RAMBUS vindicated.

[Least]

Plessy v. Ferguson.
MGM v. Grokster.

Just look at history.

 

[Terra]

Well, what would you like to know? Remember, this is a few years ago, so I'm not going to be all like "On machine #347...".

I didn't want you to be that specific..after all..it 3-4 years back, no biggie

I built a ton of machines, and the Intel/RDRAM ones never lived up to the hype. Hell, PIII/SDR machines were spanking some of them in certain areas at the time. Gaming on AMD was a nicer experience back then. It was enough to convince me that my needs as a gamer and general computer user were better served by AMD. Same performance for much less money. Oh, sure, you can benchmark and maybe get a few more megs per second of bandwidth... But it's like the difference between a 1 caret diamond at market value and a 1.1 caret diamond that's overpriced due to some hype -- they're both nice fucking rocks, but I'd be buying the 1 caret.

PIII?
Let's keep this about the CPU's designed for RDRAM...the P4's please
And I don't know if I can dig up an old thread (from a danish forum) where I compared my rig againt XP +2400's on DDR2100...back in the June/July 2002 when I got my rig.
(Damn this rig has lated me long time *L*)
They started out by saying RDRAM wasn't faster...
When I proved them wrong, it all suddenly became about the price....not the preformance.
Like I said..a lod of FUD was spread in those days...
I am NOT saying that you are doing the same, but your 1 caret and 1.1 caret (wich should be more like 1 caret vs. 1.25 caret(1990MB/sec - 1914MB/sec versus 2450MB/sec)) and your post still don't give me any idea of what kinda preformance spread we are talking about here...not even preformance estimates...
And P-III's with a 100 Mhz FSB I really can't see what has to do with anything?
You ever tried the i850 and the i850E chipsets?

Terra - I'll go digging..if the sites even stilll excists *L*

 

[|CR|Constantine]

And:



I wonder if you are so correct, that RAMBUS won the courtcase then?
If you words where true, they should have lost, not won...

Terra - Unless I have misunderstoood something in the verdict?

Just because a company did not break the law does not mean what they did was utterly despicable. They stole the foundation of a technology shouldered by many people, slapped an name on it and patented it. They had the balls to turn around and sue them for using their technology. You tell me Buttercup, this does not sound wrong to you?

I do not hold grudges and I am glad the computing industry went down the right road. Rambus is now working with AMD and it's my hope that their talented engineers bring better things to the table. But make no mistake about it our industry is about the greater good and I for one with stand on my soap box in every forum to defend that.

It's that important.

Oh and for your information RDRAM was developed for the P3 and was out in the mainstream long before the P4 came around. I know I owned the first Intel RDRAM chipset and 256MB of RDRAM when it came out.
...the last time I ever pay $1,000.00 on EBAY for 256MB memory module.

Good thing I sold it to another schmuck a few weeks later for the same amount.

 

[GilmourD]

And:



I wonder if you are so correct, that RAMBUS won the courtcase then?
If you words where true, they should have lost, not won...

Terra - Unless I have misunderstoood something in the verdict?

I think the RIAA set a precedent on that... If you could successfully sue 12 year old children, senior citizens that have never owned a computer, and people that were dead at the time of the "crime" then I could probably sue you for eating all my Ruffles even though you're probably nowhere near me.

I really honestly feel that some judges choose verdicts with a coin or dart board.

 

[Least]

...Surely they'll appeal?

 

[GilmourD]

PIII?
Let's keep this about the CPU's designed for RDRAM...the P4's please

Somebody needs to read what I actually wrote and stop spinning. I said that PIII/SDR combos were making some P4/RDRAM combos look bad.

And I don't know if I can dig up an old thread (from a danish forum) where I compared my rig againt XP +2400's on DDR2100...back in the June/July 2002 when I got my rig.
(Damn this rig has lated me long time *L*)
They started out by saying RDRAM wasn't faster...
When I proved them wrong, it all suddenly became about the price....not the preformance.
Like I said..a lod of FUD was spread in those days...
I am NOT saying that you are doing the same, but your 1 caret and 1.1 caret (wich should be more like 1 caret vs. 1.25 caret(1990MB/sec - 1914MB/sec versus 2450MB/sec)) and your post still don't give me any idea of what kinda preformance spread we are talking about here...not even preformance estimates...
And P-III's with a 100 Mhz FSB I really can't see what has to do with anything?
You ever tried the i850 and the i850E chipsets?

Terra - I'll go digging..if the sites even stilll excists *L*

What's funny is that you keep comparing RDRAM to PC2100, when PC2700 was readilly available, and during RDRAM's death throws, PC3200 was released. You can go back to the [H] home page archives and look back in later 2002.

As far as the PIII/SDR combos... If they're spanked some lower end (higher clocked) P4/RDRAM combos, that means nothing to you?

And... Hi? Have you been reading what I've been writing? I was building systems at that point. I didn't have much of a choice, since I had to build machines to customer's specs, so I built machines with all sorts of combinations (as would any other system builder).

 

[Terra]

Somebody needs to read what I actually wrote and stop spinning. I said that PIII/SDR combos were making some P4/RDRAM combos look bad.

You are talking about the first Willamette P4's right?

What's funny is that you keep comparing RDRAM to PC2100, when PC2700 was readilly available, and during RDRAM's death throws, PC3200 was released. You can go back to the [H] home page archives and look back in later 2002.

No need to..I remember it quite, well as I purchased my system bacj then and browsed to find the best solution at the time.
RDRAM PC800 vs DDR PC2100...
RDRAM PC1066 vs. DDR2700...
RDRAM PC1200 vs DDR3200...Where quad channel gave RDRAM the upperhand...

As far as the PIII/SDR combos... If they're spanked some lower end (higher clocked) P4/RDRAM combos, that means nothing to you?

Not if you are talking about the Williamette P4's...

And... Hi? Have you been reading what I've been writing? I was building systems at that point. I didn't have much of a choice, since I had to build machines to customer's specs, so I built machines with all sorts of combinations (as would any other system builder).

I would like to know your experinces with P4 Northwoods, i850, i850E and SiS R696?
You talk like they never excisted?

Terra...

 

[Terra]

as for latency, i'm honestly not sure who's the winner on that

I dug this up, does this help you?

Terra...

 

[Terra]

And a general note:
You guys talk(remember that I am NOT from the US) like you justice system is dead?
Does anyone have any more information about this, I'm not trying to be stupid, but the only thing most europeans know about the legal system in the US is what we get from TV-shows/movies?

Anyone got a good link to the current state(or lack of ) of the US legal system?

Terra - I find it hard to answer without knowing to much about about it

 

[lopoetve]

It had nothing to do with preformance.
It was due to all the FUD, and people screaming about DDR was cheaper...
Sure is was cheaper...it was "illegal"...

Terra - RAMBUS didn't do you anything....it wasn't them that spread FUD and broke the law

No, it was cheaper because it was cheaper. Easier to make, differences in design, etc...

Did rambus save your mother from a well or something? Why the hell would you want the only ram we're going to be using to cost more?

 

[lopoetve]

Quad-channel RDRAM at 9.6GB/s is far form "obsolete"...even today...and that was 3 years ago SiS did that...

And the preformance was there....people just thought with their collective *biiiip*'s

Terra...

Latency is higher and cost blows. You have to realize, the AMD archetecture didn't NEED the bandwidth, and since it was winning the performance crown (overall), no one really cared about the alternative.

 

[Terra]

No, it was cheaper because it was cheaper. Easier to make, differences in design, etc...

Did rambus save your mother from a well or something? Why the hell would you want the only ram we're going to be using to cost more?

Because when I got my P4, not DDR RAM could give 3.2G/S..unless I turned to OC and got lucky with my core.
Did you even read the thread?
Why would I want to make RAM a bottle-neck in my system?

Terra - And please leave my family outside this...you don't se me talking about this either..as it's got nothing to do with the topic...

 

[lopoetve]

And:



I wonder if you are so correct, that RAMBUS won the courtcase then?
If you words where true, they should have lost, not won...

Terra - Unless I have misunderstoood something in the verdict?

Being greedy isn't illegal, it just gets you busted, like what actually happened. If it had been the better tech, it would have won, and we'd be using it now.

 

[lopoetve]

And a general note:
You guys talk(remember that I am NOT from the US) like you justice system is dead?
Does anyone have any more information about this, I'm not trying to be stupid, but the only thing most europeans know about the legal system in the US is what we get from TV-shows/movies?

Anyone got a good link to the current state(or lack of ) of the US legal system?

Terra - I find it hard to answer without knowing to much about about it

that explains a lto.... lawsuits are a frapping joke, that's what. It means almost nothing that someone won...

Remember, someone won millions because their mcdonalds coffee was hot and they spilled it on themselves

 

[lopoetve]

Because when I got my P4, not DDR RAM could give 3.2G/S..unless I turned to OC and got lucky with my core.
Did you even read the thread?
Why would I want to make RAM a bottle-neck in my system?

Terra - And please leave my family outside this...you don't se me talking about this either..as it's got nothing to do with the topic...

That was then, this is now. There is no reason to wish higher prices on us all. The technology that was cheaper and easier to make won.

 

[BillLeeLee]

Ever hear that lawsuit about the bank robber who held up a bank, and as he was escaping, a piece of the bank's scaffolding fell on him? He sued the contractors and the bank, and won. Welcome to US, land of meaningless and frivolous lawsuits.

All this Rambus stuff was years ago, and sure, if Rambus had become accepted, it could have possibly become the great performance RAM you want it to be. However, it was much more cost-prohibitive, and at the time, the Athlon processors (which don't need ungodly amounts of bandwidth like Netburst-based processors) did not need it, and even in the i850 based chipsets where it showed more promise, it was still so much more expensive. And then DDR improved, making it more suitable for Intel Pentium 4s. Not to mention it was cheaper with a much better Price/Performance ratio.

Unless you have serious financial interests in Rambus (as in you work for them, you own tons of their stock), I don't know how you can be so incredibly adamant at the prospects of having the JEDEC member companies being completely boned and we have to move back towards more expensive RDRAM.

 

[Terra]

Latency is higher...

Please look at post #48...lets stick with the facts okay?

...and cost blows.

And again, why would I make RAM a bottleneck in my system?

You have to realize, the AMD archetecture didn't NEED the bandwidth, and since it was winning the performance crown (overall), no one really cared about the alternative.

Back in those days the P4 2.4Ghz and the XP+ 2400 where a lot closer than Intel and AMD are today...(disregarding the Conroe)
Anandtech - AMD Athlon XP 2600/2400+ Date: August 21st, 2002

Terra...

 

[Terra]

that explains a lto.... lawsuits are a frapping joke, that's what. It means almost nothing that someone won...

Remember, someone won millions because their mcdonalds coffee was hot and they spilled it on themselves

I might not be from the US, but I do know that you are talking about a "myth" now
Legal Myths: The McDonald's "Hot Coffee" Case

Terra...

 

[GilmourD]

I dug this up, does this help you?

Terra...

Where did this come from? I could start making pictures in Windows Paint, too.

 

[pxc]

I dug this up, does this help you?
http://img163.imageshack.us/img163/9862/rdram7cm.jpg[/I MG]

Terra...[/QUOTE]
Notice the bus width on the quad channel RDRAM 1200 is only 64-bits wide (which could be done with 2 32 bit RIMMs, each is dual channel by design), while the dual channel DDR400 is 128 bits wide.

Feeding a 128 byte cache line with dual channel DDR400 takes eight times "5-7 cycles, 25-35ns" (40-56 cycles, 200-280ns). The same task takes 16 times "2-7 cycles, 10-35ns," or 32-112 cycles, 160-560ns with quad channel RDRAM 1200. Understand that and you'll see why RDRAM 1200 was DOA.

RDRAM is dead in the PC and server market because Rambus alienated most memory manufacturers with high royalty rates, shady DDR1 patents and poor implementations (yields were very bad, leading to high prices), and Intel ending chipset support didn't help :D. DDR2 gave a better solution as a successor to both SDR and DDR SDRAM, with lower power and higher bandwidth post DDR1. XDR isn't a bad product, but don't look for it outside consoles and a tiny minority of future servers.

pxc - doing the math for you

 

[therapy]

Where did this come from? I could start making pictures in Windows Paint, too.

How dare you question his graphs.

I have made a graph which will show that DDR-SDRAM has a greater Sandwiches/# of Hair Follicles ratio.

 

[Terra]

Notice the bus width on the quad channel RDRAM 1200 is only 64-bits wide (which could be done with 2 32 bit RIMMs, each is dual channel by design), while the dual channel DDR400 is 128 bits wide.

Feeding a 128 byte cache line with dual channel DDR400 takes eight times "5-7 cycles, 25-35ns" (40-56 cycles, 200-280ns). The same task takes 16 times "2-7 cycles, 10-35ns," or 32-112 cycles, 160-560ns with quad channel RDRAM 1200. Understand that and you'll see why RDRAM 1200 was DOA.

RDRAM is dead in the PC and server market because Rambus alienated most memory manufacturers with high royalty rates, shady DDR1 patents and poor implementations (yields were very bad, leading to high prices), and Intel ending chipset support didn't help . DDR2 gave a better solution as a successor to both SDR and DDR SDRAM, with lower power and higher bandwidth post DDR1. XDR isn't a bad product, but don't look for it outside consoles and a tiny minority of future servers.

pxc - doing the math for you

You did factor in that DDR400 is 400Mhz and RDRAM PC!200 is 1200Mhz right?

My RDRAM PC800 is only 16 bit..but at 800MHz

Terra...

 

[Terra]

Where did this come from? I could start making pictures in Windows Paint, too.

http://www.gamepc.com/labs/view_content.asp?id=sis659tp&page=2

Terra...

 

[Terra]

How dare you question his graphs.

I have made a graph which will show that DDR-SDRAM has a greater Sandwiches/# of Hair Follicles ratio.

.oO(Should I say anything?)
.o(Nah....)
.oO(My post #63 will make him look stupid all by it self...)

Terra - That's what you get for trying to save a site for bandwithleeching..no more "Mr. Nice guy" from me...tired of personal attacks...

 

[Least]

What's funny is this article excellently states possible problems with RDRAM that in fact, became a problem.

"The question is, will normal, everyday gamers have a chance to get this chipset, and if so, will the prices for the motherboards and memory be reasonable?"

"While more memory bandwidth is always a great thing, it's only a small piece of the performance puzzle. If the processor and/or applications do not require the excessive bandwidth, much of it may go to waste."

"The two main concerns we have are availability and compatability, as we have very little doubt that the chipset will be able to perform very well with such a high-bandwidth interface."

Good technology? Yes. Prohibitively expensive? Yes.

 

[MrGuvernment]

dam hair follicols! and sandwiches!

Did they use mustard or mayo in their test's ?

 

[CCUABIDExORxDIE]

You did factor in that DDR400 is 400Mhz and RDRAM PC!200 is 1200Mhz right?

My RDRAM PC800 is only 16 bit..but at 800MHz

Terra...

1200mhz x 16bits x 4-quad channel = 9.6gb/s

667mhz x 64bits x 2-dual channel = 10.7gb/s

and yea...ddr2-667 is gonna be the standard for AM2...so quad pumped RDRAM aint got nothing on ddr2. except higher price.

 

[CCUABIDExORxDIE]

dam hair follicols! and sandwiches!

Did they use mustard or mayo in their test's ?

they used ranch dressing...and the sandwiches were the spicy chickens from Wendies. who knows how this would have turned out with ketchup and hamburgers.

 

[Terra]

1200mhz x 16bits x 4-quad channel = 9.6gb/s

667mhz x 64bits x 2-dual channel = 10.7gb/s

and yea...ddr2-667 is gonna be the standard for AM2...so quad pumped RDRAM aint got nothing on ddr2. except higher price.

You do know that you are comparing 3 year old tech to new techology, right?
If you argument was to hold true, that would have to use the basis that RDRAM wouldn't "evolve" in 3 whole years?
Do you think that would have been a likely senario?

Terra...

 

[Terra]

What's funny is this article excellently states possible problems with RDRAM that in fact, became a problem.

"The question is, will normal, everyday gamers have a chance to get this chipset, and if so, will the prices for the motherboards and memory be reasonable?"

"While more memory bandwidth is always a great thing, it's only a small piece of the performance puzzle. If the processor and/or applications do not require the excessive bandwidth, much of it may go to waste."

"The two main concerns we have are availability and compatability, as we have very little doubt that the chipset will be able to perform very well with such a high-bandwidth interface."

Good technology? Yes. Prohibitively expensive? Yes.

If you try and vote on that page, you will se a funny poll too

Terra - the poll still works

 

[Least]

You do know that you are comparing 3 year old tech to new techology, right?
If you argument was to hold true, that would have to use the basis that RDRAM wouldn't "evolve" in 3 whole years?
Do you think that would have been a likely senario?

Terra...

Well, they'd first have to figure out how to manufacture it for a reasonable price.

 

[CCUABIDExORxDIE]

You do know that you are comparing 3 year old tech to new techology, right?
If you argument was to hold true, that would have to use the basis that RDRAM wouldn't "evolve" in 3 whole years?
Do you think that would have been a likely senario?

Terra...

yea i dont think they woulda evolved in 3 years...cause no one would shell out cash to buy RDRAM, theyd go for cheaper DDR ram.

also...ddr2 is just under 2 years old btw....so ya its onla year difference from when it launched to now.

and the big difference is..ddr2 is actually available and useable, unlike RDRAM

 

[Terra]

Well, they'd first have to figure out how to manufacture it for a reasonable price.

Supply and demand...
And again, we are talking about a 3 year old technology, that still holds up very well...even compared to DDR-II 667...

And this is next part is not directed at you, Least

But the way some people "debate" here is way off.
Personal attacks, the mentioning of my family(???), ridicule based on hot air, ect.

Terra - Like I always say, when you can't muster a real argument, don't say anything...

 

[GilmourD]

Supply and demand...
And again, we are talking about a 3 year old technology, that still holds up very well...even compared to DDR-II 667...

And this is next part is not directed at you, Least

But the way some people "debate" here is way off.
Personal attacks, the mentioning of my family(???), ridicule based on hot air, ect.

Terra - Like I always say, when you can't muster a real argument, don't say anything...

Fill in the following:

Cite an example of being personall attacked: _________________________________
Cite an example of your family being mentioned in a derogatory way: _________________________________
Cite an example of an argument wasn't backed up by pure fact: _________________________________

 

[Terra]

yea i dont think they woulda evolved in 3 years...cause no one would shell out cash to buy RDRAM, theyd go for cheaper DDR ram.

For you argument to be true, no one should ever have bourght RDRAM.
Is this the case?
Or RDAM should have stopped at RDRAM PC600
Is that the case?

also...ddr2 is just under 2 years old btw....so ya its onla year difference from when it launched to now.

But DDR-II is the the latest RAM for PC's right?

and the big difference is..ddr2 is actually available and useable, unlike RDRAM

So you are saying that I can't buy RDRAM today?
So you are saying that the RDRAM in my rig dosn't work?

Terra...

 

[Terra]

Fill in the following:

Cite an example of being personall attacked: _________________________________
Cite an example of your family being mentioned in a derogatory way: _________________________________
Cite an example of an argument wasn't backed up by pure fact: _________________________________

That would only bring this thread way off topic, and be cause for a lock
I am here to debate, not throw with mud..
My comments where meant to direct the thread back on track, but you take a close look at post #61.

Terra - Can we now get this thread back on track...with arguments please?

 

[CCUABIDExORxDIE]

For you argument to be true, no one should ever have bourght RDRAM.
Is this the case?
Or RDAM should have stopped at RDRAM PC600
Is that the case?

no one ever should have bought RDRAM! period! waste of money...ddr was just as good..and for alot cheaper. AMD knew this and didnt utilized RDRAM.

But DDR-II is the the latest RAM for PC's right?

yea it is. so? point being?

So you are saying that I can't buy RDRAM today?
So you are saying that the RDRAM in my rig dosn't work?

Terra...

you can buy RDRAM today...but itll cost you an arm and a leg and you wont be able to use it in either a LGA775 rig or AMD rig of any kind.

and the last part is debateable

 

[(cf)Eclipse]

PC3200=DDR400
Data Rate=400MHz
FSB=200MHz
Peak Bandwith=3.2 GB/s(giving 6.4GB/s i dualchannel config)
Even if we go:
PC4200=DDR533
Data Rate=533MHz
FSB=266MHz
Peak Banwith=4.2 GB/s(giving 8.4GB/s in dual channel config)

Quad channel RDRAM(Wich DID come on the market ~3years ago):
PC1200=RIMM 4800
Data Rate=1200MHz
FSB= 600MHz
Peak Bandwith=2.4 GB/s (giving 9.6GB/s in quad channel config)

In my math that still gives RDRAM an advantage of 1.2GB/s...3 years ago...

yes, it is good bandwidth. but there's more to things to performance than just bandwidth...

Both PC1200 and PC2700/PC3200 appeared around the same time, I fail to se your logic, sorry?
And PC800 PC and PC2100 where "playmates" ~4 years ago...

right... cause we all remember pc3200 modules being for sale in late '02. get your timeline straight before trying to cite it.

3200x1.4x2=9.96GB/s...wich means that dualchannel PC3200 needs and OC of 40% to exceed quad channel RDRAM...that isn't OC'ed.

nice math. dual channel ddr actually needs to be 300mhz to get a theoretical bandwidth of 9.6ghz.. which is a 50% oc. while this is even more in rdram's favor for bandwidth, it's always fun arguing numbers with someone who doesn't multiply correctly

But a P-III is unable to utilize the fullbandwith of RDRAM PC800 as the P-III is FSB 100Mhz or am I mistaken?

ok, as i said, the cpu is blind to the ram. to get a certain amount of work done, it needs the data to do it. regardless of how the cpu is set up internally, it'll still need the exact same amount of data to do the exact same amount of work.
the P4's architecture is really designed for high cpu clock speeds. check out the difference between single and dual channel ddr.. surprisingly small? guess bandwidth means jack after all
i call your attention to an old review that compares a few different things:
dual channel pc1066
dual channel ddr400
dual channel ddr266
single channel ddr333
http://www.extremetech.com/article2/0,1697,1156952,00.asp
looks like ddr trounces rdram for media encoding, which is odd because of how rdram is set up, streaming data should be it's strong point

2 things to note:
1. surprisingly small spread between all rigs, considering how focused you seem to be on bandwidth.
2. ddr400 wins the majority of the tests against what was available at the time for testing: dual channel pc1066 rdram.

"but wait!" you might chime in, "it's not quad channel pc1200". this is indeed true. let's look at a review that shows the actual performance of this setup
but wait! from what i can tell, the board never actually made it to market. if it did, the only site that reviewed it was Tom's, and i have good reason not to trust any of their data from that time period.

so some more things to point out:
read this [H] article. despite having a lot more bandwidth on hand, the dual channel pc1066 rdram rig barely does better than it's single channel ddr333 counterpart (compare the abit board to the i850E)
http://www.hardocp.com/article.html?art=MzUyLDEsLGhlbnRodXNpYXN0

No need to..I remember it quite, well as I purchased my system bacj then and browsed to find the best solution at the time.
RDRAM PC800 vs DDR PC2100...
RDRAM PC1066 vs. DDR2700...
RDRAM PC1200 vs DDR3200...Where quad channel gave RDRAM the upperhand...

not sure if i really want to go over this again. pc3200 competed against pc1066, and quad channel really didn't make it to market.

Being greedy isn't illegal, it just gets you busted, like what actually happened. If it had been the better tech, it would have won, and we'd be using it now.

exactly. enthusiasts have proven time and time again that they'll pay out the ass for better performance. if rdram really was better, why did it not beat ddr in terms of sales and success?

and yea...ddr2-667 is gonna be the standard for AM2...so quad pumped RDRAM aint got nothing on ddr2. except higher price.

actually, i'm 99% certain that ddr2-800 is the standard. 12.4gb/s and lower access latency ftw

And again, why would I make RAM a bottleneck in my system?

it's honestly surprisingly hard to make memory a bottleneck in the system, unless you don't have enough

 

[pxc]

You did factor in that DDR400 is 400Mhz and RDRAM PC!200 is 1200Mhz right?

My RDRAM PC800 is only 16 bit..but at 800MHz

Your chart already has those figures. Why would you post it if you didn't understand why both clocks and time were included?

For example, DDR400 (200MHz DDR) with a 5 cycle transaction time = 5 * 1/200MHz = 25ns. Same with a 7 cycle transaction time: 7 * 1/200MHz = 35ns. Do those numbers look familar yet?

RDRAM 1200 with a 2-7 clock transaction takes 10-35ns (in RAMBUS speak, the 30-32 "timing bin", a suffix after the RDRAM 1200 module part number). Unfortunately, a cache line fill requires twice as many quad channel 64-bit interface RDRAM transactions as a dual channel 128-bit DDR interface, which makes RDRAM look and perform downright bad. Why isn't RDRAM as fast as DDR could be at 1200MHz (ignoring the bus width)? The packet based transfer method introduces additional latencies.

BTW, PC800 latency is horrible. It's around 67.5ns for the first access, much, much worse than DDR1. Dual channel PC800 is inferior in every way to single channel PC3200 except theoretical peak bandwidth which are equal. See how that worked out in real life applications.

I'm all for loving your hardware and all, but RDRAM was a failure after dual channel DDR was supported. And for you to pick on a virtually unused version of it as proof of how good it was is really stretching it.

 

[(cf)Eclipse]

Unfortunately, a cache line fill requires twice as many quad channel 64-bit interface RDRAM transactions as a dual channel 128-bit DDR interface

i applaud you in your knowledge of how the low-level stuff actually works in ram.. something i definitly need to brush up on. can i ask you a favor, and either post here or send me a pm explaining how you go about figuring out that kind of stuff?