[HowTo] SO-DIMMs Extreme: Boost 1600 to 2133 and beyond / JEDEC & XMP

Can you please upload your .thp file after you've modded it?
And could you take a guess at what would happen if I flashed it onto my Samsung ram (1333mhz)?

 

Nice, being able to do the job that the manufacturer should have done in the first place!

I can only highly recommend to get the comercial version of Thaiphoon Burner...it made CL9 RAM out of my CL11!

 

Heah but what if u will mess-up RAMs by doing this?

 

As long as you flash one stick at a time, there's not much of a risk. I've flashed OCZ, Corsair, Kingston, and Crucial RAM plenty of times. At worst, I've had to boot up with the functional RAM stick, and insert the badly-flashed stick after POST to re-flash it once I get into windows.

I made my current OCZ RAM XMP-enabled ages ago, just to see if it could be done. I'll do the same with my Value RAM later today.

 

You should of course try to stay within reasonable limits. Should the system refuse to boot you just add a RAM that works and it forces the other one to run at regular timings (bios usualy forces both sticks to run at the speed of the slowest)
In case you badly mess-up you could boot with a working stick and hot-plug the one that you need to re-program after boot and re-flash it.


@TURBODUDE


I was wondering,
since one CL has a nano second range like let`s say CL9 ranges from 69 to 76 (MTB value), do those really run at different nano seconds or does the system bios interfere? So that if it detects a value that it considers to be CL9 it will run it at certain ns speed.

Because i have gone to the lowest CL9 value (69) before it is CL8 and the system runs rock stable, but if i take it one notch down to CL8 (68) it already won`t boot at all.

So in how far does a change of the MTB value really have an effect within a CL class?

THX

 

Here you go
http://softnology.biz/work/CMSX8GX3M2A1866C10.zip

Prema

First of all, MTB is not a unit for measuring timing delays. It just helps to encode a floating point number to decimal data format to be stored in 8-bit SPD registers. For example, if the minimum CL time (tAAmin) is 8.625 ns and MTB timebase is 0.125 ns then tAAmin is 69 MTB units and can be easily stored in a SPD register as a data byte (0-255). Secondly, tAAmin time should not be out of its own possible range. For example, if CL9 is in the range of 76 to 69 MTB, you can set tAAmin to the lowest 69 MTB and get the highest possible performance. But if you set tAAmin to 68 MTB (8.500 ns) you violate the boundary, because 8 clocks for CL is too low delay and is not supported by SDRAMs .

 

THX & REP +1

One more question:

What defines the boundry of possible range then?

 

hi i ran it and what should i do now?
thnx

 

I updated my first post again. Hope no one is against it.

 

Question:

By flashing the ram timings does Memtest86+ detect the timings when running stability testing?

Looking to get 8-8-8-24 @ 1600Mhz with Taiphoon Burner.

 

Call Timing Table Editor, set focus on the "Min CAS Latency Time" listbox and scroll up/down with mouse wheel until the DefCAS field changes to another CAS. And you'll get the possible range.

 

Sorry looks like i ask the qeustion the wrong way.
What i do not understand is that the RAM runs ROOCK STABLE at 8.656 ns (CL9/69) with T1 command rate but won`t boot at all at 8.531 ns (CL8/68) with T2 command rate and even witch all other parameters much looser. Thats just a difference of 0.125ns...
Same with the tRCD it is stable at 9,781ns (10T/78) (as 9-10-9-27), but won`t boot at 9.656ns (still 10T/77) even if i raise the whole thing to 11-10-11-34?
It feels more like running into some artificial wall than really reaching the technical limits.
Thanks for your patience!

 

Prema, I assume that it won't run at CL8 because it means 8 tacts for 8.531 ns. I don't know this for sure but think that it means 1.066 ns each tact. But when you choose CL9 8.656 it will work at 0.962 ns each tacts. Your RAM needs at least 9 CL tacts on certain speed. while chips are so ROCK SOLID that they may work easily (perhaps) on the fastest possible speed of 1 tact which is 0.962 ns.

I just think hat I'm right but I may be fully wrong.

 

That indeed makes a lot of sense for the CAS!
Still wondering about the tRCD which remains within 10T and still refuses to boot no matter how slow i set the rest...

 

OH MY GOD, 18 pages about one product and not one real world benchmark is contained in here.

I would be one of the people telling you that unless your using a integrated GPU there is no reason to have faster ram than what comes stock with your laptop and rather than say that since HT already tried to, I wanted to see the benchmarks but nobody has posted any...

Wonder why?

 

Yeah it started be boring for me too som module modding for 3% cant be arsed..
And benchmarks? I think i have posted som ( depends which u wanna see) winsat shows 24620 MB/s same as the kingston no diff. at all

 

I think what Vicious meant was real world usage benchmarks, for example, running games and checking min/max and average fps. I've seen some claims in Asus gaming that faster RAM helped keeping the minimum fps a little higher, if that is truly the case, then the RAM may be worth it to some.

 

Yes this. RAM benchmarks don't tell us much. When I became interested in SSD's all I saw were SSD benchmarks and nothing more. Once you get to real world usage you would see for your average consumer the "slow" SSD's performed just as well if not better than the "fast" SSD's.

Please offer some gaming benchmarks before/after comparison would really tell us something. Even a run of 3DMark Vantage and 11 at 1333 and 1866 if nothing else. I'll even provide links to reasonable gaming benchmarks if desired.

 

HyperX 1600Mhz PNP





I'm running



Will do a few vantage runs later but need to wait till it's minus outside cause I'll be pushing the cpu for 30k

But I want Vengeance 1866!!!

 

James D,

Here is the link to the datasheet on Micron D9PFJ chips:
MT41K256M8DA-125 - Micron Technology, Inc.

Perhaps, you'll find it useful for your first post.

 

PC Mark 7 results

 

If you could tell me why would it be useful maybe? I may be not so well skilled on "Hard" as you could expect.

There is no point to follow their orders for other benchmarks than you already posted. Faster RAM is faster RAM. Quicker RAM is quicker RAM. When sysem becomes snappier you can't bench it properly but you can see that programs which are cached to RAM open quicker than they were. faster response including on Alt+Tab and other. System is more stable when you watch HD movie and do other stuff at the same time.
Also there is no point to tell about benefit for integrated GPU. Everyone who OC his GPU knows that the biggest benefit of OC is increasing memory clock (at least for DDR3 and GDDR3).

I feel that this thread is "How to" type instead of useless "Why to".

 

PCMark 7 is good. Although many of us would still like to see before and after 3DMark Vantage / 11 / Cinebench GPU or other game type benchmarks since it was stated that it would benefit the IGP significantly. I'd like to see 1333 vs 1866 personally to see if there is a measurable difference.

 

That's what I told. Except just making dozen of different benchmarks (when many of them won't even show any difference) someone needs also to do them before and after upgrade of RAM which is at least 1 hour or even more.

1 thing is asking someone for long-time benchmarks. Another is for specifying what exactly you want from the beginning to decrease or even eliminate flood.

6 long post already have been written while could be just 1 short like :"GalaxySII (for example), please post XX, XX, XX benchmarks with new and old RAM if you can please" withoutwords like this "OH MY GOD, 18 pages about one product and not one..."

P.S. Yes. I kinda worry about every thread created by me.

 

Well, the most interesting thing in the datasheet is that D9PFJs are 800MHz chips with a low voltage supply of 1.35V. So, Corsair just overclocked them to 933MHz after increasing the voltage to 1.50V.

 

I was asking, not "telling orders", don't know why you're being so defensive. I did not say anything like "OH MY GOD, there's been 18 posts and nothing..." And I was NOT only asking YOU, I was asking any other users that managed to get the 1866MHz. If nobody wants to do it, then that's their prerogative, it's their laptops, their RAM, just seeing if anyone is willing to validate the claims of 1333 vs 1866 in gaming. All the information provided here is great from a technical standpoint and much appreciated. Just was asking for some real world confirmation.

Running 3DMark Vantage takes 5 minutes. Swapping RAM takes 1 minute. You invest at most 15 minutes in checking the different RAM. Nemix77 made claims of it being as fast as 6620G. It would be good to know for other users' sake so they can make informed decisions.

Here's one article with 1600 vs 1866: http://www.neoseeker.com/Articles/Hardware/Reviews/gskill_ripjawsz_ddr3_1866_16gb/8.html

With dedicated card no difference: http://www.anandtech.com/show/4503/sandy-bridge-memory-scaling-choosing-the-best-ddr3/6

HD 3000 two games result improvement 1333 vs 1866: http://www.legitreviews.com/article/1599/8/

General comparisons of IGPs: http://www.hardwarecanucks.com/foru...a8-3850-apu-review-llano-hits-desktop-21.html

 

Yes, I also noiced his and also that default CL timings are 11. Looks like this modules are just the best for Overclocking... at least Corsair think so.
If you want I will add it.

4) And me too were not talking about only you
3) - here
2) - here my P.S. and I just want this thread being much more informative per post.
1) - I don't know word weaker than orders but stronger than just asking. Before you say smth reread 4) and 3) again
5) It takes hour. TRUST ME. You have to bench all of these, then exchange RAM, then again bench. Then again exchange RAM. Also create all screenshots and make a post. And also planning all these steps. It will take at least 1 hour unless you are a professional benchmaker.

 

Its pretty obvious that having faster ram will give better ram benchmarks. We knew that before the thread was even created.

What we want to see is proof of the claim that faster/more expensive ram actually gives higher game performance in the value of higher frames per second to any significant degree.

It also needs to be done with a benchmark that is pre packaged and static so you can run the exact same test more than once on the different hardware or different speeds of ram.

Unless you (anybody with the hardware) can give some kind of evidence like this than none of the clams contained in this thread have any bearing or support.

I must have done over a thousand game benchmarks by now through my ages of reviewing products and testing things, its not that hard and its always the first thing I do if I need to put performance into perspective for my readers.

I am pretty sure that if I was part of this thread I would have been happy to RUSH those benchmarks to prove my point and to justify my purchase. The fact I have seen no posts like this has me.... curious.

 

Thanks for the datasheet, they're 30nm 1.35v chips exactly as I guessed.

@ ViciousXUSMC

I think the whole point to the thread was to confirm the timings on the Vengeance are indeed better than those found on HyperX's.

Which makes the Vengeance not only faster but also cheaper then HyperX DDR3-1600/1866's.

There are users on this forum who spend a lot more for slower HyperX DDR3-1600/1866 memory, then again the HyperX's have ben available since second half 2011.

 

So what is the general consensus? I'm eyeing the 1866Mhz CL10 ones and I'm wondering if I would get the Cas down any lower in my laptop bios? I'm using the M17x R3. Looking to perhaps creating a full Corsair system. I got a Corsair SSD too.

I've been going through tonnes of ram lately and got so much spare if anyone wants. I've got some very rare 2GB sticks: HyperX 1600 CL7 and 1866 CL9's.

 

The HyperX 1866 are cl11 and run cl9, so there is a good chance you may reach cl8 with the corsair.

 

man of course u got plenty tests and comparing done good for you but u obviously are paid for that and we end users we are not.. if u doing this as your job of course u got to do it for your readers..
I cant waste time of doing stuff which i wont be paid for that simple

 

It's not his job and he doesn't get paid for it... usually. Nor do I, although I'm about to publish comparison of 1333 vs 1600 vs single channel vs dual channel RAM for both Llano and Brazos.

 

Found that it is possible to flash 1600 Corsair RAM by 1866`s profile and it will work. At least on low 1066 speed Also i found that it has Corsair`s memory chips instead of micron which has higher rated sticks. I think this is why price is twice bigger.
Will update post later.

 

What do you mean on low 1066 speed?

 

it was laptop with only 1066Mhz memory controller nearby.

 

I received today a Vengeance 8gb 1600 kit, installed it on my Macbook Pro and it feels a bit quicker than the HyperX I replaced. Geekbench went from 6900 to 7200.

I guess the 9-9-9-24 vs 9-9-9-27 makes a difference, apart from the fact that the HyperX would run @ 1333Mhx and the Vengeance ones run @ 1600, go figure...

I´ll try Cinebech to see how it fares.

 

Corsair just can't run 1333 without flashing them. But why HyperX didn't run 1600 while Corsair can? Hmmm.
Anyway. What Cpu do you have?

 

It´s in my sig, i7-2620m.

 

I see placebo effect...

 

If you are talking about me then I guess you are wrong.

Kingston HyperX ran @ 1333 whereas the Vengeance run @ 1600.
Geekbench went from 6900 to 7200.
Cinebench score also improved.

That´s not placebo, that´s proof, not lightyears ahead in performance but proof nonetheless...

 

Memory upgrades comonly show a large synthetic difference (30%) but a resultant performance increase of less than 1%.

 

Here we are again
I have already told what are performance changes thanks to faster RAM:
snappier (what is cached in RAM) like opening apps or Alt+Tab and other.

 

That's what I've been saying. Real world application and games benchmark differences would be helpful.

I understand "snappiness" which is something that is hard to explain to someone about an SSD, but there's also measurable differences that people can relate to as well. While I was doing my testing on the AMD 6620G (see sig) with different RAM, I really didn't notice a difference in system performance with 1066 CAS 7 vs 1600 CAS 9. But for GPU performance it was noticeably faster and measurable. It would be nice to see even a few of those benches I did with 1333 vs 1866 with an HD 3000.

 

The bandwidth of these modules is more than enough for that and even if you look at latencies there is not much overall difference.

 

enough is what? How do you measure it? Enough for who? for you? have you ever thought that other people don't agree with you? Don't buy if you don't like and don't argue with that one who already bought because you are no one to people who already bought it unless you buy it.
That' like a monkey in a zoo telling that temperature of water she is swimming is warm enough so it don't understand meaning of sauna.

But don't think that if you have already bought this RAM your words became truth. It is true for you and your feelings. But not for other.
I am fine with 20 fps in game on high resolution while other say that 30 fps is very low so they need to lower effects to get "nice" up to 50 fps.

 

The point is that it's hard for users to make a decision if it is worth the extra money to them unless there are some real world benchmarks that validate its performance other than taking someone's word that it's "snappier".

 

Question:

Does Thaiphoon Burner firmware flash permanently keeps the SPD timings on the ram like flashing overclocked BIOS for video cards?

 

Until you reflash it again.

I changed settings little bit on my sodimm ddr 333 kingston 1gb module to the same as on my crucial 1gb pc3200. Same timings but slightly lower (nanosec) latency.

 

You're exaggerating the difference it would make. In almost any situation, the gain you'll get from going with high speed memory is a few percent at best, something which is nigh impossible to notice, let alone appreciate. If you can "feel" a difference between memory modules, then unless you're one of an increadibly rare few who actually use programs which could make use of the bandwidth, then what you're feeling is a mere placebo.

If you have integrated graphics, then I can entirely understand going with faster memory, but beyond that it's pretty much pointless.

As for benchmarks, of course they will show a difference, but you have to understand that the majority of benchmarks simply generate an arbitrary number which puts weight on some componants while valuing others less, essentially making it pointless. Benchmarks like PCmark and the like are essentially designed so you can show off your hardware, not as an actual judgement of performance. If you want to compare performance then you need something which gives a linear score (automatic fail for PCmark).

Do some FFT tests, matrix multiplication, SuperPi calculations etc. then post back once you've obtained some viable results. If there's more than a few percent difference between the two (if any at all) then I'll take all the above back.