OC Report :: OCZ Blade 3x2GB PC3-16000 CL7

[3oh6]

It feels like a lifetime ago that I did my last OC Report and in this industry, it pretty much has been. Well unfortunately for you, my almost 100% completely healed white ass is back, and I have brought some toys. OCZ has always been a name most of us here recognize but one that seems to have fallen off the radar at the upper end of the benching circles as of late. Maybe it is just me but I don't hear OCZ anymore in the uber benching threads. With high performance i7 triple channel kits in full swing, OCZ has also joined the party at the top. Today I am in the OCZ spot of the 3x2GB PC3-16000 7-8-7 world, also known as Elpida Hyper country.


OCZ has reserved the absolute best of the best to their new Blade series of memory. The Flex EX series doesn't have this kit, the Reaper series doesn't have this kit, only the Blade series offers the 3x2GB PC3-16000 7-8-7 modules that are guaranteed to be Elpida J1108BASE, more recognizable by the single word...Hyper.

For those new to my OC Report's, the first rule of an OC Report is to not call an OC Report a review. The second rule of an OC Report is that it isn't a bloody review. Review's have proper 24/7 stability testing - or at least they should - and stuffy benchmarks that no one here cares about. An OC Report is just that, a report on the overclockability of the product in question. In this case, it will be memory and my overclock of choice will be slightly different from memory OC Reports of past. I will simply be running a single instance of 32M SPi. Reason being, 8x32M SPi is so bloody close to 24/7 in some cases I might as well use Prime Blend or HCI Memtest. So in keeping with the heritage of the OC Report, single 32M SPi stability is all I am asking for from each timing set and voltage.

Directory:

• Specifications & Pricing
• Photos :: But I Read It For The Articles
• Test Bed :: It's Classified
• It's Go Time :: 32M SPi OC Report
• Nothing Stuffy Here :: Benchmarking Results
• Conclusion :: Deep Thoughts...By 3oh6

I want to mention this now, these modules were sent to me directly from OCZ in beautiful Californ-I-A for review. I know nothing other than that. If you want to complain about these being a cherry picked kit, search for an OCZ rep here and fill their inbox, not mine. OCZ has never given me a reason to believe I receive anything but random retail samples in the past, and no reason to start now. I believe in full disclosure so there you have it...and you didn't even have to twist my arm. The mainstream review without my foul language, offensive demeanor, or blatant insults can be found at HardwareCanucks.com in due time.

[3oh6]

Memory is no big mystery. It is spec'd to run at a certain frequency, at certain timings, on certain chipsets, with a specified amount of voltage. There isn't a whole lot to it as far as manufacturer specifications. What IC's are powering a set of modules, however, is a different story entirely. Here is a list of the specifications you will find at the OCZ web site.


As mentioned in the opening, this is the top rated kit of the Blade series; or any OCZ series for that matter. Specified to run at DDR3-2000 (1000MHz) at timings of 7-8-7-20 on the i7 X58 platform with 1.65v, the OCZ Blade OCZ3B2000LV6GK are right up there with everyone else sporting top binned Elpida Hyper J1108BASE -MNH-E.


In August of 08, Elpida announced this fantastically spec'd DDR3 IC that had us all drooling at the possibilities. We were also so very skeptical with the announcement of these too good to be true 1250MHz modules. The 20-20-19 timings jokes were cracked and to be honest, there wasn't much fan fare at the time Elpida announced. Well, the jokes are not being cracked anymore. These IC's are capable of ultra low voltage operation at 1.2v as well as JEDEC standard 1.5v. The biggest difference between these IC's and others to date is the fact that they use copper interconnects as opposed to aluminum. It was an obvious progression of DRAM, but not one that anyone had done before.

Needless to say, factory spec'd parts at 1000MHz with timings of 7-8-7-20 and only 1.65v is not something any other IC can claim right now, so OCZ can't even change the IC's on these and not tell anyone. Elpida is the performance king at the moment, unfortunately they are also the king of cost. Here are some prices I could find for this kit at the time of posting...

• Newegg.ca - $510.99 CND
• Infonec.com - $519.00 CND
• Sohodiffusion.com - $525.90 CND
• Newegg.com - $430.99 USD
• Amazon.com - $473.99 USD

Obviously we aren't dealing with a mainstream product here. These prices are as high as the top DDR2 back in its time when it was king of the sand hill and Elpida Hyper shows no sign of falling. Equally spec'd kits from Corsair and others are priced similar...except the Kingston kits, but we won't open that can of worms, perhaps I just did? Either way, these are a premium product at a premium price. Hopefully prices will come down at some point but I don't have my hopes up for that to happen any time soon.

[3oh6]

Before we get to the clocking, here are just a couple photos of the package and the modules in question.

click for full size...

As we can see, the package for memory has evolved a bit with the triple channel packs. Manufacturers can no longer just stuff three of these sticks in a molded plastic shell with a cardboard insert and call it a day. OCZ has gone with two modules in one shell and a single module in another shell, then both jammed into a box. The box is of decent quality and does add another small layer of protection during transport.

click for full size...

The heat sinks that identify the modules as part of the Blade series are quite nice, and normally I don't find much of anything that simply serves an aesthetic purpose nice. Obviously the heat sinks do help dissipate heat but let's be honest...do they really need to be there? Have heat sinks ever really needed to be there? And no, I am not asking those questions, so don't answer and turn this thread into an "internet engineers" thermal dynamics discussion that holds as much water as a plastic jug that Brian uses to pour LN2 with. If you don't get the joke, it is because you are not cool enough to...deal with it.

click for full size...

These are mostly just glamour shots showing off the fact that I have a macro lens, but they do a good job showing the heat sink construction and thermal pads OCZ uses. And before you ask, no, I am not potentially killing these bloody sticks to take the heat sinks off to show you they are Elpida Hyper. You want to see the IC's? Look at the press release photo from Elpida in the specifications section. To be honest, the heat sinks look like they would come off quite easily but still, I don't want to give you the satisfaction.

[3oh6]

Here is the work horse. My ace. My right-hand man. My...you get the idea. The EVGA X58 3X SLI Classified is not only the best motherboard for high BCLK, or the best at running i7 at high clocks under LN2, or an absolute rock solid platform for air testing; the Classified also happens to clock memory pretty damn well. Here are some photos of the setup with my tired and aging Chilly 1 single stage. Yeah, I'm doing this with the processor cold...what, you thought this was going to be on air? Crack head

click for full size...



There is the jewel of my collection, the little 920 that could. It has been buried in that kneaded rubber eraser for dozens of sessions now without leaving. I am starting to wonder how much corrosion is going on inside that socket, oh well. The Blade modules are nice sized but definitely not too fat for 6 DIMM use and certainly not too tall to interfere with many - if any - heat sinks.

click for full size...



These two photos depict the setup as it will run for the entire testing of this memory. Now you may be wondering why the PCI-E and PCI card are mounted in the motherboard...and why the PCI-E card has a pot strapped to its face. You see, I prepped the board for a quick shake down of the GPU under LN2 because I wasn't expecting the memory to show up for another day. When the sticks knocked on the door, I wasn't about to tear down the GPU setup for LN2...and these modules were not going to sit idle overnight. Needless to say, I disabled the PCI-E port, popped in the PCI card at it was go time.

Hardware used:
MB:          EVGA X58 3X SLI Classified (BIOS S21S)
CPU:         Intel i7 920 (845B026)
CPU Cooling: Chilly 1 single stage
GPU:         ATI Mach64 PCI
NB Cooling:  Stock
SB Cooling:  Stock
PWM Cooling: Stock
PSU:         Corsair HX1000W (single EPS connector used)
HD:          Seagate SATAII 80GB 8MB NCQ
OS:          Windows XP Pro SP2

Ambient Temperature: 22-23C

There we go, you have now seen the modules, you know the setup, I have insulted a few people...yup, sounds like we are ready to roll

[3oh6]

Blah blah blah...this blurb is just here to keep those people's mouth shut who insist on trying to force their stability testing of memory on everyone else. I won't name names, you know who you are cstkl1.

This is a single 32M SPi OC Report. That is it, that is all. Nothing more, nothing less. Single 32M SPi is fast, it is a level of stability that most know, and it shows what these sticks can do with the best benchmark money doesn't have to buy. If you don't like it...get lost. If you want more stability testing, or are going to say "but his CPU is cooled by phase change", or something else that is going to make me want to smack you in the back of the head Rafer Alston style...leave. Do us both a favor and just go whine to the nearest bum on the street because he is going to care as much as I do.

And one last time...these modules were sent to me directly from OCZ as a review sample. Please refer to section one, paragraph four, line three, word thirteen for further details on what you can do should you feel this kit is cherry picked. Disclaimers aside...here are the results:



6-7-6-X 1T:
Intel X58 - EVGA Classified:
Click link for full size screenshot...
892MHz :: 1.657v | 917MHz :: 1.701v | 942MHz :: 1.758v | 967MHz :: 1.802v | 987MHz :: 1.856v

Off to a great start with 6-7-6. I was hoping for 1000MHz once I saw how scaling was going but it just didn't work out. At least all signs point toward 6-7-6 1000MHz being do-able for 32M at 1.90v. Might need a sprinkle more voltage. As it stands, I managed to keep tRAS at 18 the entire test range but did have to open RTL up to get the last 987MHz stable. I don't think it quite needed to be 57-59-61, but that is where I started and just didn't feel like working down from there.


7-7-6-X 1T:
Intel X58 - EVGA Classified:
Click link for full size screenshot...
1022MHz :: 1.657v | 1052MHz :: 1.701v | 1062MHz :: 1.758v

Like 6-7-6, 7-7-6 started off great with perfect scaling through 1.65v to 1.70v, not to mention the very nice jump up from 6-7-6 at the starting voltage. Unfortunately I hit a snag at 1.75v. The memory just stopped scaling at that point and I managed only 10MHz from the 1.70v clock. I played with tRAS, vTT, sub timings, all to no avail. Thinking back, the one thing I did not change - as silly as it sounds - was tRP. I should have opened that up to 7-7-7 to see if it made a difference. This was my last timing set I actually ran and it was the next morning already so forgive my brain fart as I was up all night clocking this memory...literally all night.

The clocking for the OC Report was done within 24 hours of getting the memory...and within 18 hours of actually starting. That is right, 18 hours of straight 32M SPi, I am a loser, deal with it. Eat your heart out Gautam...my sticks aren't even close to dead . This was after getting right drunk at dinner just before I started clocking, that definitely helped. So I aplogize for not trying tRP of 7...sue me. My address can be found on some bathroom stall somewhere, go find it and serve the papers.


7-8-7-X 1T:
Intel X58 - EVGA Classified:
Click link for full size screenshot...
1027MHz :: 1.657v | 1052MHz :: 1.701v | 1077MHz :: 1.758v | 1102MHz :: 1.802v | 1122MHz :: 1.856v

As much as the 7-7-6 was a bit of a letdown , 7-8-7 definitely wasn't by any stretch of the imagination. This kit just kept taking voltage and jumping 20MHz+ at a time with every 0.05v I fed it. I am completely shocked that this 920 - even on phase - was able to take this memory this high. And just for those about to ask, no, this likely isn't possible or even close to possible with the CPU on air. Don't complain when your kit doesn't do these numbers with your 920 and its ass IMC at 80C. But please prove me wrong if you can...seriously, cough up the screen shots, I really want to know if a CPU on air can clock memory this high. Don't be shy, no one will bite. Well...maybe someone here will but I won't disclose how I know.

Seriously though, another great kit of top binned Elpida Hyper, and from someone other than Corsair. It goes without saying, but I am saying it to keep those without common sense to miss it, these are results from a SINGLE KIT. This may, or may not be a good representation of what every kit will do.

[3oh6]

This will be my "results" page if you will, for benchmarks with this memory. Only the good stuff, nothing weak is going to the rim here. Check back often, something tells me this kit is going to be busy

[3oh6]

There are no deep thoughts, I am not Jack Handy. But after only having this memory for about 48 hours now, I am already impressed. This is my first kit of top binned Elpida Hyper and you know what they say...you never forget your first. I have been working with a PC3-15000 kit of Corsair Dominator-GTs which are also Elpida Hyper, but not their top bin...it shows. The difference between these two are not going to be seen by everyday use, but in SPi 32M, it is a night and day difference.



The most amazing thing is the ability of this dog meat 920 IMC to clock this memory this high. On air, this 920 cannot run 8x32M SPi with memory at 1000MHz 7-8-7...it just can't do it. Drop in my 965 and it plows through blind folded. Single 32M SPi is obviously quite a bit less load, and the CPU was under a single stage, but I was sure the 920 would gas out early anyway. If the 920 did run out of steam, then this kit is even further beyond all my expectations of it. You see everyone else running around with 1100MHz 7-8-7 sticks, but you never believe it will be you until it happens.

The Blade series should definitely be one of the choices on your list if you are looking for Elpida Hyper. I have no idea if their PC3-16000 CL9, PC3-15000 CL7, or PC3-12800 CL6 all have Elpida Hyper as other ICs can handle those clocks, but the PC3-16000 CL7 definitely do...and nothing but the top bin is worth it anyway. I have always said it about Corsair so why shouldn't the same hold true for OCZ...go top bin or go home

[r1rhyder]

Very nice start for a review. I really don't want to pump 1.856v into my new Blades or I would take you up on your offer. I will take your word for it that my 920 on air just won't do it. But who knows, maybe when the 975 comes out I will be back on phase. Maybe then. Mine will do 1000Mhz 7-7-7-20-1N no problem. Great stuff for sure.

[The-Fox]

Nice writing 3oh6
Amazing kit you got there, truly rocks

So, you can confirm/conclude that Cold (cascade in your case) helps IMC handle higher memory clocks ?
My 920 D0 tops out at ~4300 uncore speed, using watercooling (which kind of limits Dram speed), to get 4400Mhz working, I need to jack up the Dram voltage quite high, it started to react at about 1.80v but it wasn't enough, I gave up at that point
I use Kingston T1 (also Elpidia Hyper chips, probably not the best binned chips), but I believe it can do 2200Mhz CAS8

[tekjunkie]

Well, my D0 920 very close to the batch that you have(mine is 45B027) can do 4GHZ uncore with memory @ 2K 7-8-7-20 HyperPi 32Mx8 stable,on the classified on air.The kit that I used was the CL8 2k doms. They do 2K 7-8-7-20 1.69 vDimm and VTT was not that high(+175).I don't think it counts though as I did it with just two sticks in dual channel mode as one of the sticks died.I really don't know what is up with so many Elipda Hyper sticks dying.I personally had 1 stick die in both my Kingston Hypers and my doms. It's probably due to the 50nm fabrication process being not mature enough.

Needless to say great review, I mean great "OC report" Jody.

[3oh6]

Very nice start for a review.

that's not what you typed initially

my word isn't gospel, i just have to say things like that so that i don't have guys jumping in here after buying memory based on my results saying they can't do the same. not everyone understands that the memory controller dictates what memory can and cannot do...and that memory controller is sometimes weak.

Nice writing 3oh6
Amazing kit you got there, truly rocks

So, you can confirm/conclude that Cold (cascade in your case) helps IMC handle higher memory clocks ?

thanks, it really is a good kit i finally got.

there is no such things as conclusions when it comes to clocking, but the consensus is that cold helps to clock memory. i have never had memory that can push an IMC until now, so i have no conclusive testing done. at some point this board will go back to air cooling and i will see what this same setup can do on air. oh, and i am just on a single stage...weak one tuned for C2D. prolly not holding much below -30C even for single threaded 32M.

I don't think it counts though as I did it with just two sticks in dual channel mode as one of the sticks died.

no that doesn't count, and even if it was triple channel, it doesn't matter. IMC strength is somewhat untested but my initial hunch is it will be on a CPU by CPU basis, much like cold bug is. perhaps chips from the same batch may have similar IMC strength, but that can't be said with certainty right now...i don't think.

Needless to say great review, I mean great "OC report" Jody.

hahaha, thanks

[Expat GriZ]

Well Done Bubba!!! Nice report. Enjoyed the humerous writing!! As you were scrolling through the bios did you notice a XMP profile with your kit?? Mine doesn't have one, not that I'd use it but just wondering?? And in the limited time I had with my same kit before I left for work, they will do 1050 @ 1.67V 7 8 7 20 1N. So hopefully what you received is indicative of what they are selling to us consumers!!

Keep hammering it!!

[r1rhyder]

"that's not what you typed initially"

you were lurking? I went back and reread, that's why I edited. Yes, each chip does seem to have it's own color.

@Expat My set doesn't have XMP either, but no need.

[cstkl1]

love the oc review..

always love the order of ure posting
ty

[der8auer]

Great review

Amazing Kit. Maybe i'll buy it but not for i7

[eva2000]

Nice review Jody... everyone has said sub zero on cpu helps IMC and mem clocks... haven't tried it myself but looks like to be the case

Glad to hear you're near 100% healed too. Single 32M still has it's place too especially when it comes to Elpida Hyper in case of the single 32M issue.

Interesting to later see the difference between max memory clocks for:

cpu air + memory air cooled
vs
cpu air + memory TEC cooled
vs
cpu water + memory air cooled
vs
cpu sub zero + memory air cooled

I shall if time is permitting and i need a new evap mount from Kayl for 1366 with my 1/4HP SS should have results for the latter 3 configurations

[3oh6]

Well Done Bubba!!! Nice report. Enjoyed the humerous writing!! As you were scrolling through the bios did you notice a XMP profile with your kit?? Mine doesn't have one, not that I'd use it but just wondering?? And in the limited time I had with my same kit before I left for work, they will do 1050 @ 1.67V 7 8 7 20 1N. So hopefully what you received is indicative of what they are selling to us consumers!!

haha, glad you enjoyed the read...in one of those moods yesterday when i wrote it up.

you are absolutely right, there are no XMP profiles. i have updated the specs image, i swear i thought i saw XMP somewhere on the OCZ web site...and didn't even look at that part of CPU-Z to be honest. also thanks for the info on your kit, it is good to hear my sticks aren't going to be the only good ones from this series.

"that's not what you typed initially"

you were lurking?

@Expat My set doesn't have XMP either, but no need.

na, i just happened to see your post right after you posted it i think. i was copying and pasting to post in another forum. it gave me a chuckle, no worries...i won't hold it against you

yup, mine don't have XMP profiles either, edited the specs...my bad guys.

love the oc review..

always love the order of ure posting
ty

thanks cstkl1

Great review

Amazing Kit. Maybe i'll buy it but not for i7

thanks boss...let's see what they do on another platform. if i do end up getting the Rampage Extreme out of retirement, it will definitely see these sticks.

Nice review Jody

thanks George, i too will hopefully get a chance to do some air/phase testing of memory clocks...but i don't have one of those fancy TEC coolers for memory like you

[Oliver]

great going Jody , interesting read up for sure


cant wait to see some serious 32m damage ...

Regarding the ICs , when we are talking about differences seen from IMCs
and the scenarios George listed up , regarding memory clocking and how the Hypers does, its also highly worth taking batch of the ICs into consideration , especially when looking at Hypers, however with Hypers there are so many variables , but to sum it up , Elpida have hitted a very interesting IC and the amount of time being used to discuss these and not to mention bench them in the community , is crazy and how it should be

[chispy]

I always enjoy reading your OC Reports Jody , job well done. You made me laugh with your funny statements . Im looking forward for the results and i have added this tread to my favorites to keep an eye on it. Let me know what you think comparing Corsair
Dominator GT ddr3-2000 cas7 with this kit. Thanks jody.


Angelo.

[3oh6]

thanks Angelo and Oliver, glad you got a kick out of it.

@ Oliver, i would love to take the heat sinks off but the way they put these on I just don't like doing it. word on the memory grape vine is they could be 0906 or 0904 based on results...or at least they clock very much like top binned 0906 or 0904 batches.

@ angelo, i would love to compare the top bin from Corsair but i don't have a kit, plus it would be comparing one kit against one kit...which shouldn't really hold much water in an argument over which is better. still comes down to luck of the draw, but you have to buy top bin to be in that game. between the two, you will likely have very similar results. remember, they are all buying the same batch's of IC's from Elpida (or are they?), they are binning them under the same or similar criteria...results are likely similar.

with that said, my 1866 Dominator-GT's will run 32M 1050MHz 7-8-7-20-1T @ 1.75v, where these ones were able to do it with 1.70v...not much of a difference really. i actually want to run the 1866 GT's through the same testing i just did with these Blades, could be interesting results considering my GT's aren't the top bin. i might just do that after i finish some GPU benching in the next few days.

[saaya]

ddr2 2000 cas6...
with 1.85v...

incredibly well done as always jody... you really put all other forum reviews and oc reports to shame

are all blade ddr3 2000 kits using the same chips though?

[Oliver]

@ Oliver, i would love to take the heat sinks off but the way they put these on I just don't like doing it. word on the memory grape vine is they could be 0906 or 0904 based on results...or at least they clock very much like top binned 0906 or 0904 batches.

Thanks for jumping in Jody , taken the question apart, there is no option to point fingers what so ever at your report, its simply outstanding , as it seems now week 48 and 49 clocks superb with low vdimm behaviour (1000 c7), great scaling and strong subs, its very interesting and funny to compare the Hyper ICs , which after now looks like to be 3 phenotypes, after what ive seen and dicussed about, however its once again important to note the differences from IMCs , boards and specific cooling.

are all blade ddr3 2000 kits using the same chips though?

Atleast the c9s are seen with Micron d9kpt/kpv and as for the ICs , they are equal or 0.05v better than handpicked jnl/jnm

Micron IC information credit all goes to Sam

[3oh6]

ddr2 2000 cas6...
with 1.85v...

incredibly well done as always jody... you really put all other forum reviews and oc reports to shame

are all blade ddr3 2000 kits using the same chips though?

thanks buddy, as Oliver said the C9 kits seem to be Micron. i just stumbled across that info today. when i saw that, i also noticed other Blade kits coming with black PCB's. the same 8 layer PCB these sticks use, but black. now i talked with a friend of the owner of said PCB company that produces these boards and he said that stock shouldn't have been the reason for going with green over black.

so my question is, do other PC3-16000 7-8-7 Blade owners have green or black PCBs? and why would OCZ go with green PCB's over the black, which appear to be identical to the green? i mean, the black just look soooo much better

[eva2000]

maybe pcb colour is just to do with what's available at the time ?

[3oh6]

maybe pcb colour is just to do with what's available at the time ?

that's the thing, apparently there is absolutely no lack of the black PCB's they are using. if OCZ ran out, it should have only taken a single day to get them more...perhaps i am just getting bad info, who knows.