DDR3-2000 (9-10-10-24-1T)

[BePolite]

I managed to get some very cheap Samsung 'green RAM' back in the day for about €30 second hand. This was a great price for 16GB 2 years ago here on the emerald isle. It over clocks well at 1.65v but the sweet spot I am finding is 2000 as opposed to 1866 or 2133.

My question is: is there any reason not to run it at 2000, it seems when you search that frequency almost never comes up.

I'm running a 3570K @ 4.4Ghz on a Z77 Sabertooth.

Many Thanks.

 

[FreedomEclipse]

If it can handle the speed then i dont see why not

 

[BePolite]

Many Thanks! I just wondered if there was some odd timing issue at 2000. I suppose that would be documented if there where. Thanks for your reply

 

[Aquinus]

Well, from a performance standpoint there isn't much reason to overclock the memory. 1600 is more than enough for your CPU. Granted I'm not one to talk considering my memory is stock at 2133 and it's running at 2400, but I don't really see any performance improvement from it.

How are you achieving 2000Mhz? I personally don't have a divider to hit 2000Mhz without adjusting the bclk strap to 125Mhz, which you can't do with your CPU.

 

[FreedomEclipse]

nope, some games and apps like the extra bandwidth/speed that comes with faster ram. also $30 is a lot better then buying 2133mhz at todays inflated prices, the price has doubled if not tripled

 

[BePolite]

Well, from a performance standpoint there isn't much reason to overclock the memory. 1600 is more than enough for your CPU. Granted I'm not one to talk considering my memory is stock at 2133 and it's running at 2400, but I don't really see any performance improvement from it.

How are you achieving 2000Mhz? I personally don't have a divider to hit 2000Mhz without adjusting the bclk strap to 125Mhz, which you can't do with your CPU.

The Z77 Sabertooth just has the divider, oddly. My bclk is the standard 100Mhz, to be fair this is what got me thinking 2000 might be some sort of odd speed with it's own drawbacks.

Funnily enough memtest reports it as 2666 so again something made me wonder. CPUZ reports it correctly in Windows.

 

[Johan45]

Best way to find out your performance gain/loss is with Aida64

 

[Sasqui]

Best way to find out your performance gain/loss is with Aida64

Exactly. It's a dance between CAS and speed. Look at some of the older DDR3 memory reviews by Darksaber, he published charts with plenty of different CAS and Speed timing mixes.

 

[BePolite]

Exactly. It's a dance between CAS and speed. Look at some of the older DDR3 memory reviews by Darksaber, he published charts with plenty of different CAS and Speed timing mixes.

This is how I arrived at 2000 tbh - at 1866 I couldn't quite squeeze CL 8 out of it so I bumped up to 2000 with CL9. Seems there is no issue with 2000 per se so given the benches I'm getting and memtest runs 24 hours with no errors I think I'm good.

Thanks again to everyone who posted this forums is so helpful!!!

 

[Sasqui]

This is how I arrived at 2000 tbh - at 1866 I couldn't quite squeeze CL 8 out of it so I bumped up to 2000 with CL9. Seems there is no issue with 2000 per se so given the benches I'm getting and memtest runs 24 hours with no errors I think I'm good.

Sounds like it! Prime95 also is a good sideline memory test if you're so inclined. I tried getting 2400 out of my sticks and couldn't even do CL12. Since they work flawlessly at the advertised profile 2133 C9 1.65v, I'm happy with them.

 

[cadaveca]

Exactly. It's a dance between CAS and speed. Look at some of the older DDR3 memory reviews by Darksaber, he published charts with plenty of different CAS and Speed timing mixes.

I did too, until all the ICs were on the market. Once we knew what was out there, there wasn't much point in providing such info any more. OEMs had already started binning everything tightly.

Samsung LP/LV sticks got that testing though:

http://www.techpowerup.com/reviews/Samsung/MV-3V4G3/7.html


I'll do it again for DDR4, too. I'm actually pretty eager to see how it works at launch, gonna be pretty interesting.

I say go for 9-11-10-27 @ 1.5V 2133 MHz with these sticks as optimum. I can list some timings for your board if you post up BIOS screenshots, too, BePolite.

 

[vega22]

i have some corsair sticks that love 2000mhz too, cas 7 on 1.7vdimm

anything 1166 or later hates them

 

[MyNewRig]

There is performance gains from faster RAM in apps that are CPU bound and in many games, although i admit that there are diminishing returns, i see some games improve their minimum FPS very significantly with faster RAMs ..

For example i have been benching Thief Master Thief Edition, and on 1333Mhz the minimum FPS was 29.6FPS on 2400Mhz it was 34.1FPS ... on 2133Mhz it goes up to 38FPS

average FPS remain almost the same with like 1FPS difference between fastest and slowest frequencies, Max FPS gain like 2FPS from faster RAM ..

The increase in the minimum FPS means smoother game-play, i have seen another thread few months ago of a guy who done massive tests as well and he established that although average FPS remain almost unchanged, minimum FPS does improve a lot with faster RAM.

this title is not CPU bound but it does use the CPU extensively, i have other CPU bound apps that show a lot of difference, system boot times improves a lot with faster RAM, overall responsiveness when multi-tasking improves ..

Like discussed with cadaveca in another thread, 2133Mhz CL9 seems to be the sweet spot, above that sometimes you get worse results ..

 

[Vario]

Pretty good deal there ^ Ive found that as a rough guess, increasing ram speed from 1600 to 2133 usually seems about the same as +100 mhz cpu clock.

OP: I have 5 sticks of green. Might pick up a 6th to match the set, you can find it on ebay mislabeled sometimes, just make sure not to get the ECC version.

I have found they all perform identically and consistently, usually 1.5v even to get 2133 11-11-11-33 1t lazy overclock. I never dialed them in any tighter due to laziness. Sadly they run horribly on my Gigabyte Z77X-UD3H main rig, despite trying custom bioses etc. Did you adjust your secondary and tertiary timings? I would be interested to know the full timings you run so I can duplicate on itx. Was 1.65v necessary?

 

[BePolite]

I think I've misled people sorry - this is green RAM by virtue of colour rather than the 'green' stuff floating about. It's a different part number I think:

M378B5273DH0-CH9 is what I have.

Sorry.

 

[cadaveca]

I think I've misled people sorry - this is green RAM by virtue of colour rather than the 'green' stuff floating about. It's a different part number I think:

M378B5273DH0-CH9 is what I have.

Sorry.

No apology necessary. Same stuff.

Well, earlier version of the same stuff. Most users are familiar with HYK0, while you have HCH9. The ICs are slightly different, but general behavior is very close to other Samsung-made ICs.