larrymoencurly's Discussion of Greater than DDR3-1600 RAM Speeds

[larrymoencurly]

Not always. My Crucial Ballistix LV LP kit has much tighter timings in XMP than SPD.

Tighter means less strict, or less margin for overclocking,

How is "less leeway" "less strict"?
Strict is, IMO, a poor word choice. Timings are all very strict, the ram is rated at THAT timing and none other.
In my opinion, a strict timing is one that cannot be violated, raising a timing won't violate anything, tightening it violates it.

Right, and XMP profiles are tighter/have less leeway only because less strict quality standards apply to them.

 

[Bobnova]

You are making no sense to me, at all.

Can you violate the RAM's timing and tighten it? If so, it wasn't very strict.
Can you tighten a timing at XMP clocks and timings? Rarely.
XMP timings are generally very strict, especially on higher end stuff where XMP actually makes a difference.

strict
/strikt/
Adjective

1. Demanding that rules concerning behavior are obeyed and observed.
2. (of a rule or discipline) Demanding total obedience or observance; rigidly enforced.

XMP profile by definition demand the rules, their rules, are observed.


How is giving something more leeway being stricter?
How is giving something less leeway being less strict?

 

[Knufire]

Tighter means less strict, or less margin for overclocking, and I've never seen any module or chip manufacturer mention the tightness of its timings, except when Intel said its 64Kb or 256Kb 120ns DRAM averaged 40ns cells.

I think you misunderstood, tighter timings is slang for saying lower timings.

 

[EarthDog]

I think you misunderstood, tighter timings is slang for saying lower timings.

Exactly...

Though perhaps he is saying that b/c there are tighter timings you can loosen them more for potentially more o/c headroom? IDK... so lost.

 

[larrymoencurly]

You are making no sense to me, at all.

Can you violate the RAM's timing and tighten it? If so, it wasn't very strict.
Can you tighten a timing at XMP clocks and timings? Rarely.
XMP timings are generally very strict, especially on higher end stuff where XMP actually makes a difference.

XMP profile by definition demand the rules, their rules, are observed.

How is giving something more leeway being stricter?
How is giving something less leeway being less strict?

Say you make tires, but because you have stricter standards than your competitors, your S-rated tires (112 MPH max speed) are actually made to Y-rated standards (186 MPH max), giving drivers more leeway to safely exceed the advertised S rating. IOW your S rating isn't as tight as your competitors'.

Regarding memory, is it more likely to fail when run according to the XMP profile or the chip manufacturer's JEDEC profile? Wouldn't a higher failure rate indicate more lax, less strict testing standards?

 

[EarthDog]

I wouldnt expect either to fail since its rated for the XMP in the first place...I wouldnt expect anything in JEDEC or XMP to fail more BECAUSE ITS RATED THAT WAY to begin with.

 

[Bobnova]

I wouldn't expect it to fail either way. If it does, it's defective.

Are you trying to say that you meant that the testing for XMP was less strict rather than trying to say the timings were less strict?
I could have sworn you were talking about the timings, not the testing.
Two very, very, very different things there.

You probably do have to enable XMP because your memory is actually 1600 or slower (Patriot 2133 was proved to be 1600 by removing the heatsink), and XMP enables less strict parameters.

There we go, timings, not testing.

As a note, memory defaults to 1333 or 1600 because those are the highest valid JEDEC profiles, not because the chips are built for that spec.

 

[larrymoencurly]

As a note, memory defaults to 1333 or 1600 because those are the highest valid JEDEC profiles, not because the chips are built for that spec.

Micron rates some of its chips just JEDEC 1333 or 1600 because it thinks they're built for those specs and not anything faster and more profitable.

 

[Bobnova]

Micron rates some of its chips way higher, too. That argument is not valid.

Also, you quoted yourself saying what I said.

 

[larrymoencurly]

As a note, memory defaults to 1333 or 1600 because those are the highest valid JEDEC profiles, not because the chips are built for that spec.

Micron rates some of its chips just JEDEC 1333 or 1600 because it thinks they're built for those specs and not anything faster and more profitable.

Micron rates some of its chips way higher, too. That argument is not valid.

Way higher than the JEDEC specs it assigned to the chips? Maybe only for its heatsinked DDR3.

 

[Bobnova]

Nope, no heatsinks needed or specified.
I think you're confusing the IC's rating with the JEDEC specs. They're very different things.
Also, how do you know why Micron does something? Work there? Know someone who works there?

 

[larrymoencurly]

Nope, no heatsinks needed or specified.
I think you're confusing the IC's rating with the JEDEC specs. They're very different things.

Don't the IC's ratings describe what the company makes while JEDEC is for standardization?

I can't tell if Micron's heatsinked memory uses overclocked chips because the only ones I saw didn't seem to have the full part numbers printed on the chips. OTOH I doubt that Micron rates some of its heatsinked modules at 1.65V because they work as claimed at 1.5V.

 

[Bobnova]

larrymoencurly's Discussion of Greater than DDR3-1600 RAM Speeds

Not always. My Crucial Ballistix LV LP kit has much tighter timings in XMP than SPD. Tighter means less strict, or less margin for overclocking, How is "less leeway" "less strict"? Strict is, IMO, a poor word choice. Timings are all very strict, the ram is rated at THAT timing and none other. In...

www.overclockers.com

JEDEC timings are a set of timings, including frequency, that a RAM chip can say "YES, I DO EXACTLY THIS", and a motherboard can say "LOL ME TOO", and a memory controller can say "HELL YES I CAN DO THAT TOO!" and then they all work happily, guaranteed. Regardless of platform. Regardless of RAM. Any combination.
It has absolutely nothing to do with how fast the chip can go, or how fast it's built to go, it's simply a known-good profile that everything can do.
There are RAM chips rated for speeds that AMD PhII IMCs cannot even come close to touching, but the RAM is designed for and rated for. (like, anything >1600-1866MHz)

To call itself DDR3, memory MUST be capable of running at least one JEDEC spec frequency/timing set at 1.5v. That includes LV memory, if it's not 1.5v tolerant it cannot call itself DDR3.
That include ICs built for 1.8 to 1.9v (D9GTR), they MUST be capable of running a JEDEC profile at 1.5v to call themselves DDR3.
It can be built for more voltage, it can be built for less voltage, it has to be able to run 1.5v.

It has absolutely nothing to do with that the chip is actually built to do.
A car related example is the Skyline GTR motor.
Japanese "regulations" limit it to 280HP, that's the JEDEC spec.
A chip, nothing but a chip, and you have >500HP. That's what it is actually built to do.

On a vaguely related note, if Micron designs the chip, Micron makes the chip, Micron rates the chip, Micron solders the chip to a RAM PCB, Micron tests the stick, Micron rates the stick, Micron certifies the stick, and then Micron sells the stick, how can it be "overclocked"?
(Micron owns Crucial. Crucial is, essentially, Micron's storefront)

 

[larrymoencurly]

On a vaguely related note, if Micron designs the chip, Micron makes the chip, Micron rates the chip, Micron solders the chip to a RAM PCB, Micron tests the stick, Micron rates the stick, Micron certifies the stick, and then Micron sells the stick, how can it be "overclocked"?
(Micron owns Crucial. Crucial is, essentially, Micron's storefront)

I suspect that Micron's Crucial division has lower quality standards because why else would they offer DDR3 in 1.65V versions while Micron's commercial division doesn't list anything above 1.5V? And what about everybody's favorite, memory from their Spectek division? Just as good!

 

[Bobnova]

Way to ignore the (myriad of) points you don't have anything for.
So now the straw man is the "quality standards"?

 

[larrymoencurly]

Way to ignore the (myriad of) points you don't have anything for.
So now the straw man is the "quality standards"?

I'm not familiar with the chips you mentioned, and you didn't provide any proof about them (that meets the standards you normally request of others), so I didn't answer. But if you insist, I speculate it's a case of new chips being underrated because the manufacturer isn't sure of them and is seeing a lot of variation in test results, something that happened when the PC100 standard came out and many PC66 chips were able to meet it.

You're overusing the word "strawman" and are naive if you believe retail marketers like Crucial and Corsair have standards as high as Hynix or Samsung.

 

[wingman99]

All that maters is the memory meets JEDEC or X.M.P standards, use to be they should just run there advertized specifications.

http://en.wikipedia.org/wiki/JEDEC_memory_standards

JEDEC memory standards are the specifications for semiconductor memory circuits and similar storage devices promulgated by the JEDEC Solid State Technology Association, a semiconductor trade and engineering standardization organization.

http://www.intel.com/content/www/us/en/gaming/gaming-computers/intel-extreme-memory-profile-xmp.html

X.M.P Intel Extreme Memory
Profile by Processor

Intel has developed a certification program for memory vendors to test their products for compatibility against Intel XMP. To view compatible memory DIMMs from a variety of vendors, select your system’s processor:

 

[EarthDog]

I'm not familiar with the chips you mentioned, and you didn't provide any proof about them (that meets the standards you normally request of others), so I didn't answer. But if you insist, I speculate it's a case of new chips being underrated because the manufacturer isn't sure of them and is seeing a lot of variation in test results, something that happened when the PC100 standard came out and many PC66 chips were able to meet it.

You're overusing the word "strawman" and are naive if you believe retail marketers like Crucial and Corsair have standards as high as Hynix or Samsung.

Why would 99% of people care again so long as the sticks run what they say they will run on the package?

 

[Bobnova]

I'm not familiar with the chips you mentioned, and you didn't provide any proof about them (that meets the standards you normally request of others), so I didn't answer. But if you insist, I speculate it's a case of new chips being underrated because the manufacturer isn't sure of them and is seeing a lot of variation in test results, something that happened when the PC100 standard came out and many PC66 chips were able to meet it.

You're overusing the word "strawman" and are naive if you believe retail marketers like Crucial and Corsair have standards as high as Hynix or Samsung.

Personal accusations will get you nowhere good.

Hynix and Samsung can only rate the IC itself, that's only a part of the story when it comes to rating a stick. They also rate at 1.5v, not 1.65v. Much like CPUs, when you raise the voltage within the manufacturer allowed tolerances you get higher, stable, clocks. Samsung and hynix don't necessarily test at higher frequencies, either.
They're testing different things.


Finding datasheets for Hypers has turned out to be essentially impossible, even Elpida's own site lists the datasheet as TBD.
BBSE chips however, which as moocow stated all do >2000 with ease, is findable:
http://www.elpida.com/eolpdfs/E1375E50_EOL.pdf
Maximum rating? 1600MHz.

Finding datasheets for DDR2 D9s turned out to be similarly difficult. Don't try to argue they don't do crazy high clocks that even the sticks aren't rated for though...

(My personal best. D9 DDR2 result. The teams best is 741MHz, the all time record is 900MHz. All of these are D9s rated for 266-400MHz)

This is all common, constantly tested and found to be true, wisdom in the benching section.


The bottom line remains, if a stick is rated for X MHz and runs at X MHz correctly, it doesn't matter in the slightest what the ICs rating is. Also that IC rating and overclocking results do not correlate well.

 

[larrymoencurly]

Why would 99% of people care again so long as the sticks run what they say they will run on the package?

IOW memory modules are just like power supplies -- the advertised claims are always true, and all the products are good.