New RDRAM chipset "beats DDR by 50%" claim

[DX2Player]

Your joking right? Because your vast contrabution to this thread with all its 1 post wonder will be missed.

Seriously all Im lookin for is for most people to be more grounded and objective than Fanboy like, admitting that RDRAM is a very good alternative and has a future rather than the DIE RDRAM! mentality

 

[First]

Originally posted by: DX2Player

Originally posted by: Evan LiebAnyway, I'm still open to any sugggestions you may have about future motherboard reviews. Anything you'd like to see in the ABIT SI7 (R658) review or R659 (a ways off) review?

I would like to see reviews pitted up against the Asus P4T533 as well. Every review of the P4T533 shows it ahead of the Asus P4T533-C so I would think that it would be used when a top of the line review is done instead, but it isnt. The slight margins that we are talking about here would warrent such when crowning one or another the king. Its like choosing a Camaro Z28 to represent the fastest Camaro when there is the SS, the 15 extra hp that the SS has might just mean the win when its a photo-finish.

Although might not matter by time we get to the next generation of RDRAM chipsets, maybe could do on next (R655) review.

Actually we've done extensive testing, and there's isn't a bit of difference. The reason you might believe that the P4T533 is faster than the P4T533-C is because the FSB setting on the P4T533 is about 1.5MHz higher at default, about 135.5MHz FSB versus 134.00 FSB (I forgot the exact difference, but it's very close to that, maybe 1.48 or 1.49MHz). This difference alone could add as little as 24MHz or as much as 36MHz in core clock speed difference (this is calculated by knowing that all Northwood processor multipliers currently range from 16X to 24X).

 

[ElFenix]

wow, so many trolls, so little time...

 

[Ice9]

I tried, as did others, to politely rebut some of the erroneous statements Ice9 made. That was a serious mistake because in return I get this from her: "RESEARCH, people... RESEARCH. UNDERSTAND why semiconductor markets are failing. And UNDERSTAND why there's some that are standing despite poor market conditions. Geez"

Is that a know-it-all arsehole or what?

Somehow it reminds me of the attitude of shysters like Jack Grubbyman and the Citi management that fed off his criminal activity.

Sorry all. I'm otta this thread for good. Bye.

Next time, come back with some proof of your claims, please. You probably don't even know a thing about Jack Grubman and what he did, or the management team around him. But hey, you could be some expert that's a lot closer to the truth than I am... Are you?

EVERY DAY there's people who claim "Rambus did this" and "Rambus did that", and 95% of it simply isn't true. And the other 5% is nothing new to the semiconductor industry. After years of researching Rambus the company, and reading the "writing on the wall", I've got a HUGE assortment of proof to my claims.

You had none, aside from maybe stating the obvious of "Tulloch was cancelled"... If you think that doing all of my own research makes me come off like a guy with a "know it all" attitude, it's because compared to the average Rambus hater, I do know it all.

If you HONESTLY thought poor DDR sales were a result of market conditions and a poor economy, then Samsung would be in the same boat as the rest of 'em. But the truth is: They aren't. In fact, someone recently posted a cute little article about 50 executives walking out on Hynix. Why walk out on your company simply because of an economic downturn and poor market conditions? There's something else going on there.

 

[FishTankX]

Well, ICE9, I think I finally understand where your'e comming from in terms of how RAMBUS and DDR work in the market, but if RDRAM really won out over DDR I think we'd still have big problems with RAMBUS pricewars and such. (As mentioned by someone else, no need to reply)

What I do believe though is that RAMBUS, while a superior technology, isn't needed at this point in time and it'd be better to maintain a single memory standard that could be used in hammer and P4 systems and I think DDR400/DDRII will fill that role nicely when the time comes. At this point in time DDR is capable of fulfilling the needs of the P4 while working optimally with the hammer and this will provide less market confusion and will help keep manufacturing simpler. When the benefits of RDRAM become obvious and DDR-II can no longer fulfill the needs of the market, RAMBUS will be there to fill the gap. But frankly I would rather own a memory technology that would be compatible across platforms, than have one that would have optimal performance. It would make a transition to AMD easier.

Many people transitioned from AMD to during the 1.6A and 1.8A days. This wouldn't have been possible without the i845D. Saved them a bunch of money.

You might think such an argument is absurd, but it's not meant as an argument against RAMBUS. Just my personal opinion.

RAMBUS is good technology but I really think the market is ready for a single standard, that would simplyify your upgrade path. One less thing to worry about.

Once Intel sees the light by that time RDRAM will probably be the undeniable victor over DDR. And by that time it should be suitable to be a complete replacement for DDR. Until then..

 

[Ice9]

Well, ICE9, I think I finally understand where your'e comming from in terms of how RAMBUS and DDR work in the market, but if RDRAM really won out over DDR I think we'd still have big problems with RAMBUS pricewars and such. (As mentioned by someone else, no need to reply)

Competition, yes. DUMPING, no. I don't think there'd be "problems" like there are today. Today, there's no possible way DRAM manufacturers can make a profit on DDR. The biggest DDR companies are keeping DDR prices low for two reasons: Slow the adoption of Rambus, which has the potential to take control of the memory industry - and - force out competitors by driving them out of business. This will allow them to raise prices.

The thing we're waiting for now is a settlement from Infineon (not a matter of "if", but "when"). Once that happens, the remaining two will settle and pay royalties on DDR since the major precedent will have been set. There will be little reason to NOT manufacture RDRAM then, since the royalty rate is HALF that of DDR, and the manufacturing cost is about the same.

Some people may argue that, but Samsung has already claimed that RDRAM production costs are "nearly the same" as DDR for the DRAMs itself, but the module costs are cheaper because there's no advantage to using a 6 layer PCB for RDRAM. No signal issues on 4 layers.

But frankly I would rather own a memory technology that would be compatible across platforms, than have one that would have optimal performance. It would make a transition to AMD easier.

Many people transitioned from AMD to during the 1.6A and 1.8A days. This wouldn't have been possible without the i845D. Saved them a bunch of money.

You might think such an argument is absurd, but it's not meant as an argument against RAMBUS. Just my personal opinion.

Well, I don't necessarily believe this is true overall, but it's as good an explanation as any for "our" market (enthusiasts)... The only "fact" that's out there is that huge OEM's like Dell and Compaq complained to Intel relentlessly because there just weren't enough RDRAM modules to go around early on (the reason why 128MB PC800 modules were going for $1200).

If DDR never happened for the P4, and all of a sudden RDRAM would be the only option, you'd see a lot of supply problems if Samsung and Elpida were the only two making the stuff. Since the biggest players in the DRAM industry at the time "declined" intel's offer to refit their fabs, the powerplay was purely in favor of SDRAM/DDR. Intel simply can't ignore that, as depending on RDRAM would have cut into their own sales (fewer rdram modules sold, fewer P4's sold).

It didn't make *business* sense to stick with RDRAM if only 1 or 2 companies were responsible for ALL RDRAM production. Technological sense is a whole other story

 

[DX2Player]

Reading it seems there is more to it than just that. Your right they do slightly oc the FSB, I have the P4T533 board and Im looking in CPUID and my FSB is set to 133.64 at stock while my CPU is showing 2539.09, im not shure that the extra .64 FSB alone would be enough to change the scores though. Even in Sandra the comparisons show the 32bit ram scoring higher, maybe they are all wrong though.

"PCMark 2002's memory benchmark shows the RIMM-4200 based P4T533 system giving an even bigger performance gain over PC-1066 and DDR-333/266 motherboards. Judging this benchmark, we can see the P4T533 giving roughly 15% better memory performance compared to our other Pentium 4 platforms." -GamePC

"This is more than likely due to decreased latencies by having both channels integrated onto the same module." -GamePC

"As you can see 32bit RDRAM 4200 is only slightly faster then 16bit PC1066 memory. This slight increase in speed could be attributed to slightly lower latencies since the memory path does not have to access two RIMM?s on two different channels in two different locations on the motherboard at the same time, but now only has to go to one place basically." -HardOPC

"The ASUS P4T533, the only i850E board with support for 32 bit RDRAM, is still overall the fastest chipset for the desktopuser. 32 bit RDRAM is a little faster than two channels of 16 bit RDRAM and in some quite a few workstation applications it significantly outperformed the i845PE and DDR333." -Ace's Hardware

Here is a recent article at Ace's Hardware that actually does use both the P4T533-C and the P4T533 boards against the Granite Bay

And from the benchmarks you can see higher scores when using the P4T533 board.

Anyways gotta go to class, might edit this later to add more

Maybe the older BIOS version had the stock FSB higher.

 

[MadRat]

I think Ice9 wants a nice SiS655 versus i850e review with every processor possible...

I personally like RDRAM technology over the DC-DDR technology if not for the fact that DDR is being sold for too low of margins now. I don't think any foreign company should be able to dump product at below market cost in our country, even if we like it. Its bad blood for the future economic environment.

I also think that the old "pin count counts" theories are going out the door with newer technologies. RDRAM has far fewer pins and the design should be easier to implement, but in practice this is not the case. RDRAM signalling is more susceptible to outside interference than DDR signalling and therefore the design requires a tighter spec. So in practice it appears that the simpler technology is more expensive.

 

[Ice9]

I think Ice9 wants a nice SiS655 versus i850e review with every processor possible...

Nah, just the fastest

I personally like RDRAM technology over the DC-DDR technology if not for the fact that DDR is being sold for too low of margins now. I don't think any foreign company should be able to dump product at below market cost in our country, even if we like it. Its bad blood for the future economic environment.

Actually, the Koreans are far less guilty of this than WE are. Our own Boise, Idaho company - Micron - is the supreme #1 DDR price dumper. Hynix is #2, Infineon is #3.

I also think that the old "pin count counts" theories are going out the door with newer technologies. RDRAM has far fewer pins and the design should be easier to implement, but in practice this is not the case. RDRAM signalling is more susceptible to outside interference than DDR signalling and therefore the design requires a tighter spec. So in practice it appears that the simpler technology is more expensive.

Actually, the opposite is true. The biggest signaling problem belongs to dual channel DDR - the fact that every single one of its traces has to be the same length to maintain signal integrity. Yes, Rambus traces have to be SHORT, but that's far easier to do than have 128 to 256 traces on a PCB be all the same length. Particularly across multiple motherboard layers. Granite Bay had to be the most expensive chipset to launch in intel's history. It took them FOREVER to get the design working properly with 256 traces across 6 PCB layers, hence the big delays. When they first hit the scene, they were $250 or more. They've dropped to around $190-200, but they're still more expensive than any other chipset for desktops/workstations.

Regardless, Rambus already thought of this LONG ago, and now has Redwood. Redwood gets around the trace length limitation that kills Parallel designs (you can read about it here). Interestingly enough, they made it backward compatible with Hypertransport, and made it completely compatible with the implementation being used by Rambus and PLX for PCI Express. This makes it both Intel-friendly and AMD-friendly. And of course it's already Sony PS3 friendly

 

[FishTankX]

Excuse me Ice9, but I really have to question you on that point.

The fact that DCDDR is harder to implement than RDRAM isn't necescescarily true. If so, how did the Nforce come out so quickly, espically when it was Nvidia's first chipset? If so, how did SiS come out with the SiS655 at a price point so low (Not to mention most board designs can operate at dual DDR 400 which at one point was harder to get working than DC-RDRAM) as to undercut the majority of i850E boards?

And I believe the 'Expensive' pricing of the granite bay is due to the fact that it is a workstation chipset and there's going to be a price differential between workstation and P.C. chipsets. Just look at the huge difference between i850 and i860 chipsets! In terms of sililcon there's hardly any difference, besides the extra bus connections. But WHOO HOO! Do those boards cost.

 

[Ice9]

The fact that DCDDR is harder to implement than RDRAM isn't necescescarily true. If so, how did the Nforce come out so quickly, espically when it was Nvidia's first chipset?

That chipset was a long time coming... and they didn't get it right the first time.

If so, how did SiS come out with the SiS655 at a price point so low (Not to mention most board designs can operate at dual DDR 400 which at one point was harder to get working than DC-RDRAM) as to undercut the majority of i850E boards?

Easy to answer... Just because it's cheap doesn't mean it's cheap to manufacture. We're all learning that lesson from DDR, aren't we? Or do you guys still honestly believe that DDR is being brought to market in a PROFITABLE manner?

And I believe the 'Expensive' pricing of the granite bay is due to the fact that it is a workstation chipset and there's going to be a price differential between workstation and P.C. chipsets. Just look at the huge difference between i850 and i860 chipsets! In terms of sililcon there's hardly any difference, besides the extra bus connections. But WHOO HOO! Do those boards cost.

Then why the e7505? If what you say is correct, then there's no reason for the E7505 to exist.

E7505 is Intel's workstation chipset, not E7205. I don't care what anyone says. Intel is labeling it as a workstation chipset just to justify the increased cost.

 

[FishTankX]

Disclaimer:My spelling sucks. Sorry in advance.

Asus's i850 board cost 180$.

ECS's cost like, 80$ if I remember correctly.

If it costs less than 60$ to manufacture a K7S5A I have a hard time believing that Gigabyte is loosing any money on their low end SiS655 implmenetation.

Now, will someone please tell me if the SiS655 boards are 6 or 4 layer?

Oh, by the way. In the begining, all of the RAMBUS boards were 6 layers. You're telling me that the GB boards are 6 layers. Yeah, so what?

Who's fooling who?

Anyways, the granite bay is probably for low end workstation and the E7505 is high end. Not every workstation needs 2 CPU's, but alot of workstations for use in heavy enviornments might choke on the 2GB RDRAM limit in the i850E. Not to mention the price of the RAM. (Hey, don't blast me on mentioning the price of cheap DDR anymore. The problem will resolve itself, as soon as a company goes bankrupt. Or two. Heh.) Just ask Sharkeeper! In applications that would require alot of RAM and a good fast single (hyperthreaded) CPU but are hard pressed to take advantage of another CPU the GB would do just fine, and would save alot of money versus the E7505. Tell me though, what was the last desktop that Dell put an E7505 in? It just wasn't meant for that purpose, bottom line.

Last time I checked, lots of companies had single processor workstations loaded with huge amounts of RAM. It's that kinda market that the E7505 fits. You get what i'm saying?


And Ice9, not as a personal attack, but you really have to realize that Single channel DDR is 64 bit. DCDDR is 128. Not 256. Unless you're in the know about something I'm not, please stop stating that DCDDR boards have 256 memory traces. Please!

 

[Ice9]

Asus's i850 board cost 180$.

ECS's cost like, 80$ if I remember correctly.

You're focusing on how much the board costs for YOU to buy. You're not taking into account how much it costs them to make it. If these Dual Channel DDR boards were so darned cheap to make, and parallel interfaces were so much more advantageous, then the world wouldn't be moving towards more serialized interfaces.

At this point, I think all you care about is what you pay for this technology. That's ok, but it doesn't make it the most cost effective.

If it costs less than 60$ to manufacture a K7S5A I have a hard time believing that Gigabyte is loosing any money on their low end SiS655 implmenetation.

Where are you getting that $60 number? Somewhere verifiable?

Oh, by the way. In the begining, all of the RAMBUS boards were 6 layers. You're telling me that the GB boards are 6 layers. Yeah, so what?

In the beginning. They quickly moved to 4 layer designs, rather easily I might add.

Who's fooling who?

Well, you're trying to fool people into believing that a 6 layer design costs the same or less than a 4 layer design... You're also trying to convince people that a higher pincount is just as cheap to deliver as a lower pincount...

The fact that these boards WIND UP costing the same makes no difference. There's far less demand for RDRAM boards, that much is KNOWN. What they lose on each initial sale for DDR, they make up for in volume.

Anyways, the granite bay is probably for low end workstation and the E7505 is high end. Not every workstation needs 2 CPU's, but alot of workstations for use in heavy enviornments might choke on the 2GB RDRAM limit in the i850E. Not to mention the price of the RAM. (Hey, don't blast me on mentioning the price of cheap DDR anymore. The problem will resolve itself, as soon as a company goes bankrupt. Or two. Heh.) Just ask Sharkeeper! In applications that would require alot of RAM and a good fast single (hyperthreaded) CPU but are hard pressed to take advantage of another CPU the GB would do just fine, and would save alot of money versus the E7505. Tell me though, what was the last desktop that Dell put an E7505 in? It just wasn't meant for that purpose, bottom line.

What was the latest Dell that has the E7205 in it for that matter? Why haven't they JUMPED at the chance to deliver this "new and great" dual channel DDR solution to the masses?

Answer: It's too expensive.

Last time I checked, lots of companies had single processor workstations loaded with huge amounts of RAM. It's that kinda market that the E7505 fits. You get what i'm saying?

Yeah. I get it. No one is buying it, at least not from the top computer manufacturers out there. Not Dell. Not HPAQ. Not...er....anyone that matters. When 850/850E hit the streets, dell was READY with systems to ship. Gee, what happened to Granite Bay and E7505?

And Ice9, not as a personal attack, but you really have to realize that Single channel DDR is 64 bit. DCDDR is 128. Not 256. Unless you're in the know about something I'm not, please stop stating that DCDDR boards have 256 memory traces. Please!

Sorry. You're right here.

But the point of 32 traces vs. 128 traces sure does make a point just the same. And quad channel RDRAM floating in at still HALF that pincount pushing 9.6GB/sec... well, what can I say. That still impresses the hell out of me far more than what dual channel DDR does. I mean hey, drive your point into the ground all you want... the only reason you like DDR is because you can buy it for less than it costs to make. That's the only reason ANY consumer likes it.

Has nothing to do with the TECHNOLOGY itself.

 

[FishTankX]

I just don't get it. The E7205 should be comparable in cost to the i850E in configurations of 512MB of RAM, and cheaper in more generous configurations. Certianley dell must have workstaiton configurations with 1+GB of RAM, or not.. but HP sure does.

HP/Compaq E7205 workstation line

I'd also like to reiterate that the E7205 was designed as an entry workstation level chipset. Better suitd for large amounts of RAM than is the i850E. It can also use gigabyte memory modules meaning you could have 2 gigabyte memory modules and leave the rest free for an upgrade down the road. Anyways... my allegations are based on what I see. This place seems to back me up in my theories. As stated here

HP's workstation press release. Title:HP Marks Integration of Workstation Product Lines with Launch of New Systems

Anyways, I'm not saying that DDR is superior in any way. I'm just saying that it's the cheapest sollution. No company will hold a sollution that looses money for long. If the market really needs RDRAM or DDR really drives these companies into near bankruptcy, the situation will change and the Price of DDR will go up or they'll start making RDRAM if Intel decides to support it in the next generation.

And the fact is that RAMBUS reaches these incredible bandwidth's by running at ludircous frequencies. DDR reaches it's bandwidth by lentghening the bus. They're simply two different aproaches.

I don't know how DDR-II will change things, so i'm not able to comment on that. But I think that might put it back in the mainstream performance sector.

At any rate, one thing I don't understand is if RDRAM has such incredible bandwidth, why haven't videocard manufacturers jumped on it? Maybe DDR really does have it's purpose, long after RDRAM dominates the P.C. Market. DDR seems to have had a strangehold on the videocard market, that's for sure. Not meant to be a retort in any way. Just a musing. It's amazing that DDR in videocards has far surpassed even the Cache modules of not too long ago CPU's. No?? 20GB/s is amazing, aint it? Even RAMBUS doesn't have anything that will reach anywhere near that speed, as far as memories go.

I'm sure if RDRAM was adapted for videocards it could give DDR a run for it's money. But it seems that RDRAM isn't ever going to penetrate the videocard market. It seems that paralell memories will have a much longer lifespan in the videocard market.

 

[FishTankX]

And, Ice9, why should I care if the companies are loosing money? They could very well all form a cartel and raise prices to near RDRAM levels. What's stopping them? My purchasing decision certainley has no bearing on memory prices. They could very well raise prices to a reasonable level and all would be well. We would once again have price parity between RDRAM and DDR. But why isn't that happening? That's the part of your arugument I don't understand.

You say that DDR is loosing money. You say that DDR will be the death of the companies. So why don't they just stop and raise prices? Together? That's the question that you haven't answered yet.

 

[Ice9]

And, Ice9, why should I care if the companies are loosing money? They could very well all form a cartel and raise prices to near RDRAM levels. What's stopping them? My purchasing decision certainley has no bearing on memory prices. They could very well raise prices to a reasonable level and all would be well. We would once again have price parity between RDRAM and DDR. But why isn't that happening? That's the part of your arugument I don't understand.

I don't like seeing people LOSE THEIR JOBS over corporate incompetence. Despite the fact that I make a good deal of money in the IT industry, I always imagine what it would be like if I lost my job having a wife and 2 kids.

Maybe the young, idealistic folk who "just want it as cheap as they can get it" can easily forget about that, I simply can't.

But again, i'm one of the few blooming idiots on the internet who PAYS for his software, BUYS his CD's and doesn't expect everything to be delivered to me for pennies on the dollar.

IT'S JUST ME!

You say that DDR is loosing money. You say that DDR will be the death of the companies. So why don't they just stop and raise prices? Together? That's the question that you haven't answered yet.

EDGING OUT YOUR COMPETITION IS NOTHING NEW! The reason I haven't answered that is simple: It goes without saying!

Samsung WANTS micron to die.
Micron wants Hynix to die.
Hynix wants infineon to die.

Less competition means your OWN prices aren't dictated by the prices of your competition.

 

[FishTankX]

Hey man, you've got me wrong. My point is that even if I don't like the fact that the DRAM manufactuerers are plunging into impending doom, there's nothing I can do to raise prices. What am I gonna do? Tell my local DRAM seller 'Hey! I'll buy that stick for double price! Please give the massive profit back to the memory manufacturer.'? The point is that there is nothing I can do but let the RAM war take it's course. Arguing business points is pointless because I doubt either me or you will ever have much of an affect on the price of DRAM. Or the companies future. But methinks that when this RAM war is over the only two companies that won't come out completly devistated are the ones manufacturing RDRAM. Just a thought. How much of Samsung's sales is RDRAM, anyways? I'm not sure if any amount of profit from RDRAM could cover their DDR losses.

 

[Ice9]

HP/Compaq E7205 workstation line

I'd also like to reiterate that the E7205 was designed as an entry workstation level chipset. Better suitd for large amounts of RAM than is the i850E. It can also use gigabyte memory modules meaning you could have 2 gigabyte memory modules and leave the rest free for an upgrade down the road. Anyways... my allegations are based on what I see. This place seems to back me up in my theories. As stated here

HP's workstation press release. Title:HP Marks Integration of Workstation Product Lines with Launch of New Systems

I stand corrected.

However, I gotta know... What's "endian.net"?

Anyways, I'm not saying that DDR is superior in any way. I'm just saying that it's the cheapest sollution. No company will hold a sollution that looses money for long.

I certainly agree with that...

If the market really needs RDRAM or DDR really drives these companies into near bankruptcy, the situation will change and the Price of DDR will go up or they'll start making RDRAM if Intel decides to support it in the next generation.

What the market needs is consolidation.

Before the big Y2K boom, we had MORE than enough dram manufacturers. Then the orders for y2k upgrades came along, and EVERYONE wanted in on it.

We went from 4 major DRAM houses to *12* thanks to the Y2K boom.

Only 2 have gone bye-bye since the onset of Y2K. That leaves 10.

OF those 10, 7 pay royalties to Rambus and recognize them as the patentholder for DDR and SDRAM.

The last remaining 3 don't pay royalties... If you dont' know which 3 of these companies are doing it by now, then ignore this thread

So now, we've got the 3 EXCLUSIVE DDR PLAYERS trying hard to edge each other out of the market.

And the fact is that RAMBUS reaches these incredible bandwidth's by running at ludircous frequencies. DDR reaches it's bandwidth by lentghening the bus. They're simply two different aproaches.

"ludicrous"?

How about "appropriate"?

Lengthening the bus is what everyone wants to avoid, since that's where the manufacturing costs come in. Ignore it if you want.

I don't know how DDR-II will change things, so i'm not able to comment on that. But I think that might put it back in the mainstream performance sector.

DDR-II isn't shipping in volume until 2005. By then I think it will be completely irrelevant. Rambus' own technologies will have completely surpassed it, and DDR2 will still cost a huge royalty compared to Rambus Native stuff.

You keep forgetting that Rambus is entitled to royalties on those patents

At any rate, one thing I don't understand is if RDRAM has such incredible bandwidth, why haven't videocard manufacturers jumped on it?

Short answer: No one has licensed it.

Long answer: Let's say you're ATI or NVIDIA. If 8 of the ten dram manufacturers are making DDR (and NOT Rambus RDRAM because they don't want 'em having control), why limit yourself (and your customers) to RDRAM, even if it *IS* superior? You can make far more money/sales by selling GPU's compatible with DDR. If you limited yourself to RDRAM, you'd find yourself in a position of selling FEWER GPU's to OEMs as a result.

Remeber, as much as I feel RDRAM is superior, there's ONLY TWO COMPANIES MANUFACTURING IT. This has an impact on the market overall, and who's gonna support it in the end.

Maybe DDR really does have it's purpose, long after RDRAM dominates the P.C. Market. DDR seems to have had a strangehold on the videocard market, that's for sure. Not meant to be a retort in any way. Just a musing. It's amazing that DDR in videocards has far surpassed even the Cache modules of not too long ago CPU's. No?? 20GB/s is amazing, aint it? Even RAMBUS doesn't have anything that will reach anywhere near that speed, as far as memories go.

Rambus has 100GB/sec bandwidth that they can put out RIGHT NOW with Yellowstone. They've been beyond the 20GB/sec mark for quite some time. Remember, the Radeon 9700 has a 256 bit datapath for its 19GB/sec bandwidth... Rambus doesn't need that many pins to equal that amount of bandwidth. But again, NO ONE HAS LICENSED IT FOR THIS USE YET because only two companies are making it. This will change when the last "Big three" find themselves paying $1B in back royalties.

I'm sure if RDRAM was adapted for videocards it could give DDR a run for it's money. But it seems that RDRAM isn't ever going to penetrate the videocard market. It seems that paralell memories will have a much longer lifespan in the videocard market.

Yet to be seen. Too early to tell.

I can tell you this though: Rambus will outlast ANY current memory manufacturer that's exclusively making DDR And at this point, that's all they have to do.

 

[FishTankX]

Endian.net is a massive newssite. They have roadmpas showing major industry releases, and keep an updated news section.

Anyways, RDRAM reminds me of this one player I was playing on age of empires. I had taken all the resources on the map, mined them all. But he had 5 relics. So he had a constant steady flow of cache. He built his defenses well, and just held off all my soldiers. In the end, he won, because he had a never ending cash flow (like RAMBUS royalties from DDR). It seems like it's game over for DDR, doesn't it? But I think at this point no one is completly sure how this will pan out. All I can say, is that..

I'd pay more for DDR, but I'm just an 18 year old kid who makes 200$ a month, not even in college yet. You can't hold me accountable for my spending like an adult. (Not to mention I live in China where this is your average teacher's salary) so for me this temporary memory crash is a Godsend for me. Espically because I want to play Unreal 2 with only 256MB of RAM, and i've decided i'll put that on hold for now. More RAM, here I come! I'll buy expensive memory when I get a job. Probably even jump on the RAMBUS bandwaggon again. For now, i've made my decision. I'm sure with Micron's losses in the millions, loosing 40 more dollars won't hurt them too bad.

Note to all who might read this: If RAMBUS was slower but cheaper then I would continue with RDRAM. It's not wanting to 'jump ship', merely financial reasons. To tell you the truth, I like RAMBUS. Just don't like the price over here. Even if RDRAM was at U.S. levels (Which it isn't. 256MB PC800=75$ over here, minimum. At some places it fetches 100$.) i'd continue with RDRAM. But I really have no choice.

 

[FishTankX]

Oh yes, one question. If DDR-II isn't due for 2005 than why is SiS making a DDR-II chipset due around july 2003? Seems mighty strange for the board to precede the memory by 1 1/2 years

Edit:I just realized that Intel has a DDR-II chipset schedueled for 2003 Q1 but cancled it. Intresting, isn't it?

 

[Howard]

Somebody give Ice9 a cookie for all the valuable insight he's brought to this thread - and possibly revealing an injustice to the ignorant masses.

 

[Ice9]

Hey man, you've got me wrong. My point is that even if I don't like the fact that the DRAM manufactuerers are plunging into impending doom, there's nothing I can do to raise prices. What am I gonna do? Tell my local DRAM seller 'Hey! I'll buy that stick for double price! Please give the massive profit back to the memory manufacturer.'? The point is that there is nothing I can do but let the RAM war take it's course. Arguing business points is pointless because I doubt either me or you will ever have much of an affect on the price of DRAM. Or the companies future. But methinks that when this RAM war is over the only two companies that won't come out completly devistated are the ones manufacturing RDRAM. Just a thought. How much of Samsung's sales is RDRAM, anyways? I'm not sure if any amount of profit from RDRAM could cover their DDR losses.

No, you're absolutely right... There's nothing you can do about it...

But you've constantly been using its price as the basis of your argument for DDR being "superior" or "more desirable". You always seem to fall back on this point.

I'll sum up my points on this entire mess

Yes, it's sad that DDR is the dominant memory technology being that faster technologies (EVEN BEYOND RDRAM) that exist. 2 years have gone by, and a LOT more development could have been done on chipsets that support Rambus technologies. If Intel stuck to their guns and moved towards the memory innovation route that Rambus provided, rather than allowing themselves to be influenced by JEDEC, we'd ALL have 1333 FSB's by now pumping 10GB/sec bandwidth. JEDEC is GREAT at what it does

The major DDR companies that don't pay royalties to Rambus are breathing their last breaths (there's 3 of them, out of 10 dram companies total that make memory for PC's - WHY they're losing money so badly is anyone's guess.... but my guess is somewhat educated

With SDRAM and DDR having no "killer app" (as has been said elsewhere), Rambus has far more shelf life, and they don't sit on their hands like JEDEC does. They will be here LONG after DDR is gone. PS2 keeps RDRAM alive. SiS is moving to do the same thing (and could potentially embarrass Intel by having a far more viable chipset solution for their OWN DAMN CPU)... PS3 is a conduit for Yellowstone and Redwood (royalties collections starting next year). PLX and Rambus are delivering PCI express (more royalties) which intel can't help but embrace... RaSeR and associated technologies are used in networking equipment (royalties collections starting this year) and intel's newest networking offerings still use Rambus technology.

JEDEC doesn't have a future this bright


I

 

[Howard]

Originally posted by: kuk
I'm surprised no one has said this ... but ...

GUTB??? :Q

I'm ashamed that you even tried to suggest comparing an intelligent, rational person to a raving madman.

 

[FishTankX]

Originally posted by: Howard
Somebody give Ice9 a cookie for all the valuable insight he's brought to this thread - and possibly revealing an injustice to the ignorant masses.

Yes, he's certianley cleared me up on business perspective. I've always been clueless on financial dealings and such.

And Ice9, I never said that DDR was superior. I said it was faster. I said it was cheaper. The cheaper is conditional. In a few months if someone goes bankrupt the price of DDR could errupt and the price/performance crown could go back to RDRAM. I'd be fine with this. You have to realize that all i'm saying is that DDR is the most desirable sollution at this point in time. Unless it's a technological debate (which this doesn't seem to be) i'll always reserve judgment on the future.

Now I realize that the *reason* why DDR is the most desirable sollution at this point in time is partly due to the fact that a bunch of companies are trying to bankrupt eachother. This is abusrd. This is rediclous. This is *reality*. I'm not one to play 'What if's'. DDR is cheaper, and in DCDDR configurations, faster at this point in time. This is what i've been saying the whole thread. Whether or not the R659 will change the faster part is yet to be seen.

That's *all* i'm trying to say, man. Please don't read into my posts. DDR isn't technologically superior, neither is it the way of the future. RAMBUS vs DDR is just like the old P3 core v.s. the Williamette. All it needs is a little time. DDR will probably fade out of existance eventually. But at this point in time, (And my upgrade path) DDR just seems like the more desirable sollution. The moment RDRAM has the upperhand, i'll switch back over. I like RAMBUS. But under the current conditions, it's impossible to say that RDRAM is the best performance sollution for the P4 at this point.

 

[Ice9]

That's *all* i'm trying to say, man. Please don't read into my posts.

Easy, killer! I love a good debate... you've given this thread just that. You are OBVIOUSLY quite knowledgable.

I'm just as liable to buy a beer for a guy who agrees with me as a guy who disagrees with me You have *certainly* proven the cost aspects of DDR vs. RDRAM, JUST as much as i've tried to show why one is cheaper than the other.

And i'm certainly not disputing the viability of DDR... Like I said, they *ARE* pushing 19GB/sec on ATI cards... That's no small feat. And dual channel DDR chipsets ARE besting 850E chipsets. I'm not going to even ATTEMPT to prove the contrary at this point.

I just personally feel Rambus has the innovation going forward - and it seems like you agree. I just hope the dram industry lets it happen.