Micron Z9 based DDR3 blows away everything!

[Reefa_Madness]

Well, at the moment it won't really make any difference to me, as I'm not diving in for a while. I'll wait for the dust to settle on X38 based motherboards first.

 

[Goonda]

Hey Reefa, sombody said you have 700 kits of RAM and over 300 motherboard/CPU sets? Are you planning to invade a country? Also, word has it that you buy up half of newegg's inventory and that is causing shortages for everybody else? Is this true?

 

[Reefa_Madness]

That total is not even close...its too low.

And Granada better stay on the alert, that's all I'm going to say about that.

 

[Reefa_Madness]

Here is a little more Z9 action, although the thread is actually about the mobo.

http://www.xtremesystems.org/forums/showthread.php?t=150210

 

[FlahsMemory]

Penryn is going to be a beast if DDR3 yields this great already. They initialliy thought it would literally take almost to the end of the year to get them that high. Now alll they need to do is reduce the voltage and they will be golden.

 

[Reefa_Madness]

Yeah, this DDR3 thing is going to clip along at a much faster rate than anyone anticipated if it continues to progress at this rate.


By the way, I found this thread and the poster is testing Super Talent DDR3-1600 CL7 Testing ( W1600UX2G7).

http://www.xtremesystems.org/forums/showthread.php?t=152027

In his post he includes the following:

In this thread I will be posting info and overclocking scores & frequencies for the new DDR3-1600 7-7-7 kit from Super Talent with the Micron Z9 chips.

It also included a pic, however, this pic is of a chip only, and in no way can you tell if, in fact, it is part of the module that he is testing. It is "Z9".

 

[orion456]

One of the best overclocks so far is a QX6700 @ 5 ghz with LN2, which is equivalent to FSB of 512 or pc8200 memory.

You can already get DDR2 pc8500 memory for $150 or less, so why do we need expensive DDR3 hyper fast memory?

Considering that hitting 4 ghz is about as high as most people can go, which is pc6400, why do we need DDR3 at all?

Considering that no Intel chip has a default speed over 3 ghz which is pc5200 that you can get for under $100, the vast majority of people won't be buying DDR3 for a very long time.

 

[orion456]

Penryn is going to be a beast if DDR3 yields this great already. They initialliy thought it would literally take almost to the end of the year to get them that high. Now alll they need to do is reduce the voltage and they will be golden.

The present FSB is already fast enough to easily saturate the fastest Conroe quad core at its highest 12x multiplier. Even if the Penryn is the predicted 45% faster, it still won't tax the FSB @ 1333 mhz.

Faster memory will not result in significant performance gain until the FSB needs a speed boost which might not happen until we get 8 cores or more.

It's not easy to see who will be able to take advantage of 8 cores and the faster memory.

 

[Reefa_Madness]

"vast majority of people won't be buying DDR3 for a very long time."

Define your "long time"...then I'm going to put in an reminder in my Outlook at work to come back and check out whether your forecast holds true.

The switch to DDR3 has nothing to do with how high you and I can overclock a CPU.

DDR3 prices will not remain at their current levels, relative to DDR2, if the past is any indicator of the future, and it usually is.

Manufacturers will be forced, in due time, to convert to DDR3, just like they were to DDR2 from DDR1. It will not be the consumer driving these changes, but instead, the consumer buying what is available. How many people buying mainstream computers are even know that there is DDR2 and not DDR1 in their PCs? The same will happen with DDR3.

Be very specific and be very cautious with your forecast, because I have a feeling that you are also going to be very wrong if you make it too long.

 

[orion456]

"vast majority of people won't be buying DDR3 for a very long time.".

I was referring to high speed ddr3.

The question is, when will ddr3 high speed main memory be in main stream Dell computers? In computer time, 3 years is a very long time, so I'd say not for at least 3 years probably longer. To be specific, dd3-2000.

Heck, I've only been on this forum for 3 years and my memory has gone from pc3700 to pc6400, not even doubled. And, at pc3700 I was hitting the ceiling where as pc6400 is overkill so far. I think its a pretty safe bet we won't need another doubling of memory in 3 years if the first doubling hasn't been necessary so far.

I think the next big speed step will be hybrid silicon hard drives to speed thing followed by all silicon hard drives.

There's two predictions for you.

 

[Reefa_Madness]

I'll see if Outlook at work lets me go out three years to make an appointment.

Now define vast majority of people, since you've refined your prediction. Will the vast majority be over 60% of Dell users...then no, it won't even happen in three years as the "vast majority" of Dell (and Compaq, HP, etc) users didn't exceed PC3200 when the rest of us were using PC4000 and beyond.

If your position is that the vast majority of overclockers won't be using DDR3-2000 in 36 months, then I suspect that you will be mistaken. It wasn't a year ago (approx release of Conroe) that DDR2-533 was the standard, with DDR2-667/800 on the high end. Today, DDR2-800 is value ram (just bought a 1 gig kit of Kingston last week for $11, after MIR), and DDR2-1066 is common place and can be had, as you previously posted, for $150, or less.

It hasn't doubled in the last year, but the standard/value ram has gone from 266 (533) to 400 (800), which is a 50% increase and the enthusiast has gone from using 333 (667) to 533 (1066) and beyond. The scale is moving to the right and its not slowing down.

I just strongly suspect that the move to DDR3 at high-end rigs will move at a pace much faster than you anticipate...and this thread, by the way, is about high-end chips and chipsets, not mainstream DELL computing.

 

[EmAn]

if things are looking the way they are at this point then i think im going to skip ddr2 altogether and have a decent amount of money for ddr3, 8800gtx sli, penryn system in a year

 

[orion456]

I just strongly suspect that the move to DDR3 at high-end rigs will move at a pace much faster than you anticipate...and this thread, by the way, is about high-end chips and chipsets, not mainstream DELL computing.

I made my prediction for Dell computers so let's concentrate on the average overclocker.

Can we agree that the majority of overclockers today are on air and they are limited by heat, purchasing power and locked CPU multipliers to less than 3.2 ghz which means ddr2-667. If so then to get to 2000 would require 3x more speed with no corresponding increase in heat (because air cooling is about as far as it can go). I think that can't be done in 3 years and the majority of overclockers won't be able to practically use ddr3-2000 in that time frame. I guess that also means that $1500 won't be able to buy a complete computer that can practically use ddr3-2000, ie: not that you can't stick ram in and crank it up to that speed but that the CPUs will gain no significant advantage (10% being significant) from such overclocking.

If it did happen then for $1500 you could buy a 9.5 ghz processor with ddr3-2000 memory. It seems unlikely.

 

[Reefa_Madness]

Actually, while your prediction was later elected to be for main stream Dell users, your first post (#27) says nothing about DELL computers, but instead talks about 5 GHz on LN2 and 4 GHz overclocked rigs not needing any ram faster than PC2-6400.

It is all about overclockers in that post. That is where you initially made your "vast majority" comment, and to which I replied.

I keep telling you that none of this is about mainstream, DELL type players, or default FSB users, and you keep trying to decry the benefits to those users...I'll try again...this is not about those users, but instead, about the guys that are trying to get the most out of their rigs, against the clock and in the framerate count.

If you really want to be honest, you don't need a Conroe, nor even an A64 for the majority of the users out in the "real world", an XP rig or a P4 will surf the net and send and read email just fine, even play back a CD or burn one. The A64, the Conroe and for that matter, the move to dual cores made no significant improvement to the family computer. It might do things quicker, but there is enough human interaction to more than offset/slow down those improvements.

You don't have to spend your money, heck, I'm not sure I'll spend mine, but the measureable numbers (the benchmarks) quantify the results and they are real...whether they are significant enough for you is another matter and one that is strictly yours.

I will send you a PM in three years and we can determine, by checking a sampling of member sigs, whether or not, the "vast majority" of OCF members are using DDR3-2000, or higher. They are, after all, the vast majority of readers of this thread.

As to whether the DELL users have DDR2 or DDR3 will be of no interest to me.

 

[orion456]

Actually, while your prediction was later elected to be for main stream Dell users, your first post (#27) says nothing about DELL computers, but instead talks about 5 GHz on LN2 and 4 GHz overclocked rigs not needing any ram faster than PC2-6400.

It is all about overclockers in that post. That is where you initially made your "vast majority" comment, and to which I replied.

In that post, in a separate paragraph I said, "Considering that no Intel chip has a default speed over 3 ghz which is pc5200 that you can get for under $100, the vast majority of people won't be buying DDR3 for a very long time." It was in the context of default speeds not overclocking that I referred to the vast majority and that would cover Dell people. That comment is not about overclocking. The entire post is a listing of where we are right now going from extreme LN2 to default speeds. The last paragraph stands by itself as each preceeding paragraph does; they are link only as a list.

None the less, I agreed to extend the the talk to overclockers in my last post. But I take it you don't agree on my definition of the average overclocker; which was some one on air ready to spend less than $1500 on a computer system.

We also have to decide the meaning of "significant increase in performance" which I proposed as a modest 10%. 10% meaning you can increase memory speed but have to decrease the multiplier in order to be stable and in doing so you will not gain more than 10%on benchmarks. Do you have an alternate definition you would like to propose?

 

[mysterfix]

Well guys I for one feel I belong to that group "The vast majority of overclockers", and I think weather most of us end up on DDR3 2000 in a few months or in three years depends on a couple of things happening.

1) Good yields from manufacturers which will drive down the price as supply goes up, so us regular folk can afford to support our habbit.

2) Will totally depend on what is currently the hot hardware. In three years we may all be switching over to flash ram or solid state dimms or something else previously unheard of.

I do agree with Reefa though, at the current rate I think prices will be dropping very fast and everyone will be able to afford an upgrade to DDR3 high speed ram.

(Please don't shoot me I'm just throwing in my 2 cents.)

 

[Goonda]

I don't see why DDR3 is getting the attention of having great yields. It's been on GPU's for years now. The process has been perfected, right? How different is GDDR3 from DDR3?

 

[traldan]

Hrm, maybe I should get the GA-P35 that supports DDR2 and DDR3...lol

 

[Reefa_Madness]

We also have to decide the meaning of "significant increase in performance" which I proposed as a modest 10%. 10% meaning you can increase memory speed but have to decrease the multiplier in order to be stable and in doing so you will not gain more than 10%on benchmarks. Do you have an alternate definition you would like to propose?

That is silly and you know it. Check sigs...for that matter...check your own sig, or request a poll and report back as to how many guys here "overclock" the way you just suggested.

Why in the world would I have to decrease the multi in order to be stable? I would venture to say that the "vast majority" overclock their cpus, as well as their ram. Free MHz, and the increased performance that comes from it, is what we're after. With the increase in the processor speeds, we can then utilize that increased memory speed. You know that, I know that and the web has it documented in a zillion places.

I still don't understand what point it is that you are trying to make, in the context of this thread. This thread is about DDR3, mated with new processors and new chipsets to achieve performance never reached before by desktop machines. I am convinced that the new rigs will outperform the current ones and by a measureable amount. It may or may not be 10%, but that doesn't matter to me. Once again, I am not concerned about whether it is cost effective, or it provides thermal improvements or electrical efficiencies or anything else along those lines.

We agree that a DELL user won't be able to tell the difference. In case you haven't noticed, there are very few DELL users here. The next group, your "average overclockers", as a general rule does not overclock the way you have suggested so I'm left here trying to figure out what motivates your posts.

You've clarified that when you posted "vast majority" you were referring to the DELL group. I countered that I believe that the majority of the DELL group will be using DDR3 in 3 years because that will be their only choice.

You don't think that there is any benefit to DDR3-2000 today and there won't be in 3 years either. Fine, that is your opinion, but since you can't see the future and neither can I, lets just agree to come back in three years and compare notes because otherwise, this is going nowhere.

We are on two different planes here and they don't intersect at any point.

 

[orion456]

We also have to decide the meaning of "significant increase in performance" which I proposed as a modest 10%. 10% meaning you can increase memory speed but have to decrease the multiplier in order to be stable and in doing so you will not gain more than 10%on benchmarks. Do you have an alternate definition you would like to propose?

That is silly and you know it. .

The process of overclocking is often:

1. increase multi as high as possible.

2. increase fsb until overclocking fails stability

3. drop multi by one, increase fsb until overclocking fails

4. repeat until you reach the highest fsb that is possible.

That was the process I was trying to encapsulate in my one sentence. If you have a different process, let me know. My point is that under step 3 you can often increase the fsb by lowering the multi, but you gain nothing in terms of benchmarks on conroe chips. I wanted a 10% gain from faster memory upon lowering the multi or it isn't regarded as significant. The only reason I become this specific was your request that I be careful and specific.