405v1.1 0610028-1 = Micron, 64Mx8, DDR2-667, CL5, Die Rev D
CM2X1024-6400PRO XMS6405v3.1 0609115-2 = Infineon AENEON, Die Rev B, 64M X 8 Rev B
CM2X1024-6400PRO XMS6405v1.3 0610028-1 = Micron, 64Mx8, DDR2-667, CL5, Die Rev D

8500C5/C4
CM2X1024-8500C5 XMS8505v1.2 0620114-1 = Micron, 64Mx8, 2-667 CL5, Die Rev D
CM2X1024-8888C4D XMS8804v1.1 0634174-2 = Micron, 64Mx8, 2-667 CL5, Die Rev D
10-08-2006, 07:24 AM
Bluebeard

difference is binning and guaranteed speeds - binning is what makes it expensive

DDR2-667 CL5, Die Rev.D so it is Micron D9GMH

Any other D9GMH could do but to find exactly 2 Modules where 32 (2 times 16) unbinned chips meet those specs is almost impossible...
10-08-2006, 07:58 AM
vudoodoodoo

Interesting... how much are these?
10-08-2006, 08:01 AM
Bluebeard

:confused:

i´m talking about Die/IC/chip right now - not about modules...

if you´d like to know price for Corsair 6400C3 you´ll easy find out at any pricewatch
10-08-2006, 08:31 AM
kutujambak

nubie question...what is binning?
10-08-2006, 10:21 PM
Bluebeard

testing the chips for speed
10-08-2006, 11:07 PM
erwinz

nice bachus_anonym :D

excellent work.. :D

:woot:
10-09-2006, 11:07 AM
purecain

bachus- top review my man,
i have a question for you(and anyone else who's reading this), have you used any xms2-twin2x-pc2-8500c5.
i have a 2gb matched pair and am unsure of the actual chips used. i'd like to find out how far i can safely push them, and their max rated speed.
bit off topic but would appreciate any advice....
10-10-2006, 04:41 PM
Clones123

Quote:

Originally Posted by kutujambak

nubie question...what is binning?

"Speed binning" actually.

The contracted plant takes thousands of 667MHZ CL5 IC's, tests and sorts them into separate bins for 533MHz, 667MHz, 800MHz, 1000MHz ratings.

Later, Corsair tests the finished PCB's to refine them under standard, C3, C4, and Pro lables and then matches them into pairs of DIMMS with near-identical performance. That's my understanding (not that I've actually seen it).
10-11-2006, 05:06 PM
Redbeard

Quote:

Originally Posted by Clones123

"Speed binning" actually.

The contracted plant takes thousands of 667MHZ CL5 IC's, tests and sorts them into separate bins for 533MHz, 667MHz, 800MHz, 1000MHz ratings.

Later, Corsair tests the finished PCB's to refine them under standard, C3, C4, and Pro lables and then matches them into pairs of DIMMS with near-identical performance. That's my understanding (not that I've actually seen it).

Pretty close to right, we do have some ability to test individual ICs in-house in Fremont, CA, and we do a lot of internal testing on various motherboards to verify this. You'd be surprised how something from a tester may say it does X MHz and when we put it on a motherboard sometimes it's twice or half that fast. Weird.

In order of easiest to make to hardest to make with Micron ICs.

5400C4
6400C5
6400C4
8500C5
6400C3
8888C4

Yield on the 6400C3 is very, very low. 8888C4 yield is about half of that.

When you are buying hundreds of thousands of ICs at a time and screening them for various grades, the labor and time can cost a pretty penny. So while the 8888C4 uses the same basic Micron IC that a 6400C4 might use, the challenge of screening them so extensively and getting enough to build parts that we can actually put in stores is so high that they have to be sold at a premium otherwise we'd never be able to keep up with demand and make a profit on them.
10-12-2006, 12:14 PM
MACMAC

Quote:

Originally Posted by Alonso

difference is binning and guaranteed speeds - binning is what makes it expensive

DDR2-667 CL5, Die Rev.D so it is Micron D9GMH

What makes you think it's D9GMH?

It could be any D9XXX, no? :confused:
10-14-2006, 12:37 PM
fhpchris

Quote:

Originally Posted by MACMAC

What makes you think it's D9GMH?

It could be any D9XXX, no? :confused:

D9GKX is 800 mhz CL5 :)

667 cas 5 is D9GMH's specification.
10-14-2006, 07:23 PM
kutujambak

@clones123 and redbeard
thx for explaining
10-14-2006, 10:06 PM
charlie

so the bottom line is...

8888C4 is the solid freekin 18K Gold D9GMH?? The creme de la creme of D9's? Combine that with a HiPro Maximizer and you've got 500+ @ 3-3-3

C
10-15-2006, 05:18 PM
Clones123

Quote:

Originally Posted by Redbeard

In order of easiest to make to hardest to make with Micron ICs.

5400C4
6400C5
6400C4
8500C5
6400C3
8888C4

Redbeard,

Corsair's 5400C4, 6400C5 and 6400C4 are all now made with ProMOS 667MHZ CL5 IC's and not Micron, right? Personally, I'm a big fan of Team Group memory from TankGuys.biz just because I can be sure it's Micron D9 IC's.

Source: Corsair IC Inquiry Thread
http://www.houseofhelp.com/forums/s...ead.php?t=44365
10-16-2006, 09:27 AM
Redbeard

Quote:

Originally Posted by Clones123

Redbeard,

Corsair's 5400C4, 6400C5 and 6400C4 are all now made with ProMOS 667MHZ CL5 IC's and not Micron, right? Personally, I'm a big fan of Team Group memory from TankGuys.biz just because I can be sure it's Micron D9 IC's.

Source: Corsair IC Inquiry Thread
http://www.houseofhelp.com/forums/s...ead.php?t=44365

We build those modules with whatever we can build them with, and the ProMOS stuff is very good at lower frequencies right now, especially with the difficulty of finding enough Microns to build our other stuff.

Still, the stuff that doesn't pass at 8500C5 or 6400C3 will get downbinned once in a while and made into other parts.
10-19-2006, 06:47 AM
Complete

Update on this memory!!

I have a pair of these amazing sticks.

Give them v2.5 and I got them with alittle air cooling to:

3.3.3.5 @ 990

WOW!! Ill post a screen shot later!
11-29-2006, 12:31 PM
rvd

sorry to bump this thread, but are these available anywhere?
11-29-2006, 06:08 PM
clon22

Hmm..
My sticks dun seem to like 3-3-3 settings..
But they take 3-4-3...
So i got no complains either.. :D