The beginning of the end for DDR memory?

[nortexoid]

without reading much of the other posts...

u have to realize that only the p4 benefits from RAMBUS...not one other processor on the market does - including the athlon, duron, celeron, C3, P3, etc. etc.

i can't see DDR going away any time soon...also, with QBM coming out soon, there should be a good bump in "DDR" speeds.

 

[Radboy]

Bumpage for a great thread.

 

[Amused]

<<

<<

<< FOR THE LAST TIME.

Tom's hardware did NOT overclock the DDR Memory as the Rambus was in his test. >>

Huh? Yes, he absolutely did. He made this clear in the article right here, if you actually read it, and reiterated this fact in his posts to the community board on his web site. He clocked the P4 2.2 at 3.0GHz using the GIGABYTE GA-8IRXP motherboard configurated at 22x137MHz with 256Mb of Micron DDR333. The default DDR memory speed on P4 boards is 100MHz (200MHz effective), so he overclocked it to 137MHz to jump from 2.2GHz to 3.0GHz. >>



Obviosuly its you who do not read enough.

Many i845 boards have a 3:4 raitio and turbo features to overclock the ram bringing the ram overclock to 150+fsb.

Tom conviently used one of the few i845 boards without this feature.

But since you obviosuly need an example I will give you one. Sigh.

My best score on 3dmark 2001 is 9533 t500 265/590 i845 P4@2700

Amused one's best score on 3dmark2001 is 9498 t500 265/602 i850 P4@2400

And bofore you get any ideas it was amused who showed me how to overclock.

And if you knew anything about 3dmark you would know it likes two things. Bandwith and Video Card speed.


Amused and I both have t500s, his with a slightly greater overclock. CPU speed is not nearly as big of a factor.


So you can see now that you were wrong in your assessment because you took Tom's word. Try not to make that mistake again. >>



Actually, Tex, my 3DMark score is at 2300, not 2400

And yes, I left the RAM ratio at auto when I clocked it at 2300 giving me an FSB of 460. To get it to 2500 I had to set the ratio to 3:4, putting the FSB at 375 which SEVERELY hurt my 3DMark scrores and made them virtually the same as 4:4 at 2300.

 

[Amused]

My 3DMark scores:

All scores are with Gainward GF3 Ti500 clocked at 265/602

2000 FSB@400 (4X100) -- 8863
2200 FSB@440 (4x110) -- 9266
2300 FSB@460 (4x115) -- 9492

2500 FSB@373 (3x125) -- 9430 (unpublished at madonion, because it pissed me off and I closed it )

3DMark2001 is EXTREMELY dependant on FSB speeds and bandwidth.

 

[astroview]

Oh really Christoph83, please try to get your facts straight!

Rambus Inc., which develops memory-chip technology, was found guilty by a federal civil court jury of fraud for using information from an industry standards group to obtain patents. Rambus filed suit in September against Infineon Technologies, a German chip maker, charging patent infringement and seeking royalties for using its technology; Infineon countersued. The jury awarded Infineon $3.5 million in damages. But a United States District Court judge in Richmond, Va., reduced the jury award to $350,000, the maximum permitted in the state, and threw out Rambus's patent claims. Geoff Tate, chief executive of Rambus, said it would appeal the decision.
Source: NY Times, May 10, 2001

Rambus and Sun is also being investigated by the FTC to see whether they encouraged standards organizations to adopt new computer memory technologies covered by patents they owned but didn't disclose.
Source: WSJ, September 11, 2001

And if that latency going down with speed is true, I look forward to that! I'm not a huge fan of Rambus but if it performs better then its good in my book!

 

[jeffrey]

Rambus has lost the Pentium 4 stronghold and they will not regain it.

The only reason that DDR memory prices are approaching the level of RDRAM is because the demand for DDR memory has EXPLODED after motherboards for the P4 were offered that used it. Demand for RDRAM has fallen and demand for DDR memory has risen. Take a look at what's selling when you get your next Sunday's paper. I bet 9 out of 10 advertised P4 systems won't be RDRAM ones.

If you are going to mention PC1200 (600) RDRAM why not mention PC 3200 (200) ddr sdram?? Both of them are the same place in line, skipping PC 1066 (533) and PC 2700 (166) respectively!!

 

[Bovinicus]

If anyone impliments a dual-channel memory controller for the Pentium4 DDR SDRAM will provide greater bandwidth than RDRAM.

 

[Amused]

<< Rambus has lost the Pentium 4 stronghold and they will not regain it.

The only reason that DDR memory prices are approaching the level of RDRAM is because the demand for DDR memory has EXPLODED after motherboards for the P4 were offered that used it. >>



Really? If this is so, explain the resent sharp increase in SDRAM prices. Both SDRAM and DDR have gone up because of market situations among producers, not because of an increase in demand for DDR alone. If a rise in demand for DDR was the cause of the resent sharp increase, SDRAM would have stayed as flat as RDRAM. It did not.

I suspect RDRAM had pricing restrictions, and could not be sold at a loss like DDR and SDRAM has been for the past year.

 

[jeffrey]

"Really? If this is so, explain the resent sharp increase in SDRAM prices. Both SDRAM and DDR have gone up because of market situations among producers, not because of an increase in demand for DDR alone. If a rise in demand for DDR was the cause of the resent sharp increase, SDRAM would have stayed as flat as RDRAM. It did not."

It's rather simple to explain. SDRAM has risen because it's being phased-out. I still see $17 256mb dimms for sale on pricewatch though, so I don't see the sharp increase there. Crucial prices on SDRAM might have risen more, but that's only because as I have already stated they have moved on.

 

[Gstanfor]

I think the ram designers should get on with bringing dual ported memory to the mass market. This will do more to improve memory efficiency than rambus and DDR combined without having to significantly increase the clock rate.

The small process technologies that are readily available now mean that the extra transistors required per chip to implement the dual porting are virtually free, and the addition of a small amount of SRAM cache (8K - 16K max) to each chip would allow it to schedule the read/writes effectively and mask the latencies involved in charging up the memory cells.

Greg

 

[Degenerate]

The reason why DDR, and SD ram has risen, it is largely because that companies were not making any money as a result of competition in the technology slump. As a result, the companies raised their prices. RDRAm was not affected as its price dropped steadily instead of crashing like the DDR and SD

 

[Athlon4all]

<< If you are going to mention PC1200 (600) RDRAM why not mention PC 3200 (200) ddr sdram?? >>

Here's why. I have info from Johan at Ace's Hardware that PC2700 is suffering from MAJOR timing problems and it really seems that to me that PC2700 is the end for DDR. I know faster DDR is being used currently in graphic cards but I don't think that it will ever come to the retail market.

<< OEMs can equip budget systems with a single DDR DIMM whereas RD has to be used in pairs. >>

Again, I am not certain how big of an issue this is, and FOR THE LAST TIME, IT IS 850 THAT MAKES IT LIKE THIS

<< Also, most RD systems only have 2 RIMM sockets so upgrades require removing current memory and buying larger amounts for the upgrade, many DDR systems have 3. >>

You are wrong on both of them (at least for the P4). 850 boards feature 4 RIMM slots, and while it is true that KT266A boards have 3 DIMM slots, almost every 845-D board has only 2 DIMM slots.

<< The only reason that DDR memory prices are approaching the level of RDRAM is because the demand for DDR memory has EXPLODED after motherboards for the P4 were offered that used it. Demand for RDRAM has fallen and demand for DDR memory has risen. Take a look at what's selling when you get your next Sunday's paper. I bet 9 out of 10 advertised P4 systems won't be RDRAM ones. >>

You are correct, but just wait for the P4's fsb to increase a bit more, then the difference will be pretty big.

<< If anyone impliments a dual-channel memory controller for the Pentium4 DDR SDRAM will provide greater bandwidth than RDRAM. >>

Yes, supposedly one is coming with Intel's Garnet Bay chipset and it will do very well except in one category. Price. A dual channel DDR chipset will be so expensive to produce that RDRAM will become the "value" platform.

 

[Amused]

<< "Really? If this is so, explain the resent sharp increase in SDRAM prices. Both SDRAM and DDR have gone up because of market situations among producers, not because of an increase in demand for DDR alone. If a rise in demand for DDR was the cause of the resent sharp increase, SDRAM would have stayed as flat as RDRAM. It did not."

It's rather simple to explain. SDRAM has risen because it's being phased-out. I still see $17 256mb dimms for sale on pricewatch though, so I don't see the sharp increase there. Crucial prices on SDRAM might have risen more, but that's only because as I have already stated they have moved on. >>



What??? SDRAM is still, by far, the most widely used memory. The vast majority of computers sold today are low end systems that still use SDRAM.

And PW does not have 256 sticks for 17. 128, yes, 256, no. The cheapest 256 stick of PC 133 is $31.

And you might find cheap knock off/generic brands for $X but by and large, SDRAM prices rose the exact same percentage that DDR rose in the last two months. Oh, BTW, I can find cheap DDR memory for 29 bucks on Pricewatch, that doesn't mean I'd use it.

SDRAM has not gone the way of EDO, and wont for another two years, at least. It will have a higher demand than DDR and RDRAM because of the high volume, low end computer sales, and upgrade markets.

Your problem is you're confusing the general market with high end computer geeks. They're not. The vast majority do not upgrade their machines until forced to, and will be using SDRAM machines for another 5 years or more. Crucial's main market is people upgrading old machines, therefore their main sales right now are PC 100/133 SDRAM, and will be for another couple of years.

 

[fkloster]

Tex & Ammused...are guys saying that Asus's P4B 266 ddr solution coupled w/PC-2700 is faster than Asus's P4T-E running PC-1066?

 

[Texmaster]

<< Tex & Ammused...are guys saying that Asus's P4B 266 ddr solution coupled w/PC-2700 is faster than Asus's P4T-E running PC-1066? >>



I'm not because I like what Amused and Christoph have said, there is no telling what 1066 will do for overclocking.

The bottleneck for overclocking on the i850 is the 800rambus. 1066? Who knows?

My guess would be the rambus would outperform it

 

[Athlon4all]

<< My guess would be the rambus would outperform it >>

Well we'll find out soon enough

 

[jeffrey]

"And PW does not have 256 sticks for 17. 128, yes, 256, no. The cheapest 256 stick of PC 133 is $31."

My mistake.


<< If you are going to mention PC1200 (600) RDRAM why not mention PC 3200 (200) ddr sdram?? >>

"Here's why. I have info from Johan at Ace's Hardware that PC2700 is suffering from MAJOR timing problems and it really seems that to me that PC2700 is the end for DDR. I know faster DDR is being used currently in graphic cards but I don't think that it will ever come to the retail market."

64mb of PC3200 is used in every Xbox. We all know how close that is to being a PC mobo with integrated graphics. PC3200 is just as valid as PC1200. What is currently shipping in mass production with PC1200 RDRAM??? Didn't think so. Don't bring up PC1200 RDRAM if you can't accept PC3200 DDR sdram.


<< OEMs can equip budget systems with a single DDR DIMM whereas RD has to be used in pairs. >>

"Again, I am not certain how big of an issue this is, and FOR THE LAST TIME, IT IS 850 THAT MAKES IT LIKE THIS"

Dual-channel memory motherboards are more complex to manufacture and thus more costly. The advantage of RDRAM was that the paths were 16-bit vs. 64-bit for sdram. Now, Rambus has suggested increasing the bit path to memory to improve bandwith. There goes that advantage.


<< The only reason that DDR memory prices are approaching the level of RDRAM is because the demand for DDR memory has EXPLODED after motherboards for the P4 were offered that used it. Demand for RDRAM has fallen and demand for DDR memory has risen. Take a look at what's selling when you get your next Sunday's paper. I bet 9 out of 10 advertised P4 systems won't be RDRAM ones. >>

"You are correct, but just wait for the P4's fsb to increase a bit more, then the difference will be pretty big."

Just because the P4's fsb will increase doesn't mean that RDRAM will retake market share from DDR and sdram systems, even if it does perfrom marginally better.

 

[jeffrey]

PC3200 from Samsung has started in production.

PC3200 for the P4 on a SIS chipset.

 

[Athlon4all]

<< 64mb of PC3200 is used in every Xbox. We all know how close that is to being a PC mobo with integrated graphics. PC3200 is just as valid as PC1200. What is currently shipping in mass production with PC1200 RDRAM??? Didn't think so. Don't bring up PC1200 RDRAM if you can't accept PC3200 DDR sdram. >>

Fine. I never onced mentioned PC1200. I've been talking about PC1066 which is coming VERY soon. I do know that PC3200 is shipping in the Xbox but I still, am not certain that it's suitable for the retail market. Only time will tell. I still despite Samsung is beinging production of PC3200, I still am holding my breath. I will note however that PC1200 is very doable because RDRAM increases in clock speed MUCH easier than SDRAM/DDR.

<< Dual-channel memory motherboards are more complex to manufacture and thus more costly. The advantage of RDRAM was that the paths were 16-bit vs. 64-bit for sdram. Now, Rambus has suggested increasing the bit path to memory to improve bandwith. There goes that advantage. >>

You are absolutely correct about the costs of a Dual Channel motherboard, but it still is not as big as people think it is for RDRAM because you're only adding a 2nd 16-bit bus. As for the 32-Bit RIMM's well, you have to understand that in Intel's original plans, Tulloch, a Single channel RDRAM chipset was going to be the platform for the 32-bit RIMM's, so in the end, you would have cheaper chipset, not to mention a 32-Bit RIMM is hardly more expensive to produce than a 16-bit one (see here), but the overall cost of the motherboard decreases because only one channel needs to go from the memory slots to the MCH so in the end, 32-Bit RIMM's decrease the overall cost of the system. And in the end, the bandwidth of 850 and Tulloch+32-Bit RIMM's is exactley the same. It's a shame that Intel canceled Tulloch, but still.....

<< Just because the P4's fsb will increase doesn't mean that RDRAM will retake market share from DDR and sdram systems, even if it does perfrom marginally better. >>

Well, you are correct that the market share won't increase just because of a fsb increase, but you must keep in mind the current memory prices, I dunno if these prices will change any time soon, but if they stay then that alone is enough for it, and plus, you must keep in mind (of course you think that PC3200 will become a reality), that PC2700 is likely the last speed bump for DDR. And even if PC3200 does come, historically, SDRAM speed bumps take a very long time (PC133 was introduced in what 1999? That's 3 years between 133MHz and 166MHz (PC2700). And I will also note that DDR and PC133 devices are exactley the same, in other words, the SDRAM devices used on PC2100 DIMM's are the exact same 133MHz devices used on PC133 DIMM's. The difference lies in the interface between those devices and the chipset.) and in the mean time,the only possible way DDR can out perform RDRAM on the P4 is by a DUal Channel DDR chipset, but as you have said, Dual Channel DDR will cost so much, and if you thought Dia; Channel RDRAM costed a lot, wait til you stil, Dual Channel DDR! RDRAM will all of a sudden become the "value" platform. I really hope Intel comes back to their senses, and makes Tulloch. It would've done nothing but good, and I don't see why they ever canceled it. All they have to do for Tulloch is remove a RDRAM Channel, that's ALL! I'm not counting on it, but I am hoping that Tulloch will be launched one day.

 

[RedShirt]

<< I'll make one point. Yes the PC1200 revealed this weekend has a lower latency than current RDRAM. And yes DDR has a lower latency than DDR. However as Crucial says, lower latency is most likely unnoticible to the end user. Suppose you request a chunk of memory be sent to the processor. The first data that reaches the processor is delayed by the latency a few nanoseconds. From then on, the rest of the data comes without a latency penalty. Even with 0 latency memory, that same data chunk would only be a few nanoseconds faster. No one can tell the diffence in speed between a process that takes 1 second and a process that takes 1 second + a few nanoseconds.

Latency has an effect when want to overclock - usually lower latency memory overclocks better. But now you are talking a very small fraction of the market.

RDRAM is here to stay and so is DDR. I think PC1200 RDRAM will be great (especially since you won't have to buy in pairs anymore and the price isn't supposed to be any more than PC800). >>



I'm gonna have to dissagree. You can't just grab all the info you need from RAM in one swift swoop. You normally send a max of 32 bits of information from the Memory. This information gets stored into registers in the processor. There is hardly ANY space in those registers and info needs to be in there to preform calculations. So many variables are used in todays programs that you couldn't hope to have even 5% of your variables in the registers. There are a LOT of calls to memory to get information in programs. You can't just simply copy over all the information you are going to need with just a one time latency penalty, it happens all the time. If you can only send 32 bits (4 bytes .004k), at once, thats all you get. To get the next batch you gotta send an instruction again to retrieve some more.

Latency is a big issue, believe me. If it wasn't then why was everyone yelling at RAMBUS for its terrible latency in the past.

 

[jeffrey]

"Your problem is you're confusing the general market with high end computer geeks. They're not. The vast majority do not upgrade their machines until forced to, and will be using SDRAM machines for another 5 years or more. Crucial's main market is people upgrading old machines, therefore their main sales right now are PC 100/133 SDRAM, and will be for another couple of years."

Your problem is that you are confusing Crucial with Micron. Crucial is a division of Micron. They are only a division to sell Micron memory directly to the public. I don't think that Micron is adding any production capacity to SDRAM, but I bet that they have been expanding production of DDR sdram.

 

[figgypower]

Isn't DDR memory open technology, or at least more so than RDRAM... so, wouldn't DDR effectively get cheaper, faster?

 

[Amused]

<< "Your problem is you're confusing the general market with high end computer geeks. They're not. The vast majority do not upgrade their machines until forced to, and will be using SDRAM machines for another 5 years or more. Crucial's main market is people upgrading old machines, therefore their main sales right now are PC 100/133 SDRAM, and will be for another couple of years."

Your problem is that you are confusing Crucial with Micron. Crucial is a division of Micron. They are only a division to sell Micron memory directly to the public. I don't think that Micron is adding any production capacity to SDRAM, but I bet that they have been expanding production of DDR sdram. >>





Who said I had Micron and Crucial confused? Crucial is a major outlet of Micron memory, yes. That does not change these important facts:

1. The number one consumer memory product sold for both upgrades, and new computers is still SDRAM

2. SDRAM and DDR prices rose at the same rate over the past two months.

3. SDRAM production is not being phased out any time soon. It's still, by far, the most widely used memory.

4. The price increase has little to do with the recent incrase in demand for DDR (as any fool could figure out by the similar price increase in SDRAM), and far more to do with producers closing plants and ending a two year price war that sold memory for below cost in an effort to kill competition.

5. Finally, RDRAM prices remain virtually unchanged, I believe because producers could not sell it under cost due to stiff Rambus restrictions on production and pricing.

 

[KenAF]

Ohh my!

On an old P4 1.5GHz, moving up from PC800 RDRAM to PC1200 RDRAM increases performance in the Sysmark Office benchmark by 16.8%.

On an old P4 2.0GHz, moving up from PC800 RDRAM to PC1200 RDRAM increases performance in the Sysmark Office benchmark by 17.4%.

They seem to think that the differences between PC800 and PC1200 will be even greater at higher P4 frequencies. Also of note...it seems that most of the PC800 (400MHz) RDRAM they tested was capable of running at PC1200 (600MHz).

The Sysmark office benchmark isn't bandwidth limited, is it? Thus the improvement must be the result of PC1200's dramatically improved latency? Regardless, it looks like Intel may no longer be such a mediocre performer in office and productivity applications after it gets faster RDRAM in April.

 

[googly]

This thread has me LMAO to the extent that the Missus is concerned about my sanity. I'll pick on only one point - AluminumStudios "It's sad to say that non-leading edge performance oreinted customers (laymen) have significant influence because their numbers are great!". (cut & pasted).
What arrogance.

It's we laymen and women that create the profits that allow continued development that you lot are busy beta-testing. We want stable systems that allow us to do our daily business and get on with more relevant stuff like cooking and eating dinner.