Posted by
ryuzaki0
on from the finally-moving-forward dept.
jandrese writes "According to Cnet, Intel is finally getting around to supporting DDR SDRAM in their P4 chipsets. This is a good move on Intel's part, as they need to bring the cost of their P4 based systems down to compete with AMD, and moving away from Rambus is a good start."
Isn't it too late to worry about this?
by
ergo98
·
· Score: 5, Insightful
Seriously when RAMBUS was 10x the price of SDRAM it seriously hurt, but now that RAMBUS is getting close to comparable, I don't see what the point is. In my neck of the woods PC-800 RDRAM goes for about 30% more than PC2100 DDR, which really isn't that much (and dual channel RDRAM is the fastest RAM platform out there). Given that the P4s one redeeming factor is that with RDRAM it has a serious memory advantage, I really don't see what Intel is thinking: Put a P4 with DDR DRAM and it'll get clobbered even more.
Re:Isn't it too late to worry about this?
by
Cloud+9
·
· Score: 5, Insightful
(and dual channel RDRAM is the fastest RAM platform out there).
Speed isn't everything... RDRAM has a great deal more latency than DDR. In many cases, RDRAM performs significantly worse than even SDRAM.
Besides that, there's the evil factor, considering Rambus believed more in the policy of suing for royalties as a revenue model instead of producing and selling a decent product.
-- Karma: Dyn-o-mite!(mostly affected by Jimmy Walker reading your comments)
Re:Isn't it too late to worry about this?
by
$carab
·
· Score: 3, Interesting
Interesting Point, but the price of DDR RAM is still to expensive to replace SDRAM right now. However, once the Northwood gets released in a few months, OEMs may move to DDR because of the sheer volume of DDR that is being used. DDR RAM, however is still more expensive; Crucial sells 256 Megs for 50 bucks, compared w/ 30 for some PC133...But I think thats artificial Christmas Price Inflation (TM).
In terms of high end Intel systems, DDR just isn't that way to go. A couple of months ago, when Intel got to 2GHz, they were beasting on similar Athlon systems. But now, AMD has gone on a tear, heavily ramping up their Palomino core. A 1.9 XP w/ DDR beats a 2GHz P4 w/ PC800 RDRAM in every category except for memory bandwith...If the Intel was using DDR as well, the Xp's margin of victory will be even greater. The P4 relies on fast, fast memory. Give a P4 slow memory, and it will freeze (P4 and SDRAM is a horrid combo). Since RAMBUS will soon be releasing pumped 133 MHz bus memory, I think this is the memory that will help Intel more than DDR. Intel is losing, and has ALWAYS lost, the price battle. I think that if Intel cuts memory performance to reduce price, they are losing their ONLY advantage over AMD systems.
Re:Isn't it too late to worry about this?
by
nusuth
·
· Score: 3, Informative
My comments were only applicable to programs that stress memory subsystem, otherwise you won't see much of a difference between sdram and rdram anyway. So I didn't (at least didn't intend to) claim that all (even most) apps should be tuned for optimal memory performance.
I don't agree with you comment about context switching emphasising latency. Context switching takes doesn't happen frequently, time slices are too long from a cpu perspective (linux defaults to 10ms, I guess that corresponds to ten million or so cycles.) Contexts are loaded in burst rates so latency would have little effect on overall performance. Offcourse it might be that context switching occurs more frequently because programs are releasing their time slices, but that would mean you don't need performance either since your processes are idle. A busy process still gets its ten million cycles, wasting a tiny percentage on context switching - high latency or not.
Its almost 4am here, I'd better sleep now.
--
Gentlemen, you can't fight in here, this is the War Room!
Re:Isn't it too late to worry about this?
by
scriber
·
· Score: 5, Insightful
The guys at Intel aren't stupid. They invested millions into marketing themselves as a consumer brand before it was cool to do so, and it's been the best move they ever made. AMD, for all its trying, hasn't even registered on the radar of most consumers. Intel uses this to its advantage to charge a hefty price premium. After all, they're the Coca-cola while AMD is just the RC.
So Intel _is_ winning the price battle, since the winner of the price battle is the guy that gets to charge more money and still sell 80% of the processors, not the guy that sells them for half as much to push enough volume to break even. After all, Intel could sell their parts for much less than they do without actually losing money, but they don't have to.
And, as AMD's recent relabeling of their XP line has shown, clock speed is still king. Nobody has ever successfully dethroned it as the single number consumers care about above all others. Which is why Intel has won that battle as well.
P4's with DDR aren't in any way related to RAMBUS's performance as much as keeping low-margin systems affordable and still fast. That's why you'll see P4 Rambus and DDR boards out there, fighting it out for the price/performance sweet spot.
At this point, Intel is more worried about Sun than AMD, since Sun is the lone vendor not committed to Itanium/McKinley. They've also got the high-margin Xeon processors competing with Sun's mid-range offerings. This is where the interesting things are going to start happening, but you won't be hearing about it on Tom's Hardware.
Stealth Introduction?
by
LostScorp88
·
· Score: 3, Redundant
A "stealth introduction"? Doesn't Intel know about Slashdot? We geeks find out EVERYTHING, and as soon as one of us finds out and submits it to Slashdot, everyone knows. With media the way it is and Slashdot here, a quiet introduction is virtually impossible.
-------------------
DDR has better latency
by
nusuth
·
· Score: 3, Interesting
And it is latency that usually counts, peak transfer rate is not sustainable anyway. A dual channel DDR-SDRAM platform would be faster than dual channel RDRAM platform, single channel ddrs are already competing with dual channel rdrams. With 166*2MHz DDRs on the horizon, I think it is a very sensible solution.
--
Gentlemen, you can't fight in here, this is the War Room!
Rambus - now even more obsolete!
by
nyquist_theorem
·
· Score: 3, Interesting
It's unfortunate that Rambus RIMMs are even more obsolete now that Intel's new chipsets are going to DDR. Not that I'll miss them, but its always unfortunate when early adopters get hosed with proprietary hardware. Anyone remember the Socket 4 P60/P66? Drop a ton of money on the new Pentium, and watch while everyone who waited is able to upgrade while you can't. Ditto to those with Asus P4Ts - not only are they hosed on processor options because intel changed pinouts for the new P4s, but now the RAM is obsolete too.
From the article: Intel is planning the stealth introduction of a chipset that will let computer makers connect the Pentium 4 to speedy DDR (double data rate) memory.
Speedy? Isn't DDR-SDRAM slower than RDRAM? Sounds a bit fluffy to me. What they really mean, but don't clearly spell out, is that DDR is faster than the normal SDRAM the 845 supports. But its still no RDRAM. Which I guess everyone here knew anyways.
Ahh well, I'm just grumpy b/c I convinced my mom to buy a P4T/Rambus-based P4 1.7Ghz, and now I have to ditch the Ram/Mobo/CPU to upgrade it. (I'd have given her an Athlon but the dustbunnies at her place are such that I'd be afraid of her burning the place down... remember that THG vid of the flaming Athlons?)
-- --
"Ignorance more frequently begets confidence than does knowledge." (Charles Darwin)
Re:Rambus - now even more obsolete!
by
RelliK
·
· Score: 4, Troll
(*) DDR and Rambus transfer data at both the rising and falling edge of the clock cycle, thus doubling the effective bandwidth. Bacause of that they are often reffered to as 200MHz, 266MHz, and 800MHz respectively.
Anyway, the point is that DDR has greater bandwidth than Rambus. On top of that, Rambus has a pathetically high latency. Because of that Pentium 3 systems with PC133 SDRAM outperformed their Rambus counterparts most of the time.
As an aside, Intel decided to "fix" this flaw by making Pentium 4 waste four times as much memory bandwidth as Pentium 3 -- that makes P4 highly sensitive to memory bandwidth./. linked to a very interesting article discussing the P4 architecture a while ago.
Back to the point, Pentium 4 chipset uses two channels of Rambus memory that work in parallel. That gives it 2 * 1600 = 3200MB/s of bandwidth, which is greater than a single channel of PC2100 DDR (though it still has high latency). Problem is that you need to install memory in pairs (on RIMM for each channel), and each RDRAM channel can have only two memory slots. That means you are only one upgrade away from maxing out your memory. On the contrast, each DDR channel can have up to 4 memory slots and you can upgrade one slot at a time.
Also note that NVidia Nforce does the same thing for Athlons & DDR as the P4 chipset. Of course two channels of DDR have the bandwidth of 4256MB/s).
Ahh well, I'm just grumpy b/c I convinced my mom to buy a P4T/Rambus-based P4 1.7Ghz, and now I have to ditch the Ram/Mobo/CPU to upgrade it.
I'm so tempted to say "I told you so" -- which I would have if we ever spoke.
(I'd have given her an Athlon but the dustbunnies at her place are such that I'd be afraid of her burning the place down... remember that THG vid of the flaming Athlons?)
This is about the stupidest thing I ever heard. It's like claiming that Ford makes unsafe cars because their engines fry when you drain all the oil from them. Try this: remove the heatsink and fan from your P4 and see how long it takes for it to catch fire.
-- ___
If you think big enough, you'll never have to do it.
Initial Designs
by
Murdock037
·
· Score: 3, Informative
As we all know, the P4 is designed to take advantage of high memory bandwidth-- which is why the 845 chipset (or whatever) that used SDRAM was such a flop.
When the P4 was first introduced, Intel claimed that it was designed specifically for RDRAM. If this is true-- and I suppose it doesn't have to be-- then is it possible that the new DDR stuff will actually perform below RDRAM systems? Is the only advantage going to be price?
I'm not really a tech guy, so this is an honest question. I'm not a Rambus fan-- I've got a PIII with the 820 chipset, and I'm not particularly fond of it-- but could it be that the company that everybody hates is actually the better way to go in this case?
Of course, everybody around here is going to be be gushing over DDR over Rambus-- if they choose Intel over AMD, which doesn't seem likely-- but it seems that Intel is either stepping backwards or conflicting with their past words.
Re:Initial Designs
by
WasterDave
·
· Score: 3, Interesting
The P4 was designed for truly immense memory bandwidth and very large on chip cache. The current generation of P4 is suffering from the cache being significantly below it's design point, which is somewhere around the 2Mb mark. Obviously with a cache this big the latency of the ram itself isn't such a problem - hence Intel signing up to RDRAM.
Anyway, the actual question was:
but could it be that the company that everybody hates is actually the better way to go in this case?
Not really. There's not that much difference in bandwidth between DDR and RDRAM anyway. And Rambus need to go broke to remind the industry in general that we won't tolerate that kind of behaviour. Unfortunately they have some huge contracts, the PS2 being probably the biggest, so it seems unlikely they are going to go chapter 11 in the near future.
Re:Patents kill your tech off!
by
MrResistor
·
· Score: 4, Interesting
The patent drove the cost of DDR RAM up so much relative to competing technologies that the tech died.
I'm hoping you meant to say RDRAM.
Anyway, the patents had nothing to do with the price differences between RDRAM and DDR SDRAM, it was all due to manufacturing costs. I remember a little over a year ago Kingston was bragging about their 30%(!) yield on PC-800 RDRAM chips. When 70+% of your product doesn't pass QA, that's definately going to drive your costs up! Additionally, manufacturers had a fair amount of retooling to do before they could make RDRAM, and high setup costs get passed on to the consumer. As I recall, RDRAM also has a bigger die size than DDR SDRAM (I could easily be wrong, it's been a while since I cared) which would also drive up costs.
In contrast, DDR SDRAM only required modifications to existing SDRAM tooling, and since the SDRAM manufacturing processes had been pretty much perfected already yield was high from the get-go.
Rambus' royalties on RDRAM were actually pretty low. I don't remember what they were, but I remember it being under 3%.
-- Under capitalism man exploits man. Under communism it's the other way around.
RDRAM not to bad....
by
Anonymous Coward
·
· Score: 3, Informative
People seem to forget that RDRAM has gone down DRAMATICALLY in price. The days of $900 128mb RIMM's are long gone. You can order 256mb RIMM's for like $76 on pricewatch.com now
Yes DDR is still cheaper per mb, but RDRAM isn't that much more expensive. Especially when you consider how fast & stable it is in the i850 chipset by Intel.
We've all seen the THG memory bandwidth benchmarks that show RIMMS are still way ahead of DDR DIMMS in terms of bandwidth. The CPU, well that's another story, but nonetheless Intel is not taking back what they've said, they're simply offering another choice to consumers, which is a good thing. I don't think we should critisize them for offering up a system with "slightly" lower performance. How rare is it, when a company has been forced (a la AMD and VIA) to offer more choices for the increasingly price/performance consious consumers? Pour a cold one for the little guy.
Hasn't anyone being paying attention to VIA? What about the P4X266 chipset, even Tyan has a board with it: http://tyan.com/products/html/trinity510.html
Intel motherboards and chipsets are fine however you don't have to wait for Intel to come out with a DDR chipset for your P4. VIA has one already.
It wasn't so much the price of RAMBUS...
by
Derek
·
· Score: 4, Informative
...it was the principle!!
For those who missed it. In 1992 Rambus joined an industry consortium (Joint Electron Device Engineering Council JEDEC) made up of companies seeking to develop a *royalty-free* standard for the next generation of memory chips. The resulting standards (SDRAM, and DDR RAM) have been widely adopted during the past few years.
Then, about a year ago, Rambus let the lawyers loose. They claimed that, despite its participation in JEDEC, it owned patents that were being infringed upon by any company making SDRAM or DDR RAM chips without a license. Moreover, Rambus claimed it was entitled to damages in the form of retroactive royalty payments.
Well, this is really old actually.
by
CMiYC
·
· Score: 3, Interesting
Intel's chipsets for P4 have been using DDR for a while now. In fact, most OEMs have been developing P4 systems with DDR for months now. I can't tell you the last time we sold a RAMBUS probe. Further more, all of intel's future processor chipsets will use DDR as well.
I work for a test equipment supplier which will go unnamed.
Slashdot Mirror
Seriously when RAMBUS was 10x the price of SDRAM it seriously hurt, but now that RAMBUS is getting close to comparable, I don't see what the point is. In my neck of the woods PC-800 RDRAM goes for about 30% more than PC2100 DDR, which really isn't that much (and dual channel RDRAM is the fastest RAM platform out there). Given that the P4s one redeeming factor is that with RDRAM it has a serious memory advantage, I really don't see what Intel is thinking: Put a P4 with DDR DRAM and it'll get clobbered even more.
A "stealth introduction"? Doesn't Intel know about Slashdot? We geeks find out EVERYTHING, and as soon as one of us finds out and submits it to Slashdot, everyone knows. With media the way it is and Slashdot here, a quiet introduction is virtually impossible.
-------------------
And it is latency that usually counts, peak transfer rate is not sustainable anyway. A dual channel DDR-SDRAM platform would be faster than dual channel RDRAM platform, single channel ddrs are already competing with dual channel rdrams. With 166*2MHz DDRs on the horizon, I think it is a very sensible solution.
Gentlemen, you can't fight in here, this is the War Room!
It's unfortunate that Rambus RIMMs are even more obsolete now that Intel's new chipsets are going to DDR. Not that I'll miss them, but its always unfortunate when early adopters get hosed with proprietary hardware. Anyone remember the Socket 4 P60/P66? Drop a ton of money on the new Pentium, and watch while everyone who waited is able to upgrade while you can't. Ditto to those with Asus P4Ts - not only are they hosed on processor options because intel changed pinouts for the new P4s, but now the RAM is obsolete too.
From the article: Intel is planning the stealth introduction of a chipset that will let computer makers connect the Pentium 4 to speedy DDR (double data rate) memory.
Speedy? Isn't DDR-SDRAM slower than RDRAM? Sounds a bit fluffy to me. What they really mean, but don't clearly spell out, is that DDR is faster than the normal SDRAM the 845 supports. But its still no RDRAM. Which I guess everyone here knew anyways.
Ahh well, I'm just grumpy b/c I convinced my mom to buy a P4T/Rambus-based P4 1.7Ghz, and now I have to ditch the Ram/Mobo/CPU to upgrade it. (I'd have given her an Athlon but the dustbunnies at her place are such that I'd be afraid of her burning the place down... remember that THG vid of the flaming Athlons?)
-- "Ignorance more frequently begets confidence than does knowledge." (Charles Darwin)
As we all know, the P4 is designed to take advantage of high memory bandwidth-- which is why the 845 chipset (or whatever) that used SDRAM was such a flop.
When the P4 was first introduced, Intel claimed that it was designed specifically for RDRAM. If this is true-- and I suppose it doesn't have to be-- then is it possible that the new DDR stuff will actually perform below RDRAM systems? Is the only advantage going to be price?
I'm not really a tech guy, so this is an honest question. I'm not a Rambus fan-- I've got a PIII with the 820 chipset, and I'm not particularly fond of it-- but could it be that the company that everybody hates is actually the better way to go in this case?
Of course, everybody around here is going to be be gushing over DDR over Rambus-- if they choose Intel over AMD, which doesn't seem likely-- but it seems that Intel is either stepping backwards or conflicting with their past words.
I'm hoping you meant to say RDRAM.
Anyway, the patents had nothing to do with the price differences between RDRAM and DDR SDRAM, it was all due to manufacturing costs. I remember a little over a year ago Kingston was bragging about their 30%(!) yield on PC-800 RDRAM chips. When 70+% of your product doesn't pass QA, that's definately going to drive your costs up! Additionally, manufacturers had a fair amount of retooling to do before they could make RDRAM, and high setup costs get passed on to the consumer. As I recall, RDRAM also has a bigger die size than DDR SDRAM (I could easily be wrong, it's been a while since I cared) which would also drive up costs.
In contrast, DDR SDRAM only required modifications to existing SDRAM tooling, and since the SDRAM manufacturing processes had been pretty much perfected already yield was high from the get-go.
Rambus' royalties on RDRAM were actually pretty low. I don't remember what they were, but I remember it being under 3%.
Under capitalism man exploits man. Under communism it's the other way around.
People seem to forget that RDRAM has gone down DRAMATICALLY in price. The days of $900 128mb RIMM's are long gone. You can order 256mb RIMM's for like $76 on pricewatch.com now
Yes DDR is still cheaper per mb, but RDRAM isn't that much more expensive. Especially when you consider how fast & stable it is in the i850 chipset by Intel.
We've all seen the THG memory bandwidth benchmarks that show RIMMS are still way ahead of DDR DIMMS in terms of bandwidth. The CPU, well that's another story, but nonetheless Intel is not taking back what they've said, they're simply offering another choice to consumers, which is a good thing. I don't think we should critisize them for offering up a system with "slightly" lower performance. How rare is it, when a company has been forced (a la AMD and VIA) to offer more choices for the increasingly price/performance consious consumers? Pour a cold one for the little guy.
Hasn't anyone being paying attention to VIA? What about the P4X266 chipset, even Tyan has a board with it: http://tyan.com/products/html/trinity510.html
Intel motherboards and chipsets are fine however you don't have to wait for Intel to come out with a DDR chipset for your P4. VIA has one already.
Nathaniel P. Wilkerson
www.haidacarver.com
http://www6.tomshardware.com/mainboard/01q4/011008 /index.html
I thought that this chipset looked good enough.
THERE IS NO DATA. THERE IS O
For those who missed it. In 1992 Rambus joined an industry consortium (Joint Electron Device Engineering Council JEDEC) made up of companies seeking to develop a *royalty-free* standard for the next generation of memory chips. The resulting standards (SDRAM, and DDR RAM) have been widely adopted during the past few years.
Then, about a year ago, Rambus let the lawyers loose. They claimed that, despite its participation in JEDEC, it owned patents that were being infringed upon by any company making SDRAM or DDR RAM chips without a license. Moreover, Rambus claimed it was entitled to damages in the form of retroactive royalty payments.
And then the lawsuits began....
-Derek
Intel's chipsets for P4 have been using DDR for a while now. In fact, most OEMs have been developing P4 systems with DDR for months now. I can't tell you the last time we sold a RAMBUS probe. Further more, all of intel's future processor chipsets will use DDR as well.
I work for a test equipment supplier which will go unnamed.