Tested vs SPD?

[Iamazn]

In practical terms, considering what kind of memory available on the maket, your only choices are:

A. Good manual overclocking (run Samsung 1600 MHz @ 2133 MHz)

B. Bad automatic overclocking (G. Skill 1333 MHz @ 2133 MHz)

C. Manually under-overclocking the automatically overclocked memory (G. Skill 1333 MHz @ 1333 MHz)

If I had to overclock, I'd choose A first, no questions, especially because it has such a good track record, and 1600 MHz chips can probably run faster than 1333 MHz chips. And with those Samsung modules you'll get something that's rock-solid at 1600 MHz, no matter which sample you receive.

A. As I said before, I'd rather not do any manual overclocking. I'd prefer to buy a kit that has the speed/timing I want at stock...
B. To be honest, if the "stock" speed can be changed from 1333MHz to 2133MHz via XMP profile, I'd rather have that. If it can't, I can just RMA it, right?
C. What...?

What on earth are you talking about larrymoencurly? G.Skill rates their sticks for what they will run. I have never had a set of G.Skill RAM fail to run its rated speed unless it was an IMC problem, from DDR2-1066 to DDR3-2666. However there is a strong potential for a Samsung 1600 kit to not run at 2133. Plus, because it's not rated for where he wants to run, he's out of luck on an RMA.

So the kit I posted will be able to run 2133mhz CL9 without any manual overclocking? i.e. All I would have to do is load the XMP profile?

 

[Suppressor1137]

Ok, I'm going to make it veeeeeerrrry simple.


Pick a speed, pick a brand, Pick size of module, Click buy.


You will NOT notice ANY performance gain for ANYTHING over 1600 cl9s

you WILL reach 7.9 WEI with these speeds, and 8 gigs is the minimum I recommend. Games and programs are starting to use more and more ram, I'm actually finding I'm pulling 7.4 gigs of ram at 100% use.

Overclocking RAM has very Little benefit, Only really benefiting extreme high cpu clocks for synch rates and bus stability.

It really IS that simple.

Buy a Fancy heat spreader Memory kit, or buy a stripped down bare kit, The difference is so minuscule that it isn't even funny.

 

[freeagent]

I cant say this for the newer intel chipsets, but on x58 memory speed dictates uncore speed, wich inturn dictates memory latency, and memory bandwidth, not that those are a problem to begin with. Alternatively you can manually adjust wichever setting youd like. But for stock clocks, I would probably go with the fastest stuff I could find. Yes you could get away with running whatever stuff youd like, and you would more than likely not notice a difference, unless you have ocd

Do I notice a difference between 1600 6-6-6-x vs 1600 9-9-9x? A little bit..

 

[hokiealumnus]

So the kit I posted will be able to run 2133mhz CL9 without any manual overclocking? i.e. All I would have to do is load the XMP profile?

Assuming you don't have a horrible clocking IMC (not sure what CPU you're using either), yes, the kit linked in the OP will run their rated 2133 / 9-11-11 @ 1.6V as specified. You should be able to set XMP and forget it.

As far as the 2400 kit you linked, what CPU are you running? That's going to have the biggest impact on what you'll be able to get.

You should also know the other guys in this thread mentioning the speed difference above 1600 isn't very noticeable are quite right. That doesn't mean I don't overclock RAM, because for some strange reason I just really enjoy clocking RAM. Plus I overclock to benchmark on HWBot, where it most definitely does make a difference.

Also, so larrymoencurley is (hopefully) satisfied with our test methods, I use a hardware-based test solution, the Ultra-X R.S.T. Pro 3 to ensure stability. Since you place lots of emphasis on cost, while these are available by quote only, I saw mention of a price ~$800 a couple years ago. Corsair, Hynix, Kingston, Micron, Patiriot, Samsung & SuperTalent all use them. It's "PC-based", but certainly no Memtest86/86+ and I'd call it sufficient considering Ultra-X's customer base.

 

[EarthDog]

http://www.overclockers.com/forums/showpost.php?p=7225659&postcount=7

Went over that before with Larry. Hopefully there is a more thorough understand with this level of explanation.

 

[Iamazn]

Assuming you don't have a horrible clocking IMC (not sure what CPU you're using either), yes, the kit linked in the OP will run their rated 2133 / 9-11-11 @ 1.6V as specified. You should be able to set XMP and forget it.

As far as the 2400 kit you linked, what CPU are you running? That's going to have the biggest impact on what you'll be able to get.

You should also know the other guys in this thread mentioning the speed difference above 1600 isn't very noticeable are quite right. That doesn't mean I don't overclock RAM, because for some strange reason I just really enjoy clocking RAM. Plus I overclock to benchmark on HWBot, where it most definitely does make a difference.

Also, so larrymoencurley is (hopefully) satisfied with our test methods, I use a hardware-based test solution, the Ultra-X R.S.T. Pro 3 to ensure stability. Since you place lots of emphasis on cost, while these are available by quote only, I saw mention of a price ~$800 a couple years ago. Corsair, Hynix, Kingston, Micron, Patiriot, Samsung & SuperTalent all use them. It's "PC-based", but certainly no Memtest86/86+ and I'd call it sufficient considering Ultra-X's customer base.

I plan on getting the 3770k.

 

[hokiealumnus]

You should be ok up to 2400 depending on motherboard. Some 3770K's don't play well above that. It's very rare that one can't get to 2400. Any Ivy Bridge chip + Z77 mobo combo on the market should do 2133 without issue.

 

[larrymoencurly]

What on earth are you talking about larrymoencurly? G.Skill rates their sticks for what they will run. I have never had a set of G.Skill RAM fail to run its rated speed unless it was an IMC problem, from DDR2-1066 to DDR3-2666. However there is a strong potential for a Samsung 1600 kit to not run at 2133. Plus, because it's not rated for where he wants to run, he's out of luck on an RMA.

G.Skill rates it sticks for what they will run -- barely, and even they say they ship sticks with up to 2 bad bits.

Why should G.Skill 2133 MHz modules made with 1333 MHz chips be more likely to run reliably at 2133 MHz than Samsung chips made with 1600 MHz chips? It's not as if G.Skill does super-strict testing because their testers use nothing but PCs.

 

[hokiealumnus]

G.Skill rates it sticks for what they will run -- barely, and even they say they ship sticks with up to 2 bad bits.

Why should G.Skill 2133 MHz modules made with 1333 MHz chips be more likely to run reliably at 2133 MHz than Samsung chips made with 1600 MHz chips? It's not as if G.Skill does super-strict testing because their testers use nothing but PCs.

Wow, you are seriously anti-G.Skill or pro Samsung, one of the two.

They're more likely to run reliably at that speed because that is where they are guaranteed to run. If they do not, you can get another kit for the life of the sticks, free of charge (potentially shipping I suppose). That's why it's more likely...because no company in their right mind would ship kits that are going to cost them money by having to RMA them.

I use a PC to test RAM too, with the aforementioned R.S.T. Pro 3. So do all those other companies listed in my post. Why is it so bad to test RAM that's to go in a PC with a PC, as long as you have proper equipment to do so?

Why are you obsessed with the chip rating anyway? They are binned to run the speed at which they are rated. It's similar to GPU companies that come out with overclocked cards. It's the same GPU, but binned to be guaranteed to run at the speed rating at which they're sold. This is no different.

EDIT - How do you even know the sticks the OP is proposing have those ICs? From what I've seen, G.Skill has been swapping around Hynix and Samsung ICs as they're available recently, especially in the 2400 & 2600+ kits. Reference this thread @ XS; there are kits that run at rated speed with very little headroom and some that clock quite high; it's luck of the draw over rated speed.....but they'll definitely run rated speed, which is my entire point and is what the OP is after - set it and forget it XMP.

EDIT II - I guess what I'm trying to figure out is why you keep harping on this? The OP has stated he wants to set it and forget it. Meaning he wants to set the XMP profile, save & exit and go on about life. He literally can not do that with the Samsung sticks. Not possible. They don't have XMP programmed into them. He will be forced to overclock the sticks, which he doesn't want to do. I completely don't understand any logic to your continuing along this line.

 

[larrymoencurly]

A. As I said before, I'd rather not do any manual overclocking. I'd prefer to buy a kit that has the speed/timing I want at stock...
B. To be honest, if the "stock" speed can be changed from 1333MHz to 2133MHz via XMP profile, I'd rather have that. If it can't, I can just RMA it, right?
C. What...?

So the kit I posted will be able to run 2133mhz CL9 without any manual overclocking? i.e. All I would have to do is load the XMP profile?

That's the way it's supposed to work.

If there were better choices, I'd avoid manually overclocking, too, but the chips in those RipJaws aren't tested very strictly for their XMS profiles but are simply run in PCs, often at room temperature, rather than in expensive machines that can analyze the access time for each individual bit. I've seen a lot more errors with modules made from overclocked or subprime chips than with plain-Jane modules full of prime chips, even when the latter were overclocked. Also you're more likely to have success overclocking memory just 33% (what the Samsungs need to go 2133 MHz) than 60% (what G.Skill does with its 2133 MHz modules).

If you're going to rely on RMA for your quality control, choose Kingston because if you talk with higher level tech support, they send out modules that have been tested more thoroughly than normal.

 

[larrymoencurly]

Wow, you are seriously anti-G.Skill or pro Samsung, one of the two.

No and no. I just don't like subprime or factory overclocked chips because they're not reliable. It doesn't help that the module companies all brag about producing only the finest kind, American GI!

They're more likely to run reliably at that speed because that is where they are guaranteed to run. If they do not, you can get another kit for the life of the sticks, free of charge (potentially shipping I suppose). That's why it's more likely...because no company in their right mind would ship kits that are going to cost them money by having to RMA them.

At least not gross amounts of errors that make customers blame the memory, but friends have had problems getting Crucial to exchange Ballistix and Corsair to exchange XMS more than a few times in a row. BTW Harbor Freight often warrants its tools better than Sears or Snap-On do, so the quality must be just as good or good enough, right?

I use a PC to test RAM too, with the aforementioned R.S.T. Pro 3. So do all those other companies listed in my post. Why is it so bad to test RAM that's to go in a PC with a PC, as long as you have proper equipment to do so?

While Ultra-X RST cards are a lot better than just a PC, they're still not anything like the machines the chip makers use. Corsair also uses RST cards but has managed to send out heaps of duds in the past 5 years, and in the past year I've been getting maybe 30% bad DDR from them (lots of old Dells needing more memory).

Why are you obsessed with the chip rating anyway? They are binned to run the speed at which they are rated. It's similar to GPU companies that come out with overclocked cards. It's the same GPU, but binned to be guaranteed to run at the speed rating at which they're sold. This is no different.

I'm not obsessed unless you're crazy.

There's binning and then there's binning, and I'd trust a chip company's binning a lot more than any PC-based binning. If binning was all the same, how do chip makers justify testers that cost thousands of times as much as RST cards (even costlier when throughput is considered), and why does Kingston charge industrial customers extra for testing with the more expensive machines?

EDIT - How do you even know the sticks the OP is proposing have those ICs? From what I've seen, G.Skill has been swapping around Hynix and Samsung ICs as they're available recently, especially in the 2400 & 2600+ kits.

I don't, and I was referring only to RipJawsX 2133 MHz, which have been sold in versions overclocked by as much as 60%.

Reference this thread @ XS; there are kits that run at rated speed with very little headroom and some that clock quite high; it's luck of the draw over rated speed.....but they'll definitely run rated speed, which is my entire point and is what the OP is after - set it and forget it XMP.

How can you say they have very little headroom if you don't know the headroom? You admitted that you weren't allowed to remove the heatsinks. Also how would modules not hand-picked by G.Skill test out?

EDIT II - I guess what I'm trying to figure out is why you keep harping on this? The OP has stated he wants to set it and forget it. Meaning he wants to set the XMP profile, save & exit and go on about life. He literally can not do that with the Samsung sticks. Not possible. They don't have XMP programmed into them. He will be forced to overclock the sticks, which he doesn't want to do. I completely don't understand any logic to your continuing along this line.

Because a lot of memory is junk, ECC is rarely used, and bad bits cause problems that aren't always so obvious, and XMP profiles aren't tested nearly as strictly as what chip makers do. What's not to understand?

 

[hokiealumnus]

Regarding headroom, if you read through the thread, it's headroom over rated speed to which I was referring. Not headroom over your perceived 1333-rated chips.

I really do hate to say something like this, especially regarding a member that has been here for a decade (congrats btw, just added your Decennial banner!), but Iamazn, I'd recommend you just disregard what larrymoencurly is saying.

You can get RAM from whomever you wish, G.Skill or Kingston, or whoever; take your pick. If you want the experience you are after - setting XMP and done - buy sticks with XMP profiles rated at the speed, timings and voltage you want. What he is recommending is precisely what you don't want to do, so I'd recommend you just disregard it.

FWIW, I'm sorry larrymoencurley and do not mean this in a mean-spirited way at all. I understand you think you have a point, but in this case you're just giving the wrong advice. You may have a point that's worth discussing, but it should really be discussed elsewhere, not in this gentleman's thread. He asked a simple question and deserves a simple answer if he can get one. In this case, it's completely possible and the right way to go.

 

[larrymoencurly]

FWIW, I'm sorry larrymoencurley and do not mean this in a mean-spirited way at all. I understand you think you have a point, but in this case you're just giving the wrong advice. You may have a point that's worth discussing, but it should really be discussed elsewhere, not in this gentleman's thread. He asked a simple question and deserves a simple answer if he can get one. In this case, it's completely possible and the right way to go.

I understood what he wanted to do -- run the memory at 2133 MHz to get better game performance, which is why I didn't tell him to not do it and why I recommended overclocking (probably the first time I've done that). And I actually was trying to keep things simple. Normally I would have said that memory speed hardly mattered and it would be better overclock the CPU a little and buy those Samsungs and run them at their stock 1600 MHz.