RAM brands

[jerryshardware]

Hello,
What do you think are good brands for RAM?
I am sure there are good and better brands.

Please reply, thank you.

 

[EarthDog]

Not really... ram is ram (take a look around this section) for the most part.

That said, popular brands are Corsair, GSkill, Mushkin, Kingston...

 

[I.M.O.G.]

Agreed, ram is ram.

Consumers look at brands for RAM, because they buy based on reputation mostly.

But the super-interested in RAM enthusiasts don't look for certain brands, they look for certain IC's. PSC, BBSE, Micron, etc... Depending on what their interest is. Like super tight timings, or high frequency, or both. The manufacturer makes little difference, except certain brands/models sometimes become reputable for having certain IC's onboard, then that specific model is sought after by some.

In my opinion anyone not benchmarking should buy whatever ram is cheapest, in the quantity they desire, that runs at the frequency they want. The brand means very little, aside from support/RMA ability.

 

[larrymoencurly]

Any brand of module fine if its chips aren't overclocked and you can easilyidentify the chips as Micron/Elpida, Samsung, ProMOS, PowerChip, Nanya/Onitera, or Hynix. Otherwise test like crazy with both MemTest86/86+ and Gold Memory, in both AMD and Intel motherboards.

RAM is not RAM. Some is made from chips that passed testing in a million dollar machine, others made from chips that were sold as rejects or as whole wafers that weren't completely tested by their manufacturers and then tested only with PCs in a RAM testing sweatshop.

 

[I.M.O.G.]

RAM is not RAM. Some is made from chips that passed testing in a million dollar machine, others made from chips that were sold as rejects or as whole wafers that weren't completely tested by their manufacturers and then tested only with PCs in a RAM testing sweatshop.

Any detail on this million dollar ram tester? We have an RST from Ultra-X for review use... It costs a couple grand, and it is the same device used by some RAM companies. I haven't heard of any million dollar ram testers, but its something I'd be interested in reading about.

All ram is batch checked tho... To ensure its functional and assign it a rating. No brand checks every stick, let alone every module.

Performance wise - ram is ram. You aren't going to feel a difference between budget ram and premium ram on a usual desktop computer. Its just the way it is. You can run a synthetic test and find noticeable advantages, but you just aren't going to realize a noticeable fps increase or desktop feel from better ram.

 

[EarthDog]

Any detail on this million dollar ram tester? We have an RST from Ultra-X for review use... It costs a couple grand, and it is the same device used by some RAM companies. I haven't heard of any million dollar ram testers, but its something I'd be interested in reading about.

All ram is batch checked tho... To ensure its functional and assign it a rating. No brand checks every stick, let alone every module.

Performance wise - ram is ram. You aren't going to feel a difference between budget ram and premium ram on a usual desktop computer. Its just the way it is. You can run a synthetic test and find noticeable advantages, but you just aren't going to realize a noticeable fps increase or desktop feel from better ram.

QFT

 

[larrymoencurly]

RAM is not RAM. Some is made from chips that passed testing in a million dollar machine, others made from chips that were sold as rejects or as whole wafers that weren't completely tested by their manufacturers and then tested only with PCs in a RAM testing sweatshop.

Any detail on this million dollar ram tester? We have an RST from Ultra-X for review use... It costs a couple grand, and it is the same device used by some RAM companies. I haven't heard of any million dollar ram testers, but its something I'd be interested in reading about.

Advantest, probably the only remaining maker of RAM test equipment used by chip companies. I saw one of their used machines on Ebay for $500K.

Kingmax said in 2004 that it bought 3 Advantest machines, for $4.5M each.
http://www.xbitlabs.com/news/memory/display/20040521145326.html. The article, from a Russian website, doesn't say whether that's $4.5M US or $4.5M New Taiwan, but the latter would make it about $150K US.

If RAM is RAM, why do some companies bother testing with $$$ machines (even charging OEM customers extra for it) and in high temperature chambers while others use just PCs at room temperature, even allowing 2 errors per module? Then there's the matter of overclocking, and I don't mean by the user but by the module makers when they write aggressive timings to the SPDs, sometimes 60% faster than official chip specs.

 

[EarthDog]

Ram is ram for 99% of users. What is in the sticker is what it runs at. Doesn't matter if its Kingston, crucial, gskill, or hynix, psc, or samsung ic's. If its rated at ddr3 1600 cl9, that is what the user is going to get regardless of what machine it was tested in.

now you talk overclocking that is a completely different story, although in that case I would tell the op to get a higher speed kit suitable for his goals instead of searching for highly overclockable ic's... not that overclocking memory is even worth it outside of benchmarks.

 

[Mandrake4565]

Sort of on this subject, if ram is ram. When a motherboard co says that "brand X" ram works in their motherboard. Would that mean, any ram with the same specs as "brand X", will work with that motherboard?

 

[EarthDog]

Yes. If its on their list it means they tested it. It does not mean that if it is not on the list it wont work.

 

[Woomack]

Sort of on this subject, if ram is ram. When a motherboard co says that "brand X" ram works in their motherboard. Would that mean, any ram with the same specs as "brand X", will work with that motherboard?

It means that RAM on their list is compatible with boards on the list just because they were testing it. Even though you have 2 sticks on the same chips, they can have different SPD but most lower rated kits have almost the same SPD and in 99% cases is compatible with JEDEC specification that is the same for almost all RAM.
XMP = settings tested by RAM manufacturer so in this case it's like "brand X" tested memory on "brand X IC" using their testing mobos to achieve stable settings that you have as XMP profile.

Btw. memory testing on expensive machines has nothing to do with memory binning by manufacturers. Chips are produced within some specification but producers like G.Skill, Kingston, Corsair etc are making selection.
Chips that we get to our home PCs are within standard specification so are designed to work from 0 - ~80*C. Industrial grade IC is designed to work from -40*C up to ~120*C ( or something near depends from IC ).
All CPUs or RAM have some single errors in calculations. That's why you have ECC memory ( additional chips on RAM and ECC in CPU cache ).

larrymoencurly, you back with your conspiracy theory every some months but really we should be happy that we can overclock memory and it's not locked with really loose timings like 1600 11-11-11 ...
All DDR3 memory could be in 1333 9-9-9 or 1600 11-11-11 but memory manufacturers are testing it for higher speeds so you can pick if you want 1600 kit or 2600 even though IC is often similar. Look at Hynix BFR/CFR. Almost the same series for 1600 and 2600 kits and often designed for the same speed by Hynix so 1600 11-11-11. Price that you pay is for their selection, better package, sometimes better PCB, heatspreader and additional "hand" testing.

 

[larrymoencurly]

Ram is ram for 99% of users. What is in the sticker is what it runs at. Doesn't matter if its Kingston, crucial, gskill, or hynix, psc, or samsung ic's. If its rated at ddr3 1600 cl9, that is what the user is going to get regardless of what machine it was tested in.

So major brand, lifetime-warranted memory rated for 1600 MHz and made from 1333 MHz chips (common) will be just as reliable and compatible as memory rated for 1600 MHz but made from 1600 MHz chips, assuming no overclocking?

Btw. memory testing on expensive machines has nothing to do with memory binning by manufacturers. Chips are produced within some specification but producers like G.Skill, Kingston, Corsair etc are making selection.
Chips that we get to our home PCs are within standard specification so are designed to work from 0 - ~80*C. Industrial grade IC is designed to work from -40*C up to ~120*C ( or something near depends from IC ).
All CPUs or RAM have some single errors in calculations. That's why you have ECC memory ( additional chips on RAM and ECC in CPU cache ).

larrymoencurly, you back with your conspiracy theory every some months but really we should be happy that we can overclock memory and it's not locked with really loose timings like 1600 11-11-11 ...
All DDR3 memory could be in 1333 9-9-9 or 1600 11-11-11 but memory manufacturers are testing it for higher speeds so you can pick if you want 1600 kit or 2600 even though IC is often similar. Look at Hynix BFR/CFR. Almost the same series for 1600 and 2600 kits and often designed for the same speed by Hynix so 1600 11-11-11. Price that you pay is for their selection, better package, sometimes better PCB, heatspreader and additional "hand" testing.

I don't have a conspiracy theory. I'm just saying chip manufacturers do their binning to stricter standards than module manufacturers do, and that's probably why plain Jane no-heatsink modules made from branded chips tend to be more overclockable and a lot more reliable than the finned monsters. Still I'd like to see a legitimate review comparing the two types of modules for overclocking, and by "legitimate" I mean a review not based on hand picked modules on loan from the manufacturers.

Kingston and Corsair told me their DDR2 and DDR3 memory is rated to work up to 70C, not 80C, and a Corsair factory tour article showed testing being done at 58C. Those chips are usually rated by their manufacturers for 85C (same as Industrial), or 95C if the refresh rate is doubled.

 

[Knufire]

If they sell it to you, and it doesn't run at the speed/timings that the sticker says, then it's defective and you return it until you get a set that does. For non-benchers, overclocking memory is largely pointless and what IC is on the RAM isn't really consequential to anything.

 

[EarthDog]

So major brand, lifetime-warranted memory rated for 1600 MHz and made from 1333 MHz chips (common) will be just as reliable and compatible as memory rated for 1600 MHz but made from 1600 MHz chips, assuming no overclocking?

Yes. Its rated to run at that speed so it should run at that speed, and it does, correct.

larrymoencurly, you back with your conspiracy theory every some months

+1. Getting a bit tired of seeing this every so often. I have records less scratched than this.

 

[larrymoencurly]

So major brand, lifetime-warranted memory rated for 1600 MHz and made from 1333 MHz chips (common) will be just as reliable and compatible as memory rated for 1600 MHz but made from 1600 MHz chips, assuming no overclocking?

Yes. Its rated to run at that speed so it should run at that speed, and it does, correct.

If that was the case, why didn't the chip manufacturer rate those 1333 MHz parts for 1600 MHz? Could it be that they failed testing at 1600 MHz? I mean real testing, not running in a PC overnight.

 

[moocow]

If that was the case, why didn't the chip manufacturer rate those 1333 MHz parts for 1600 MHz? Could it be that they failed testing at 1600 MHz? I mean real testing, not running in a PC overnight.

Why would they bin through all the chips when they can sell them wholesale and let other companies sort out which chips can do crazy frequencies and which ones wont?

It is pretty much the same reason that Intel doesn't bin through all their chips and figure out which ones will do 5ghz out of the box and other ones will do 3.4ghz. They rate them for the worst-possible circumstances like server usage. A chip rated to 3.4ghz at 110*C will likely run for the same length of time as a chip doing 4.5ghz at 60*C. Dram IC providers do the same thing, they rate them low in case there are some bad chips that would not be able to meet the expectations of others in the same batches.

 

[EarthDog]

Larry, I get you, I do. But my point is, if the sticker says 1600MHz on it at CL9 and 1.5v, that is what I expect them to run at with no errors. I dont care that its made from 1333MHz 1.35v chips and the voltage raised to be stable at 1600MHz. I dont care. It works as described on the label and that is all that matters. Ram is ram in that respect. 99% of people dont care about the issue you have brought up a couple times. Its just ram, it runs at its rated speeds, I dont care how it got there, ya know?

 

[larrymoencurly]

Larry, I get you, I do. But my point is, if the sticker says 1600MHz on it at CL9 and 1.5v, that is what I expect them to run at with no errors. I dont care that its made from 1333MHz 1.35v chips and the voltage raised to be stable at 1600MHz. I dont care. It works as described on the label and that is all that matters. Ram is ram in that respect. 99% of people dont care about the issue you have brought up a couple times. Its just ram, it runs at its rated speeds, I dont care how it got there, ya know?

99% of the people don't care because 90% of them don't experience RAM problems (isn't the yield for RAM wafers now ~90%?), and most of the remaining blame problems on software or power surges. Most products that are just as good are not.

 

[larrymoencurly]

Why would they bin through all the chips when they can sell them wholesale and let other companies sort out which chips can do crazy frequencies and which ones wont?

It is pretty much the same reason that Intel doesn't bin through all their chips and figure out which ones will do 5ghz out of the box and other ones will do 3.4ghz. They rate them for the worst-possible circumstances like server usage. A chip rated to 3.4ghz at 110*C will likely run for the same length of time as a chip doing 4.5ghz at 60*C. Dram IC providers do the same thing, they rate them low in case there are some bad chips that would not be able to meet the expectations of others in the same batches.

But Intel and AMD actually bin like crazy, not just by clock speed but also temperature range and voltage. Why would they do that if their chips could easily exceed their clock ratings by huge amounts? OTOH CPUs almost always run very reliably at their rated specs, while quite a few memory modules -- virtually all of them heatsinked or made with private label chips -- do not.

 

[EarthDog]

99% of the people don't care because 90% of them don't experience RAM problems (isn't the yield for RAM wafers now ~90%?), and most of the remaining blame problems on software or power surges. Most products that are just as good are not.

well, I appreciate your insight, again...but im sticking with ram is ram, lol!