OCZ Brings Out 600MHz DDR SDRAM Memory

[Cole]

:banana: ooooow .... is all I can say !

http://www.xbitlabs.com/articles/editorial/display/cebit2004-4.html

OCZ Technology Group has been seriously innovating the memory market recently. Within the last couple of month the firm unveiled Enhanced Bandwidth technology as well as rolled-out Ultra Low Noise 2 print circuit boards for memory modules. Right now the company is showcasing DDR SDRAM memory modules functioning at 600MHz – a truly unbelievable speed for DDR and much higher than DDR2 is capable of achieving right now.

[Mrstickinit]

Holy Shietballs!!! Dead link though :(

Maybe they can do 2-3-3-6 at 500Mhz...... or even better.

[JCviggen]

wouldnt PC5000 be DDR625 instead of DDR600 ?

Its nice I suppose, if the price is low enough...at the end of the day its still high latency stuff...and your mainboards needs to be capable of reaching 300 FSB for it too.

The "innovation" part seems kind of rediculous however...hand-picking sticks and slapping a heatspreader onto them isnt that difficult

[Mrstickinit]

That's why I said it would be better to run it at 250 (500Mhz) with tighter timings. There's not many peeps that can say they've ran their boards at 300Mhz FSB.

[nailbomb]

The "innovation" part seems kind of rediculous however...hand-picking sticks and slapping a heatspreader onto them isnt that difficult

Jc, isn't there a little bit more to it than just handpicking sticks. Stuff like the design of the PCB, voltage regulation, etc?

[JCviggen]

Originally posted by nailbomb
Jc, isn't there a little bit more to it than just handpicking sticks. Stuff like the design of the PCB, voltage regulation, etc?


To be honest, i dont think so...mostly fancy names. I could be wrong, but i have had some really cheap PC4000 from Taiwan which ASUS sent me and it nearly hit 300 as well, random sample.

[Cole]

Originally posted by JCviggen
and your mainboards needs to be capable of reaching 300 FSB for it too.

That was no problem for my old Abit IC7 and P4 2.4C ... so I for one would have loved to have some :p:

[Soulburner]

This should be labeled PC4800...

Damn the marketing! But they're good at it, i'll give em that.

[SupaMan]

right you are, this should certainly be labeled PC4800

[saaya]

Originally posted by Soulburner
This should be labeled PC4800...

Damn the marketing! But they're good at it, i'll give em that.

yepp!

[Spec3]

HOLY!! I gotta get me some to play with :D

[Tomsawyer]

Cough Coughing up some serious bread for that I bet lol

[OPPAINTER]

If it runs 300 stock then I would think it could do 320 maybe, who knows.
But I think 320 even at junk timings should do pretty good against 250 agressive.
Of course your probably limited to P4C-800, if the mobo can even do 320 1:1. I know it can do 5:4 past 320.

OPP

[SupaMan]

Originally posted by Soulburner
This should be labeled PC4800...

Damn the marketing! But they're good at it, i'll give em that.

i have to ask though, what makes their marketing good?

[lalPOOO]

OPP has a point, OCZ usually leave about 16mhz for overclockers, I think either ryan or tony said that on here a while back.

I can't help but think that 50% of the sales on this ram will come from uneducated babons who say "shopkeep, bring to me the fastest ram you have" and the guy says, "well we have this pc5000 stuff" "WOW! that will run so fast in my pc, I want some of that!" Someone doomed to never ever find out that just because the ram can do 600mhz doesn't mean it is. Thats not really ocz's fault though.

SupaMan: Ocz's marketing is good because they're EVERYWHERE, with ads and other stuff.

[faruquehabib]

what are the stock voltage requirements for it?? 2.8v?

[Slickthellama]

Originally posted by Mrstickinit
Maybe they can do 2-3-3-6 at 500Mhz...... or even better.

doubt it, high mhz rated memory typically does not like tight timings at any speed. They are designed for raw MHZ not timings.

[lalPOOO]

They've gotta be 2.8v and I don't see why JC is saying this isn't innovation, I sure as hell can't run my pc4000@300mhz@2.8v and I don't know of any other memory that can. Its probably not just handpicked chips, probably some stuff going on with the pcd design/components.

[Naughtyboy]

Well I Can run my A-Data PC4000 @ 300mhz to with 3.0vdimm....so even cheapo ram can do thoose speeds.

But my conserns is this friggin on going damne BS about "tight timings is the best no matter what"...well Newsflash for ya guys.....I don´t care if god him self says tight timings is the best...thats a friggin lie...nothing else.

Tight timings isn´t always the best...it all depends on what you are using your rigg for. Also...if you run at speeds high enough, there is a point where the amount of of mhz on memory out performs tight timings...and BASTA....thats it.

So stop running around spreading BS like tight timings IS DA BOMB when people ask.....instead ask what they are using ther computers for...what kind of appz and such.

[blinky]

Originally posted by [NH]Naughtyboy
Tight timings isn´t always the best...it all depends on what you are using your rigg for. Also...if you run at speeds high enough, there is a point where the amount of of mhz on memory out performs tight timings...and BASTA....thats it.
amen, brotha

[***Deimos***]

For those of you that haven't yet gone to the link posted at the beginning of the thread, according to xbitlabs coverage of GDC, the DDR600 (300Mhz) is made possible by their fancy ultra low noise design. However, they are noting that it is getting more difficult to go higher because of the noise in power signal from motherboard.

You can tell that i865/i875 boards have really good inductance/capacitance matching, minimal clock skew and good grounding planes to decrease the crosstalk since people can hit well over 350 fsb and over 280 on the memory. However, I dare to imagine that if AMD goes ahead with DDR500 support, and "over-engineers" for Mhz like Intel, then we may very well see DDR650 (325Mhz) - after all the memory controller is on chip and made on 13nm SOI. However, that would require much stricter specs... meaning, OCZ/Corsair/Kingston would have to do their fair share to make sure all the signals are crystal clean.

[macci]

Tight timings isn´t always the best...it all depends on what you are using your rigg for. Also...if you run at speeds high enough, there is a point where the amount of of mhz on memory out performs tight timings...and BASTA....thats it.
Can you give some examples where 3-4-4-8 outperforms 2-2-2-5 (other than 'buffered' sandra)?

[FUGGER]

I have several 300 5:4 2-2-2-5 and 300 1:1 SPD scores posted.

Ill check my scores and unhide the 5:4 and 1:1

Nice stuff OCZ

[Naughtyboy]

Originally posted by macci
Can you give some examples where 3-4-4-8 outperforms 2-2-2-5 (other than 'buffered' sandra)?

Yes....when I´m doing video conversion....atleast on my rigg I get ahole lot more performance losing the timings and uping the fsb 1:1.

Sorry for not being able to show it to you in numbers...atm I don´t have any...but as soon as I can i´ll give´em to ya

I can understand that in your world of 3dmarking timings is everything..no ofense....but sorry my friend.....if you are saying that low timigns is best no matter what...then with all due respect in the world...you are simply lieing.

There are appz...and realworld situations where appz like to strech out in high memory bandwidht...instead of getting acces to the info "fast".....

[st0nedpenguin]

Never mind the RAM :eek: (http://www.xbitlabs.com/articles/editorial/display/cebit2004-4_2.html)

During the whole CeBIT exhibition I noticed plethora of mainboards based on the newest Grantsdale (i915P/i915G) and Alderwood (i925X) chipsets with corrupted Socket T. There were so many damaged Sockets for Intel’s LGA775 processors around that I asked from a mainboard maker about this issue. Apparently, the problem with pins in Socket T is real and is pretty nasty.

After a number of processor installations tiny pins bend and the whole mainboard becomes useless. This is certainly not a problem for PCs that are never upgraded, but users who modernize their computers should be very careful with installation of new CPUs. Furthermore, system integrators should probably get themselves prepared for pretty tough microprocessor installation. The issue should also be extremely horrible for test laboratories, where plethora of chips may be changed on a single mainboard per single day.

There are mainboard makers – ASUSTeK, Soltek, Jetway, etc – who tend to solve the problem rather radically – they install Socket 478 on their Grantsdale mainboards. This certainly limits upgrade options for particular products, but undisputable solves the problem of damaged pins. Loads of mainboard makers are not satisfied with Socket T, as every such socket costs them about $7-8 and brings additional issues, including those with warranty.

Besides, there is a rumour going around the show that mainboard makers are also not happy with DDR2 and increased complexity of PCBs it brings.

http://www.xbitlabs.com/images/editorial/cebit2004-4/lga775_damaged_big.jpg

Way to go Intel...

[Atilaptops]

This has got to be one of the most debated issues. My personal opinion is you better have 2-2-2-5(or 11 as it may happen) for an fsb of 260 or lower, maybe 270. But if you're gonna go to 300 then loose timings is ok. It that wierd 270-300 range that's sort of wierd. Then again AMD systems almost always perform beteter with tighter timings and they have to run 1:1. I hope to be running 250 fsb in DC with my RAM soon. I think that is about as good as it gets before uber timings become near impossible. Of course raw MHz do win, otherwise DDR-2 will suck.

Btw, do you gain more performance if you run your RAM 2:3, say on a 2.8e? Just wondering?

[QuadDamage]

sorry to say that but tRCD @ 4 stuff will never touch LL ram, no way. tRP @ 4 will only worsen things. Cas and tRAS isn't that important though.
i can't remember exactly since it's been ages since i last touched slack ram but there's a 10fps diffrence in Lobby between tRCD @ 2 & 3 not to mention 4.

[pkrew]

If you're around 300fsb then I'd say it would be pretty close, but everything that I've read shows the 5/4 tight is slightly better. For that matter who knows before you start to overclock if you'll hit around 300. IMHO its better to get some LL ram if you can find some. Then you're good to go for most situations. The only posible exception would be someone who's not into volt modding their mb.

[STEvil]

The socket T thing has been known for a while... even the pics on toms had bent pins, lol!

[saaya]

ocz probably DID improve the pcb etc, but who knows how big of an impact it has?

ive heard of a lot of memory that hits 300mhz with those timings and ~2.8v

didnt ocz say they are working on a way to keep trcp low even at high speeds? maybe it does run 300 3-3-4-4 or even 3-2-4-4 with enough volts...

[Soulburner]

Originally posted by QuadDamage
sorry to say that but tRCD @ 4 stuff will never touch LL ram, no way. tRP @ 4 will only worsen things. Cas and tRAS isn't that important though.
i can't remember exactly since it's been ages since i last touched slack ram but there's a 10fps diffrence in Lobby between tRCD @ 2 & 3 not to mention 4.
Exactly, when it comes to most applications, especially gaming situations, even memory running a divider like 5:4 at 2-2-2-5 is faster than memory at the same FSB 1:1 3-4-4-8, ALWAYS, at ANY speed. 2.5-3-3-7 1:1 is pretty close but it needs to be clocked pretty high.

To the guy that said otherwise, I would like to see some benchmarks published proving it.

[lalPOOO]

Originally posted by [NH]Naughtyboy
Yes....when I´m doing video conversion....atleast on my rigg I get ahole lot more performance losing the timings and uping the fsb 1:1.

Sorry for not being able to show it to you in numbers...atm I don´t have any...but as soon as I can i´ll give´em to ya

I can understand that in your world of 3dmarking timings is everything..no ofense....but sorry my friend.....if you are saying that low timigns is best no matter what...then with all due respect in the world...you are simply lieing.

There are appz...and realworld situations where appz like to strech out in high memory bandwidht...instead of getting acces to the info "fast".....

When you compared the speed difference was that with a cpu going the same speed or different?

edit: and someone needs to come up with screenshots or proof of people hitting 300mhz at 2.8v (not calling you a liar saaya, just saying that I don't think that any other ram will do it)

[saaya]

afaik a lot of hynix 4ns chips get there

http://www.modlabs.net/index.php?location=articles&url=ddr500

in this roundup it looks like most memory only reaches around 270mhz with 2.8v, but the latest hynix chips clock better... since ocz uses hand picked chips im not surprised to see them running 300 with 2.8v, i think those hand picked chips would run at least 295 on every pcb with the same vdimm and vref... i really dont think the pcb makes a big diference, there are bh5 sticks with crappy pcb that beat high quality pcb sticks with bh5... and i dont think that those bh5 chips would oc even higher on a better pcb

[QuadDamage]

lalPOOO,

actually Geil PC4000 with that heavy silver heatspreader will do 300fsb 1:1 at 2.8V but the bad news is it doesn't like volts 2.9+ and 295-300fsb is the max depending on stick.

[nailbomb]

The Geil is Hynix D43?

[saaya]

yepp it is, almost all memory above 3500 is hynix d43, as theres no other memory that will go that high ,except for kingmax, twinmos and samsung.

samsung has just released 3700 memory but it needs more volts, and twinmos and kingmax chips dont clock as high as hynix so most manufactueres choose hynix as its also the cheapest. but twinmos and especially the new kingmax memory can run much better timings (2-2-2 at ~200/2-3-3 at ~233/2.5-3-3 at~250)

i for myself cant wait for the new kingmax memory that can run 200+ 2-2-2 and 250+ 2.5-3-3 with only 2.5v!!!

[nailbomb]

samsung has just released 3700 memory but it needs more volts

Jah, did you see the Anadtech article? That was interesting.

Are you talking about Kingmax's BGA stuff, or what?

[saaya]

kingmax DDR466 and DDR500 EXTREME memory with tiny bga chips. i asked them and they said its released already but theres no shop that sells them so far... i asked them where i can order some and they didnt reply :/

anandtech reviewed the stuff and it looks extremely promising. this stuff could be what bh4 would have been like.

[Naughtyboy]

Originally posted by lalPOOO When you compared the speed difference was that with a cpu going the same speed or different?

Basicly yes..it was on a P4 ES.



And guys I´m not saying that saying that 2225 isn´t good.....for gaming and 3D it IS the way to go...hell...I got my self 1gb of LL ram for my gaming rigg.

What I´m saying is that it isn´t ALWAYS the way to go....there are appz and times when losen the timings a bit and get uping the mhz will be more useful.

[QuadDamage]

yeah in PCMark :D

[Geforce4ti4200]

"The only posible exception would be someone who's not into volt modding their mb."


he hit it on the nail. vmodding a64s are very risky, my friend says use more than 3 volts and you risk taking out the ondie memory controller, bh5 ram is quite useless at below 3 volts too, getting about 220fsb at <3 volts. What if you had like a 2.4c at 3.6 or an a64 3200+ at 3GHz? youd want that ram for 300fsb. Just use 2.8v and you are in business! That ram probably still costs less than bh5 too. It is my experience that from 2-2-2-11 to 2.5-3-3-11 is worth about 25fsb so 3-4-4-11 at 300fsb is equal to 2-2-2-11 at 250fsb. for athlon xp rigs, its useless cause nf7-s and dfi mobos dont usually like past 250fsb anyway and the dfi comes with 3.3v so bh5/6 is the way to go. but for p4c or athlon a64 you want high fsb at 2.8v

[MightyOne]

@saaya

The Kingmax DDR500 Hardcore Chips are available at www.overclock.co.uk for 264€ (1GB Dual Channel) - I think I'll send my crappy Kingston PC3000 back and get those ... might be a little overkill as my NF7/DLT3C Combo does only 225MHz FSB stable w/o mods, but then again: I'll do the VDD, VDIMM and L12 Mod soon, and then I'll see whether those RAMs really like the volts! :D

[Crankster]

Geforce4ti4200, not really... The thing is you have to adjust the voltage on the pot, with the computer turned off of cource. If you change it in the bios, it is hen your risk your precious chippie

[macci]

Lets see some 3-4-4-8 ram do over 4GB/s unbuffered (http://www.akiba-pc.com/e107_images/articles/30c/memsettings.gif)? :D

252x15 1:1 2-2-2-5:
http://www.akiba-pc.com/e107_images/articles/30c/sandra_252_unbuf_1024.gif

200x15 1:1 2-2-2-5:
http://www.akiba-pc.com/e107_images/articles/30c/200x15_unbuf_1024.gif

[QuadDamage]

nice. End of discussion i guess:D

[Naughtyboy]

Originally posted by macci
[B]Lets see some 3-4-4-8 ram do over 4GB/s unbuffered

I havent got any high FSB system or cpu atm....but Sami ya know I´m gonna give it a go.... ;) :)

Also why are you guys always using 3-4-4-8 as measurement for lose timigns...2,5-3-3-7 is also "slower" than 2-5-2-2 but still whole lot faster than 3-4-4-8.
I never said anything about wich timing to use...thats from situation to situation....but ofcourse as tight as possible at certain speed.

Kickazz scores btw..... :banana: :banana:

[flapperhead]

i wonder if thy're using the new hynix d5 or the newest 550 stuff hynix will be realeasing to the public in april..

[Tony]

Originally posted by macci
Lets see some 3-4-4-8 ram do over 4GB/s unbuffered (http://www.akiba-pc.com/e107_images/articles/30c/memsettings.gif)? :D

252x15 1:1 2-2-2-5:
http://www.akiba-pc.com/e107_images/articles/30c/sandra_252_unbuf_1024.gif

200x15 1:1 2-2-2-5:
http://www.akiba-pc.com/e107_images/articles/30c/200x15_unbuf_1024.gif

You need to talk to Radelon Macci..he has posted benches higher than that with D43 at 293fsb.

FOR ALL

The Kingmax is a PCS die and the yield is up and down..we have looked at them and will do again.

Winbond have packed it in now..so if you have BH or good CH keep onto it.

OCZ are constantly looking to improve latency at high FSB's..we have 3-3-2- at ddr500 now and ddr533 soon.

We are also looking at a ddr650 IC...:)..LOL

[Geforce4ti4200]

wow 2-3-3-6 at ddr500 from OCz now? nice!!!!!!!

[Penti]

Originally posted by bigtoe
You need to talk to Radelon Macci..he has posted benches higher than that with D43 at 293fsb.


Here is his thread. 4GB/s Unbuffered at 292.3MHz (http://www.xtremesystems.org/showthread.php?s=&threadid=17803&highlight=sandra)

Pretty impressive for 2.85V.

[macci]

You need to talk to Radelon Macci..he has posted benches higher than that with D43 at 293fsb.
MAn your right - its a total of 1MB/s faster overall (ALU+FPU) :D
I think I'll need to find some 3-4-4-8 RAM for myself..
ANy idea where I might find some good stuff?? ;)

[Soulburner]

At 293fsb.

If you notice, macci wasn't running anywhere near 293fsb ;)

[zakelwe]

I really don't like the increases in latency manufacturers continue to market with ever bigger numbers, soon we'll have CAS4 and just to make it sound good a new improved " buy me " style name, like, say DDR2 .

;)

Seriously though , if company's are trying to get the latencies lower as mentioned above then good luck OCZ.

At the moment I am sticking to my one reminaing 250+Mhz 2-5-2-2- stick of bh5 and hope I don't kill it and seeing what goes on.

270 at 2-6-3-3 and around 2.95-3.2v might be a nice target, any chance ?

Regards
Andy

[pkrew]

True, he was only at 252 fsb and I'm sure he can get a lot more out of it. How do you run Sandra unbuffered?

[macci]

http://www.akiba-pc.com/e107_images/articles/30c/memsettings.gif
disable those settings

[saaya]

Originally posted by MightyOne
@saaya

The Kingmax DDR500 Hardcore Chips are available at www.overclock.co.uk for 264€ (1GB Dual Channel) - I think I'll send my crappy Kingston PC3000 back and get those ... might be a little overkill as my NF7/DLT3C Combo does only 225MHz FSB stable w/o mods, but then again: I'll do the VDD, VDIMM and L12 Mod soon, and then I'll see whether those RAMs really like the volts! :D

270€ :eek: well then again its pretty much the best memory available after good bh5.

please pm me and let me know how it does!

[pkrew]

Wow, all I could manage was 3710/3736 and that was using 264 1/1. Very impressive. Was that due to using 1G vs 512mb in the memory. Either way you're a magician. I need some of your skills.:toast:

[lutjens]

Actual rated 300 MHz memory......:slobber: :slobber: :slobber:

I like...:)

[Tony]

Macci

He has seen higher than that posted bench, as have I.Thing is i play with so many boards now etc i hardly ever keep the benches.I did run 4400Beta at 303fsb 2.85V and that returned 4150+ but i didn't keep the bench..sorry

Regarding lower latency..3-2-3-5 at ddr500 is what i can run with the 3500Platinum EB...Yield is improving also and we hope to see the same at 533 soon.Not bad for ram designed to be a CH5 replacement..

[macci]

pkrew, extra mem bank and boot up FSB make the difference ;)
on AI7 board I hit 3709/3746 (http://www.akiba-pc.com/ai7/ai7_232unbuf.gif) at 232FSB 1:1 (4x256MB)

[macci]

bigtoe, great numbers :)
check your pm :D

[pkrew]

Originally posted by macci
pkrew, extra mem bank and boot up FSB make the difference ;)
on AI7 board I hit 3709/3746 (http://www.akiba-pc.com/ai7/ai7_232unbuf.gif) at 232FSB 1:1 (4x256MB)


Thanks for the tip. I only have 3 sticks that don't have ram sinks glued to them, but I'll try and get four together in my AI7 and give it a try.

[Soulburner]

Originally posted by macci
pkrew, extra mem bank and boot up FSB make the difference ;)
on AI7 board I hit 3709/3746 (http://www.akiba-pc.com/ai7/ai7_232unbuf.gif) at 232FSB 1:1 (4x256MB)
I've noticed this too, but on my ASUS all it takes for that extra boost is using slots 2/4 instead of 1/3...I don't even need to use all 4.

It gives higher Bandwidth in Sandra, but all other applications are unchanged.

[pkrew]

Well, tried it with my Asus and was able to boot into windows at 255. Unfortunately PAT was disabled. I tried clock gen but I couldn't keep it from increasing the AGP clock so it was no go. I was able to get 3850/3876 at 255 on the P4C800. I also tried it on the AI7. I don't have a modified PSU on that, but was able to get to 237 at 3.2v and a score of around 3800. I couldn't get clock gen to work and increasing the fsb using uguru didn't seem to change the Sandra score. I just booted at 237. It seemed that PAT was enabled regardless.

[Soulburner]

pkrew, PAT was not disabled or your unbuffered would have been much lower I think.

CPU-Z will show disabled but it is not disabled, it is a bug.

[pkrew]

That maybe true, but it was enabled when I only had two sticks in and it also says it disabled using memtest. The other thing is I got almost the same score with only 237 on the AI7 and cpu-z reported PAT enabled. But you're right the scores were pretty high.

BTW, is there a program that I can use like clock gen that wont raise the AGP freq on the P4C

Edit: The other thing is that cpu-z reported PAT enabled on the P4C with 4 sticks when I set it to 200fsb. When I went over 200fsb with four stick it reported disabled.

[LikwidKool]

hey bigtoe that pc3500 EB was doing 2-3-2-6 at what volts? I am liking what I see. Would love to get my hands on some! Too poor right now from buying bh-5!!

[Soulburner]

I could be wrong then...but I have seen this tested before.

All 4 sticks would make CPU-Z say disabled, but benchmarks remained the same...(except Sandra).