Cheap DDR3 Capable of 2000MHz+

[Reefa_Madness]

@MrGenius

TB didn't give you a screen that looks like the attachment below? It displays after you click on "Read" and select one of the channels listed. In my attached S/S, note that if you look under "Dram Components" you see the Part Number, however, it doesn't show if it is BFR or CFR. However, at the bottom it does show timings for various different clock speeds, including 800MHz so it has to be a DDDR3-1600 part (CFR) otherwise it would have only listed timings up to 667MHz (BFR),

As for the red GSKill module, starting from the left, we have a mfg date of year 2015 week 02 (YYWW format). Looking at the codes for after April 201, the next 4 digits of 140H indicate Hynix and based on your pic your 4GB module appears to be a single-sided module therefore we are looking at a 4Gb density chip like MFR, instead of a 2Gb chip like CFR (or Samsung's HCH9). Because they are from 2015 it is possible that they could be MFR. There is (was?) a C-die and a Q-die. IIRC, there is also an AFR variant. Tried to locate a confirmation of that but couldn't find it.

Every now and then you get a kit that has a letter instead of a digit in that last spot of the 4-digit IC code like yours does. I seems that I've noticed the letter in kits originating outside of the US and others have expressed similar findings, but it really isn't confirmed.

So, in summary, none of those "name brand" modules are Samsung, but instead are Hynix.

BTW, when trying to identify what's under the hood, this thread, also from XS, is slap full of good info.

http://www.xtremesystems.org/forums/showthread.php?285767-DDR3-IC-thread

 

[MrGenius]

TB didn't give you a screen that looks like the attachment below? It displays after you click on "Read" and select one of the channels listed. In my attached S/S, note that if you look under "Dram Components" you see the Part Number, however, it doesn't show if it is BFR or CFR. However, at the bottom it does show timings for various different clock speeds, including 800MHz so it has to be a DDDR3-1600 part (CFR) otherwise it would have only listed timings up to 667MHz (BFR),

I get this with the Team sticks. So they're BFR then.

I get this with the G.Skill.

I didn't realize there was any useful info to be gleaned from any of that. I was partially right though. Since there ain't much.

 

[Reefa_Madness]

@MrGenius

I don't know where my brain was yesterday, perhaps off on holiday, but I misread some of the info off the TB screen. Several corrections are needed.

First off, when I looked at your Team labels I saw "8GB" and thought that was the total kit capacity, when in fact that is the individual module's capacity. I went back and looked at it again because the info on the TB screen didn't match up with what should be on an 8GB kit. It shows a 4Gb density module with a rated speed of 1333. Your double sided Team 2666 CL11 are Hynix MFR. The info was staring me right in the face so I have no excuse. You can read all that from the screenshot under the captions of "Speed Grade" and "Die Density". Another one of my misreads...the timings at the bottom of the image are first, the JEDEC standards and then below those are the timings the manufacturer has binned them for, in your case, 1333MHz or DDR3-2666 at CL11. For what it's worth, these are some of the best binned 16GB kits that were available, as most were CL12 at these speeds. I previously alluded to those being one more of the JEDEC timings when determining whether it was a 1333 or 1600 part and that was incorrect.

I didn't realize there was any useful info to be gleaned from any of that. I was partially right though. Since there ain't much.

Yeah, it isn't perfect. It is just one of the tools that you use, in conjunction with other available info, to make the best educated guess. In the end, removing the spreaders and looking at the actual chips is the best way. Even that isn't 100% as sometimes the chips have been relabeled. At that point you're left with measuring the size of the chips, the method used to mount them to the PCB and other technical means. I've never gone to that level. Most I'll do is remove the spreaders, and only on kits that I know the warranty is void, for example if I bought them second hand.

Note that the GSkill module doesn't have that last group of timings that the Team does and that would be because it's speed and timings ratings from GSKill are the same as the JEDEC standards for DDDR3-1333. Don't know if some of the other missing info on the GSkill is due to it not being programmed into the SPD, or read failure on the part of TB. It would be interesting to see what info CPU-Z could return for that module, as this is the basic stuff usually available from the CPU-Z Memory tab.

Speaking of available info, in case you don't know, Corsair uses version numbers in the same way GSKill uses their IC code. Kingston also used a numbering system on their packaging, and a letter code on the module labels. More about that is detailed here:

https://www.overclockers.com/forums/showthread.php/755487-DDR3-Memory-Chip-Information-Sources

Again, I keep forgetting the purpose of this thread was to make note of the Samsung chips available on eBay and we have certainly gone off on a tangent. Once again my apologies to the OP.

And in keeping with the HCH9 theme of this thread...for those in the US, or with friends in the US, I ran across these just a few minutes ago during my daily trip down eBay's "Memory Lane".

Well binned HCH9 from GSkill:

https://www.ebay.com/itm/G-SKILL-Tr...h=item4434613f77:g:7mEAAOSwnjZcTi6Z:rk:9:pf:0

 

[MrGenius]

Bah...the nightmare continues...

Got the M378B5273DH0-CK0 sticks with the BCK0 chips in today (Sorry about the pic quality...my phone needs charged...had to use the backup cam).

Good news first? They boot @ 2666 12-15-15-47 2T with 1.76V. If you can call that good. They were cheap too...I guess. $45 shipped for 16GB.

Bad news? Oh...there's plenty of that. My motherboard hates them. Or something. I don't quite get it. The problem...I think... is the amount of RAM it detects and, therefore, it not running them in dual channel. If I reset the CMOS and go into the BIOS it says 16GB installed. I can check the memory pages that say there's 4 x 4GB installed with all the correct info. But as soon as I save and exit the BIOS...8GB go missing...ish. It's a total WTF? situation.

What do I mean by that? Depending on what I check it says one thing or another thing. Almost all of it being wrong in one way or another. CPU-Z says I have 16GB installed, 4GB per slot x4, with all the rest of the details as they should be...EXCEPT running in single channel. HWiNFO says exactly what CPU-Z says. How TF can I have all 4 slots filled and be in single channel? That's not even possible. MSI Control Center says I have 4 slots filled, 4GB per slot, dual channel, but the timings for channel A are all out of whack(channel B timings are fine, as in the same that everything else reports). AIDA64 says I have 4 x 4GB DIMMs installed for a total of 8GB? What the math? MemTest64 says Modules 0 and 1 are empty, but Modules 2 and 3 are 4GB DIMMs. Thaiphoon Burner sees a 4GB DIMM in each of the 4 slots. But Windows 10 also says I only have 8GB installed. And, like I said, if I go back into my motherboard BIOS after saving and exiting it says there's only 8GB now...but it also says all 4 slots are still filled with 4GB DIMMs. There doesn't appear to be a dead stick or sticks. They all seem to work. But my system just says "fock that snot...not havin' it! not all of it anyway!".

I dunno. I tried flashing a different BIOS on my mobo. Swapped the sticks around in the slots. Nothing. Same deal. I don't know if I even have the patience to try and figure this one out. I'm thinking I'll probably just return them and get my money back. Seller offers returns within 30 days and pays for the return shipping.

 

[DR4G00N]

These are not bad for the price, I have some from years ago. The particular speed bin isn't really important so long as it's D-die and they're all hit and miss oc'ing wise.
Only one of my six sticks will do 2800 9-12-12 @ 2.1V The rest need 10-13-13 @ 2.2V. 2666 9-12-12 @ 1.9V is bit more doable with 4 dimms.

One thing to note, 2GB dimms will oc better but they will be a bit slower overall compared to the 4GB dimms.

 

[MrGenius]

...they're all hit and miss oc'ing wise.

You got that right. It's an incredibly naive thing to think that you can get 4 of these sticks to overclock the same and/or with the same voltage. I'm learning that the hard way. You've got to bin each individual stick to see what they'll do. "This week or that week is a better batch" they say. BULLSHIT!!! You'll be goddamn lucky if you can get 2 from the same week's production to overclock the same and/or with the same voltage. Let alone 2 from different week's productions. Thinking you can get 4 out of 4, even from the same week? YOU'RE HIGH!!! So far it's looking like out of this latest batch I got, I've got 2 from 1338 that'll do 2666, one from 1338 that'll only do 2600, and the one from 1324 might barely do 2400. None of them with "tight" timings and/or "low" voltage. All barely do CL12 at those speeds with around 1.6V to 1.76V(haven't got stupid with the volts yet, and I'm not sure I intend to, I'm looking for daily-drivers). And they're ALL D-die(the BCH9s and the BCK0s).

EDIT: So...yeah. If you were guessing that's what the problem was. Good guess!!! You were right! Looks like my motherboard is smart enough to kick the duds to the curb and run the fastest pair in single channel all on its own. Which is pretty cool...in a way. But...not exactly what I'd want it to do.

 

[eidairaman1]

Bah...the nightmare continues...

Got the M378B5273DH0-CK0 sticks with the BCK0 chips in today (Sorry about the pic quality...my phone needs charged...had to use the backup cam).
View attachment 115728

Good news first? They boot @ 2666 12-15-15-47 2T with 1.76V. If you can call that good. They were cheap too...I guess. $45 shipped for 16GB.

Bad news? Oh...there's plenty of that. My motherboard hates them. Or something. I don't quite get it. The problem...I think... is the amount of RAM it detects and, therefore, it not running them in dual channel. If I reset the CMOS and go into the BIOS it says 16GB installed. I can check the memory pages that say there's 4 x 4GB installed with all the correct info. But as soon as I save and exit the BIOS...8GB go missing...ish. It's a total WTF? situation.

What do I mean by that? Depending on what I check it says one thing or another thing. Almost all of it being wrong in one way or another. CPU-Z says I have 16GB installed, 4GB per slot x4, with all the rest of the details as they should be...EXCEPT running in single channel. HWiNFO says exactly what CPU-Z says. How TF can I have all 4 slots filled and be in single channel? That's not even possible. MSI Control Center says I have 4 slots filled, 4GB per slot, dual channel, but the timings for channel A are all out of whack(channel B timings are fine, as in the same that everything else reports). AIDA64 says I have 4 x 4GB DIMMs installed for a total of 8GB? What the math? MemTest64 says Modules 0 and 1 are empty, but Modules 2 and 3 are 4GB DIMMs. Thaiphoon Burner sees a 4GB DIMM in each of the 4 slots. But Windows 10 also says I only have 8GB installed. And, like I said, if I go back into my motherboard BIOS after saving and exiting it says there's only 8GB now...but it also says all 4 slots are still filled with 4GB DIMMs. There doesn't appear to be a dead stick or sticks. They all seem to work. But my system just says "fock that snot...not havin' it! not all of it anyway!".

I dunno. I tried flashing a different BIOS on my mobo. Swapped the sticks around in the slots. Nothing. Same deal. I don't know if I even have the patience to try and figure this one out. I'm thinking I'll probably just return them and get my money back. Seller offers returns within 30 days and pays for the return shipping.

Could be Timings too tight for the amount installed and Clock speed.

I saw that happen on my platform when I tried my 16GB kit of Ripjaws X 2133 at 2400 speed with a CAS of 9. The bios would detect all of it but windows at that point could only utilize only half of it.

Once I bumped it back to 10 it was fine.

Now at 2133 my kit can run at 9.

I am unsure if it can run 8 or 7 though.

 

[MrGenius]

This "kit"(not even a good name for it AT ALL) is just a clusterfrick nightmare. Every stick is so different from every other stick as far as timings, clock speeds, and voltage they want to run stable at...I can hardly get 2 of them to run...OVERCLOCKED AND OVERVOLTED...with the same settings at the same time. Total garbage for OCing in my book. They run stock as perfect as they possibly could though. Good thing I didn't pay a lot for 'em. Cuz I sure ain't gonna get a lot out of 'em...

I don't know what else to even say at this point. I have learned a valuable lesson though. I'll NEVER go down this road again. That's FOR SURE!!! "Buy generic RAM and OC it to the point that it performs as well as, or better than, the spendy stuff" they said. Yeah...right. Go to Vegas and gamble the money instead. You've got just about as much chance of coming out a winner. There's nothing wrong with budget RAM per say. It's the idea that you can get more than what you pay for that's...well...a crock of shit. You either get lucky...or you don't. And that's all there is to it. Trying to get lucky does not a good idea make.

Turns out I'm and idiot...see post below.

 

[eidairaman1]

This "kit"(not even a good name for it AT ALL) is just a clusterfrick nightmare. Every stick is so different from every other stick as far as timings, clock speeds, and voltage they want to run stable at...I can hardly get 2 of them to run...OVERCLOCKED AND OVERVOLTED...with the same settings at the same time. Total garbage for OCing in my book. They run stock as perfect as they possibly could though. Good thing I didn't pay a lot for 'em. Cuz I sure ain't gonna get a lot out of 'em...

I don't know what else to even say at this point. I have learned a valuable lesson though. I'll NEVER go down this road again. That's FOR SURE!!! "Buy generic RAM and OC it to the point that it performs as well as, or better than, the spendy stuff" they said. Yeah...right. Go to Vegas and gamble the money instead. You've got just about as much chance of coming out a winner. There's nothing wrong with budget RAM per say. It's the idea that you can get more than what you pay for that's...well...a crock of shit. You either get lucky...or you don't. And that's all there is to it. Trying to get lucky does not a good idea make.

I havent used value ram in some time, I know kingston used to be top for that. I found mushkin just to be as good or better.

But yeah random builds these people pull ram from= box o chocolates

 

[Regeneration]

My second order from China arrived and I've been lucky again.





M378B5273EB0-CK0 with K4W2G0846P-HC15 chips.

Don't have time to test it yet, but a Google search shows the chips are used on G.Skill 1600 CL7 and 2133+ kits:

https://ru.gecid.com/prtart.php?id=36984

https://www.techpowerup.com/forums/threads/gskill-ripjaws-2133-is-using-gddr3.208003/

https://www.hkepc.com/11617/RipjawsX_DDR3-1600_C7登場_G.SKILL_F3-12800CL7D-8GBXM/page/2

 

[MrGenius]

Well...I've sort of had a change of heart about the "kit" I was having problems with. As I've been able to sort out the issues to a fairly satisfactory extent(basically just had to unlink the timings between channel A and B, so they'd boot and all be detected). And, once again, in the end they aren't nowhere near as bad as I thought they were. And it turns out they too are a pretty spectacular bargain in terms of price vs. performance. I didn't really get the boost in performance I was after. But...this "kit", like the other, performs pretty much the same as the "kit" I've been trying to upgrade from. Which is damn good, considering I only paid about half as much as the first Samsung "kit". And only about a third as much as the spendy "kit" I'm trying to upgrade.

Yadda yadda yadda...this is how they stack up.

16GB Team Xtreem LV 2666 @ 2400 10-12-12-32 1T w/ 1.58V $150

16GB Samsung M378B5273DH0-CH9(BCH9) 1333 @ 2400 11-11-11-30 2T w/ 1.81V $80

16GB Samsung M378B5273DH0-CK0(BCK0) 1600 @ 2400 12-12-12-34 2T w/ 1.5V $45

As you can see...pretty much equivalent in performance(within margin for error). The only thing significantly different is the price.

But...I've got one more Samsung "kit" on the way too. So there's still a chance I picked a REAL winner(in terms of price to performance ratio).

 

[Regeneration]

Glad to see you're happy with the cheap Samsung memory.

The one i got today (M378B5273EB0-CK0) runs at 11-11-11-30 and somehow beats my expensive G.Skill 2400 10-12-12-31 kit.

 

[DR4G00N]

BTW for D-die you want to run the tRCD & tRP two or three increments higher than the tCL, eg. 9-12-12, 10-12-12, 10-13-13, ect.

@MrGenius
The likely reason they're not doing great is probably because your on Ivy. Ivy IMC is not nearly as good as Haswell especially when loaded with all four dimm's. 2400 @ 2T is probably the best you'll get with four dimm's without loosening the timings up a lot more. M-power is also not the most amazing mem clocker overall, Asus Z77 or Z77 OC Formula smoke it.

Newer DDR3 platforms will always oc mems better than older ones. These samsung's on X58 need 2.2V for 2220 8-10-10 1T as an example.

 

[Bones]

I'll throw this in as an example from a run I did sometime ago, these being a set of G. Skill PSC chipped sticks with a stock rating of CL6-8-6-24.

This example is a newer kit of DDR3 from Corsair, you can see what they are as indicated by the CPU-Z SPD tabs, rated as PC 2800 sticks.

For a set of G. Skill DDR3 PSC chipped sticks the 5th through 8th numbers of it's model string are "1040". If you see a set of DDR3 sticks from G. Skill with that on them, they are PSC chipped sticks and what you'd want to get.
I'll also say unless I "Know" what I'm seeing are the actual sticks I'm buying I'll pass on listings that say they have some but what you see may or may not be what you get....
Or those using stock photos.

 

[MrGenius]

BTW for D-die you want to run the tRCD & tRP two or three increments higher than the tCL, eg. 9-12-12, 10-12-12, 10-13-13, ect.

@MrGenius
The likely reason they're not doing great is probably because your on Ivy. Ivy IMC is not nearly as good as Haswell especially when loaded with all four dimm's. 2400 @ 2T is probably the best you'll get with four dimm's without loosening the timings up a lot more. M-power is also not the most amazing mem clocker overall, Asus Z77 or Z77 OC Formula smoke it.

I'm not having any problems running the tRCD and tRP the same as the tCL. That works just fine. The problem I have with this motherboard(and my Z77A-GD65 Gaming) is once it decides what the tCL(and CR) should be...that's as good as it'll do. Everything else I can mess with to an extent. Also voltage only helps with the clocks and not the timings AT ALL. But it is what it is. And as far as 2400 @ 2T with 4 DIMMs...it'll do better than that actually. I can run the 4 x 8GB Team sticks @ 2400 10-12-12-32 1T w/ 1.58V stable as a table all day long.

 

[bogmali]

If you don't have anything to contribute besides hurling insults-stay away please. Thread cleansed

 

[MrGenius]

The M378B5273DH0-CH9(HCH9) sticks came in today. I've had time to test them, and form my final conclusions on this subject. Foremost of which is the realization that I'm limited by what my Z77 motherboard(MSI Z77 MPower) and CPU(i7-3770K) will do with these Samsung ICs. And limited by the Z77 platform in general. I've not been able to get any RAM(Samsung or other) 100% stable above 2400. I've gotten 2600, 2666, and even 2800 to boot and run some quick benches. But 2400 is where I cap out for 100% stability. Regardless of how much is installed, and/or how fast it's rated for/capable of, and/or how "tight" the timings are(or aren't), and/or whatever amount of voltage I throw at it. Which is all stuff I pretty much knew already. I was really just trying to get something that performed a little better @ 2400(than my Team Xtreem LV 2666 kit). Which I just barely might have succeeded at. At a fairly high cost(in terms of $). But whatever. I'm over that. It's all good in the hood!

So...this is how it all played out. I was able to get all 12 4GB Samsung sticks to run stable @ 2400 with 1.5V. The only difference was the timings I could get them running stable with. I was able to get 8 Samsung sticks to boot @ 2600 with a little bit more voltage. And the other 4 Samsung sticks to boot @ 2666 with a little bit more voltage. None of them(Samsung) would boot above 2666. That boiled down to this(as far as what sticks do what):

Samsung M378B5273DH0-CH9(BCH9)
Boots @ 2600
Stable @ 2400 11-11-11-31 2T 1.5V
Price = $70 for 16GB ($4.38 per GB)

Team Xtreem LV 2666
Boots @ 2800(1 stick out of 4) or 2666(4 sticks out of 4)
Stable @ 2400 10-12-12-32 1T 1.58V
Price = $300 for 32GB ($9.38 per GB)

Samsung M378B5273DH0-CH9(HCH9)
Boots @ 2600
Stable @ 2400 11-12-12-33 2T 1.5V
Price = $80 for 16GB ($5 per GB)

Samsung M378B5273DH0-CK0(BCK0)
Boots @ 2666
Stable @ 2400 12-12-12-34 2T 1.5V
Price = $45 for 16GB ($2.81 per GB)

As far as performance(it's basically as shown above...but specifically as below):

1. Samsung M378B5273DH0-CH9(BCH9) are just barely faster in latency than Samsung M378B5273DH0-CH9(HCH9) & Samsung M378B5273DH0-CK0(BCK0)(and performance as close as possible to identical to those in all other aspects), and barely faster write speed than my Team Xtreem LV 2666 sticks @ 2400 10-12-12-32 1T 1.58V(and performance as close as possible to identical to those in all other aspects).

2. Samsung M378B5273DH0-CH9(HCH9) & Samsung M378B5273DH0-CK0(BCK0) perform identically to each other(or as close to it as possible), with barely faster write speed, but slightly slower latency, than my Team Xtreem LV 2666 sticks @ 2400 10-12-12-32 1T 1.58V(and performance as close as possible to identical to those in all other aspects). So...basically a tie for second place for those 3.

And...the winner by a nose is...Samsung M378B5273DH0-CH9(BCH9)!!!

THE END

 

[Regeneration]

Don't forget about the price. $100+ vs. $34 for dual. Memory vendors ripped us off all this time.

 

[MrGenius]

Don't forget about the price.

Added.

 

[MrGenius]

Well...I thought it was all said and done. Turns out it wasn't. Close...but not quite.

What happened was I decided to try and see if I could push any of these Samsung's to a little higher clock speed than I got them up to previously. Like start with whatever they did before and add a little BCLK to it. Since I was running them all, as I always do for daily use, with the BCLK @ stock(100MHz). Anyhow...in the process of which things turned out a bit differently than I stated previously. I'll just cut to the chase. This is how that all turned out(except for the Samsung M378B5273DH0-CK0(BCK0), which I sold to another member here yesterday, and couldn't retest):

Samsung M378B5273DH0-CH9(BCH9)
Boots @ 2746(+106% OC) 11-15-15-48 2T 1.81V(4 x 4GB)
Stable @ 2400(+80% OC) 11-11-11-31 2T 1.5V(4 x 4GB)
Price = $70 for 16GB ($4.38 per GB)

Team Xtreem LV 2666
Boots @ 2866(+7.5% OC)* 12-14-14-37 2T 1.98V(1 x 8GB), or 2666(rated speed)* 11-13-13-35 2T 1.65V(4 x 8GB)
Stable @ 2400(-10% UC)* 10-12-12-32 1T 1.58V(4 x 8GB)
Price = $300 for 32GB ($9.38 per GB)

Samsung M378B5273DH0-CH9(HCH9)
Boots @ 2744(+105.9% OC) 11-15-15-48 2T 1.83V(4 x 4GB)
Stable @ 2400(+80% OC) 11-12-12-33 2T 1.5V(4 x 4GB)
Price = $80 for 16GB ($5 per GB)

Nothing changed stable performance wise. I was just able to clock the Samsung's quite a bit higher than I did before(needed some more SA and IO volts to get them in the 2600+ range). And I failed to mention that the one 8GB Team stick that clocks higher than the other 3, actually clocks a bit higher than I mentioned previously. But, in actuality, I haven't tested each individual Team stick for the highest clock it will run yet(I just know only one of them will boot @ 2800+, and the rest boot @ 2666 at least). I also added some CPU-Z Validator links for proof, plus the voltages, # x capacity of DIMMs, and timings for the Boots @(but not stable @) clocks too. I have a feeling I could have gotten the Samsung M378B5273DH0-CK0(BCK0) sticks to boot @ 2800+ CL12. But, like I said, SOLD. Oh well. Wouldn't have done me any good if they did. Just bragging rights...

*Technically these were built using chips rated for 1333. So @ 2400(+80% OC), @ 2666(+100% OC), and @ 2866(+115% OC)...if you prefer.

 

[Bones]

I've noted with mine sometimes using a little BCLK can squeeze a little more from them rather than stepping up to the next divider.
Everything has a limit, some of it is in how you get it there.
Good stuff here guys, glad to see success with this.

 

[Reefa_Madness]

@MrGenius , it occurred to me that you've shorted the overclock % of the Team 2666 kit just a tad.

Looking back at the TB screen from your post #27 it shows that those Hynix MFR are really a DDR3-1333 part, exactly like the Samsung HCH9 are a 1333 part. The only difference is that Team has binned the chips...then overclocked them 100% and sold them as 2666. You took them a little further when you booted up at 2866. So comparing apples to apples, by my calculations that's more along the lines of a 115% overclock over the chip manufacturer's speed rating for them.

 

[MrGenius]

@MrGenius , it occurred to me that you've shorted the overclock % of the Team 2666 kit just a tad.

Looking back at the TB screen from your post #27 it shows that those Hynix MFR are really a DDR3-1333 part, exactly like the Samsung HCH9 are a 1333 part. The only difference is that Team has binned the chips...then overclocked them 100% and sold them as 2666. You took them a little further when you booted up at 2866. So comparing apples to apples, by my calculations that's more along the lines of a 115% overclock over the chip manufacturer's speed rating for them.

True. But I chose a different perspective. One of "they've been requalified to do this speed now". So any more than that I can consider a bonus OC. And when I run them at a lesser speed I consider it an underclock, since they've been proven to be able run faster. Technically, it's wrong. I know. But it's right, in a way, too.

I guess I'll make a note of it.

 

[emissary42]

@MrGenius: The Z77 MPower should not have issues with stability @ 2666/2800 with a 3770K unless the CPUs memory controller is really crappy. Have you tried manual subs and voltages?

Don't forget about the price. $100+ vs. $34 for dual. Memory vendors ripped us off all this time.

Not all retail kits are rip-offs. You will have a hard timing finding original Samsung 2Gbit D-Die that can to 2666+ 10-12-12 with tight subs under 1.65V fully stable. Good retail bins can do that without issue and will have voltage tolerance to hit 2800/2933 9-12-12 with 1.85/2V. The binning costs are also factored into the retail prices among other more obvious things like warranty and such. If you understand the scaling properties for certain ICs you can very quickly see which kits are really worth their price premium... because the manufacturer had to run a very tight bin to produce them. in some cases it probably was less than 1% of the chips that cleared certain specs, which in result made those kits quite expensive (and some also kinda rare).

 

[MrGenius]

@MrGenius: The Z77 MPower should not have issues with stability @ 2666/2800 with a 3770K unless the CPUs memory controller is really crappy. Have you tried manual subs and voltages?

I've tried everything. I think it basically boils down to what I consider stable and/or safe for 24/7 OC. I'm not willing to run what I consider to be excessive/unsafe volts for a 24/7 OC. Which is a matter of opinion. But I base my opinion roughly on as follows.

That's the main thing anyway. The other thing that applies is the performance gain, or lack thereof. So far I'm not able to get what I consider to be better performance running higher than a 2400 memory divider. Which is probably due to not being able to get "tight" timings and/or CR1 with memory dividers above 2400 no matter what I do. Not only with my Z77 MPower, but also with my other Z77 board(a Z77A-GD65 Gaming, which is no slouch either, and has been capable of everything the MPower can do, but the Vcore VRMs run a bit hotter doing it). As far as one 3770K vs. another IMC wise...I've had 1 other(death from Vcore overvolt) and it was exactly the same as this one in that regard. I've also had a total of 4 3570Ks(killed 3, 1 still living) that were capable of doing the same as my 3770Ks in the memory department(on either of my Z77 boards). So...I've either never had an Ivy Bridge K with a weak IMC...or they've all been weak. IMO though...there's no such thing as a weak Ivy Bridge K IMC. All the same IMO, and/or IME. An opinion that I noticed was shared by @der8auer as well. Whom stated, much as I've experienced personally, that an Ivy Bridge K should be capable of what he calls a 24/7 memory OC of at least 2400...but any higher than that is very unlikely. Anyhow...I digress...back to the performance @ 2600+. I can't get better performance in anything but slightly faster latency running 2600+ compared to 2400. Which, like I said, is probably due to the "loose" timings and/or not being able to get CR1. Write speed usually stays about the same, but read and copy speeds either also stay about the same or start tanking hard @ 2600+(depending on which particular DIMMs). So even if I could get 2600+ "stable", it's either not going to gain me much in performance(if anything at all), and will invariably need more voltage(in one form or another) than I'm willing to run with 24/7. Not gonna run the SA or IO with more than 1.1V, or the Vmem with more than say 1.7V for a 24/7 OC. Maybe I'm being too cautious, or maybe I'm not. Regardless, nothing to be gained performance wise anyway so far as I can tell. When the sweet spot performance wise appears to be 2400 with .94V SA, 1.05V IO, and 1.5V-1.58V Vmem.

Oh...forgot about what I consider to be "stable". 20 runs of MemTest64 with 8GB of 16GB(or 16GB of 32GB) and NO ERRORS.

EDIT: Actually I lied about the CR. I can get the Team sticks to run @ 2600 11-13-13-34 1T with 1.65V. But performance, in general, is slightly less than/the same as @ 2400 10-12-12-32 1T(aside from possibly slightly faster latency, or possibly not even, basically within margin for error IMO). And they don't pass my stability testing method anyway(or at least not with voltage I'm comfortable running 24/7, and I haven't even tried with more than that, nor do I intend to, pointless).