Samsung 30 nm Green PC3-12800 Low Profile 1.35 V DDR3

[cadaveca]

If you need any tips or anything for clocking, let me know, i'd be glad to help. You always welcome to hop in the teamspeak in the evening for some guidance, too.

 

[PaulieG]

Nice review Dave. For many of us who started playing with these sticks months ago, we were amazed at how tight you could get the timings at 1600/1866 while running 16GB. Hell, I was running 1866 8-8-8-24 1T at 1.45v!! I would have liked to have seen some of this in the review. However, I do understand that you do have to draw the line somewhere as to how much information to add.

 

[ucanmandaa]

If I run into any trouble I will ask for help thanks . I got them from ebay so it will take at least a week for them to arrive. I am usually "out to a bar or something" almost every night but if I find the time I'll join you in the teamspeak.

 

[cadaveca]

Nice review Dave. For many of us who started playing with these sticks months ago, we were amazed at how tight you could get the timings at 1600/1866 while running 16GB. Hell, I was running 1866 8-8-8-24 1T at 1.45v!! I would have liked to have seen some of this in the review. However, I do understand that you do have to draw the line somewhere as to how much information to add.

I have a plan!

I did consider adjusting timings tighter, and did try that as well, however, they affected clock scaling in such a way that the bandwidth gains on X79 did not justify doing so. It's not a matter of how much info is included, it's about optimizing performance based and what the voltage and timing gains give.

This is not the only reason I scale the timings the way I do, but I want to save that for the next reviews, as there is some good info to be had about that, and why I do things the way I do. There is no quick and easy way to explain it, so I've chosen to add a bit into each review in such a way that every review will add on to the next. As time goes on, I want to pull all the clocking information out of the reviews and group them into one large guide that then references the reviews, but i do have to get W1zz's approval first.

as to your own clocks, I was not able to reach 1866 @ 8-8-8-24. My sticks only reached 1706 MHz:

There is at least one other user in the original thread we had going about these sticks that was unable to reach those clocks as well. I tested the DIMMS with 1.2 V to 1.75 V before deciding what to use for the clock scaling tests. I considered performance gains, votlage scaling, clocks, and everything else.

I keep track of my time; I spent 71 hours total on this review.

 

[revin]

I have a plan!

There is at least one other user in the original thread we had going about these sticks that was unable to reach those clocks as well. . I considered performance gains, votlage scaling, clocks, and everything else.

I keep track of my time; I spent 71 hours total on this review.

Err, me

Wow, how many was guiding me?!! Again respect to all who helped me I just cant get over how baddazz these sticks are.
Reminds of the killer Corsair XMP 2.2.2.5 sticks back in the day..................

 

[Feänor]

That is what i call a complete review. One with a x79 platform would not need anything more to play with this ram.

I have three 4gb unmatched sticks and while i did not test at lower speed (< 1800), i managed to get 1850 9-9-9-24 and 2100 10-10-10-27 (both 1,6v) out of them. I'm pretty sure it is my 970's imc who's the limiting factor. They really seem to shine on Sandy bridge (-e or not).

 

[mastrdrver]

What is you uncore for those speeds and kind of vvt are you using for those speeds?

Thinking about getting 3 so I ran run 12GB on 3 sticks instead of 6.

 

[LagunaX]

Been running mine 9-10-10-28 1T 1.6v 2133 for over a month now:
http://www.techpowerup.com/forums/showthread.php?t=159320
Even put some bling bling on this Honda Civic too:
http://www.techpowerup.com/forums/showthread.php?t=160687
Nice review!

 

[manofthem]

I keep track of my time; I spent 71 hours total on this review.

Bless your heart, my good man, great review indeed. Knew it would be since I've been using them as well, but it's still good to see it in official review form

 

[Steven B]

props to you mate for this review, i know you went out of your way and bought this kit.

I also know a good review when i see one, and this is one heck of a memory review

 

[Feänor]

What is you uncore for those speeds and kind of vvt are you using for those speeds?

Thinking about getting 3 so I ran run 12GB on 3 sticks instead of 6.

Uncore running at 3515 mhz and using 1,395 vtt. I do not dare to go higher 24/7.

 

[Aquinus]

Very nice review. You wouldn't have noticed how warm the memory got, did you? Considering they don't have any stock heat spreaders I assume it must be pretty low.

Very nice review, it almost makes me wish I didn't get my 2133mhz G.Skill kit.

 

[cadaveca]

props to you mate for this review, i know you went out of your way and bought this kit.

I also know a good review when i see one, and this is one heck of a memory review

Heh, I didn't buy this kit. Actually, I'd only ever consider buying something for review if the requests were overwhelming, and I didn't have other reviews to work on...I've got two other kits waiting here to finish up, and another on it's way.


But W1zz and DarkSaber are pretty good at making sure I have what I need...and this kit was sent to me from DarkSaber. I'm not sure how he got it....

Very nice review. You wouldn't have noticed how warm the memory got, did you? Considering they don't have any stock heat spreaders I assume it must be pretty low.

Very nice review, it almost makes me wish I didn't get my 2133mhz G.Skill kit.

I saw 46c while running 1.75 V, neasured with a thermal diode attached directly to the top of one of the middle memory ICs. Taht was on the Crosshair V though, that has the pins to attach the diodes to.


The G.Skill kit, to me, I feel more comfortable overvolting. The Samsung's have stock 1.35 V ICs, whic hare rated by Samsung as 1.35 V and 1.5 V compatible. They also say the 1.5 V is good to 1.575V. 1.6 V should be fine, but I do not know for sure. When I can get 2133 11-11-11-28 @ 1.35-1.4 V, I don't see much point to shoving 1.6 V to them for the minor timing change that makes very little difference in bandwidth. There's nothing wrong with it, but i feel we might see some sticks from OEMs like Corsair and G.Skill that use these IC's and higher votlages, and I do have to make sure I am making recommendations that work for less-advanced users as well. If threse sticks booted in on JEDEC profiles with 1.5 V, I'd feel a bit more comfortable, but they didn't, they used 1.35 V.

 

[Taurus_G4]

How good are these against Mushkin Redline Ridgeback 993997 16GB (4x4GB) DDR3 PC3-17000 9-11-10-28

 

[cadaveca]

How good are these against Mushkin Redline Ridgeback 993997 16GB (4x4GB) DDR3 PC3-17000 9-11-10-28

NO idea, i do not have that Mushkin kit, and haven't had a chance to play with it. It might depends on how the secondary timings are set up, which IC the Mushkin uses, etc...

What I can say is that the Samsungs are going to cost you far less for the same capacity(AFAIK, the mushkin's sell for $179.99, with which you could very nearly buy 32GB of the Samsungs).

I relyl don't have any idea as to perofrmance or anything until I am at hte ponit of making graphs..I do stabiility testinhg without looking at hte actual speeds or scores, just focus on whether the kit is stable or not. A big part of my final score is how I fell when doing the testing, and how the sticks respond to the changes i make trying to bring stability back.

Then, I typicalyl write up the main aprt of hte review, the part before the testing. Then I make the graphs, get them hosted, and then look at the results adn copare against the numbers i already have.

Technically, i thought the G.SKill kit was going to be the better of the two, but performance-wise, the Samsungs win. However, I feel more confident when clocking the G.SKills...so it depends on what you are really looking for...heck, I'll buy ram just because i like the heatsink(or lack thereof).

If Mushkin decides to send me a sample, I'll let you know what I think then.

 

[Taurus_G4]

You're right.. I really hope they do that, they don't have a single benchmark for those modules.

 

[cadaveca]

You're right.. I really hope they do that, they don't have a single benchmark for those modules.

First, let me apologize for my atrocious spelling in the post above. Geez.


As to your comment, I kinda find that surprising, you know? I am using a Mushkin set in my reviews, however, this is a kit that I myself purchased, and once I have enough results, those results will be removed from my future reviews, too.

Not that I have any bias against Mushkin..in fact, i really like the kit that is used in my current reviews, however, it's a cobbled-together kit, not a real quad kit. I am open to reviewing products from any company.

 

[Wrigleyvillain]

Now $47 at MC which will be like $50 after tax. Still, hard to complain. I'll pick up a kit or two if I can get my ass over there.

 

[LagunaX]

The best was $39.99 at CompUsa/TigerDirect free shipping...

 

[LAN_deRf_HA]

Does running such a low ram voltage improve cpu overclocking and thermals on 1155/2011?

 

[cadaveca]

Does running such a low ram voltage improve cpu overclocking and thermals on 1155/2011?

Alittle bit, yes, but it also depnds on case airflow, IMHO. The memory itself does consume less pwoer, and run a lower voltage, so it doesn't get as hot, and neither does the memory VRM. Crank up the voltage with a huge OC, and they put out a decent amount of heat, but nothing overwhelming.

 

[LAN_deRf_HA]

I mean isn't it less strain on the cpu. Not from the memory putting out less heat but from less voltage going through the IMC. Doesn't the ram voltage tie directly into the procs now? Isn't that why they recommend not using crazy ram volts on 1155/2011?

I know when you're at the extreme end on 1155 you can end up having to choose between a higher memory clock or higher cpu clock since it's all tied together. If these are less stressful on the IMC in theory it could enable a higher clock on one or the other right?

 

[cadaveca]

I mean isn't it less strain on the cpu. Not from the memory putting out less heat but from less voltage going through the IMC. Doesn't the ram voltage tie directly into the procs now? Isn't that why they recommend not using crazy ram volts on 1155/2011?

I know when you're at the extreme end on 1155 you can end up having to choose between a higher memory clock or higher cpu clock since it's all tied together. If these are less stressful on the IMC in theory it could enable a higher clock on one or the other right?

It's not really the votlage that dictates the IMC load. vDIMM is reduced before hitting the IMC. I haven't run into a situation where vDIMM affects CPU clocking ability on LGA2011, so I haven't had the oppotunity to investigate such. Not one of my CPUs, even 1155, is affected by ram clocks when CPU clocking.

The CPU itself and the memory speed are linked due to how the L3 works, as the L3 is last-level cache before ram, and runs at the CPU's speed(because of ringbus design). The higher the CPU speed, the faster the L3, the more load is placed on the IMC. It's NOT about the IMC on the DIMM side at all.

Ram is a passive device, and as such, doesn't really have much effect on CPU unless it's not capable of feeding the CPU the data it needs. Any issues with clocknig related to ram speed are 100% the fault of the IMC itself, and not the DIMMs.

Now, if a DIMM requires less CURRENT, that might apply, but not that i have seen as of yet.

 

[Millennium]

Much more discussion and results in this thread by the way. I'm really happy with my (non LP) memory...