Poll: Who is more popular to bash these days: 3dfx or Intel?

[Quickfingerz]

this is an original poll that ended up being a intel bashing poll.

 

[Sunner]

PCR, do you seriously think its fair to compare the EV7 to a P3 or P4 system?

8 channel RDRAM is nice and all, but lets face it, it aint gonna happen in PC's cause its too friggin expensive.

I might just as well say that Intel's servers all suck cause they cant match the performance of a full blown Sun E10K.
Sure, the E10K will cost quite a few arms and legs, and probably your more precious parts as well, but it is faster.


To get back to the topic at hand, if RDRAM indeed needs 8 channel interleave to show its potential, then it just aint for PC's.
I do think that it will be interesting to see what kind of performance the EV7 will bring though, but whatever it is, it wont be cheap.

 

[superbaby]

It's not Intel's fault that Rambus doesn't perform well.

Apparently the engineers at IAG (Intel Architecture Group) said that the Rambus spec was very very restrictive and they actually had to work around it, not work with it. This explains why the MTH was needed and why there were so many problems with the i820 chipsets.

Also the correlation between the 133FSB and 600/800mhz bus on the the RDRAMs isn't exactly valid. You are thinking in terms of the BX chipset, where the CPU FSB is directly related to the speed the memory operates at. Like the EV6 I believe the i820 works in somewhat of a similar fashion - the FSB and memory bus are independent of each other.

 

[PCResources]

<< PCR, do you seriously think its fair to compare the EV7 to a P3 or P4 system? >>



I have always stated that i do not think RB is a good solution for desktops, but for high-end servers and workstations, i think it's the future.



<< I might just as well say that Intel's servers all suck cause they cant match the performance of a full blown Sun E10K. >>



I never said that all Intel's servers suck, it is the use of RB in Intels implementation i think suck (but maybe they can get it right with the P4).



<< Apparently the engineers at IAG (Intel Architecture Group) said that the Rambus spec was very very restrictive and they actually had to work around it, not work with it. This explains why the MTH was needed and why there were so many problems with the i820 chipsets. >>



So the problems with the i820 was because of RB? If i remember correctly the problems occured when you didn't use RB on the i820. If intel didn't believe that RB was a good solution, that it is too restrictive and that they couldn't work &quot;with it&quot; but had to work &quot;around it&quot; then why did they ever even consider it? Again, this only shows that the Intel implementation of RB wasn't a very good one.



<< Also the correlation between the 133FSB and 600/800mhz bus on the the RDRAMs isn't exactly valid. You are thinking in terms of the BX chipset, where the CPU FSB is directly related to the speed the memory operates at. Like the EV6 I believe the i820 works in somewhat of a similar fashion - the FSB and memory bus are independent of each other. >>



This is all fine as long as the FSB is FASTER than the memory bus, like the EV6 is faster than the PC133 modules, and therefore it performs better at an async 133 than a sync 100 memory bus.

But when the FSB is slower than the memory bus, that's another story. If the CPU can only access the memory at 133 it really doesn't matter if you have 200 DDR, 400 DDRII, 800 RB or whatever, now does it?


I do NOT believe that RB is the way to go for desktops, low-end/middle range servers or workstations (at least not yet) but there are situations when you really could use the performance of a multi-channel RB bus. I will be using the A21364 for one of my clients as soon as i can get hold of one, because he really, really needs the extra memory performance.

Patrick Palm

Am speaking for PC Resources

 

[superbaby]

PC Resources,

Where do you get your information from? The MTH (for SDRAM) in the i820 was added AFTER the issues with RDRAM occurred, which is why there are problems with it. There are problems with and without RDRAM on the i820, same with SDRAM. It was added because RDRAM at the time caused the system to be unstable blah blah blah. As a fix Intel released the MTH to support SDRAM which actually DID work after several updates.

As for Intel's supposedly crappy implementation of RDRAM:

http://www.electronicnews.com/enews/Issue/FreeIssues/2000/10232000/0110231f-1.asp



<< &quot;The issues were not defects within the MTH. The issues were with the Rambus channel itself and the use of large packages at channel speeds. Technically, the problem has been with microwave-like resonance effects in the component packages, connectors and in the structures formed by these when placed on printed circuit boards.&quot; >>

 

[PCResources]

<< Where do you get your information from? The MTH (for SDRAM) in the i820 was added AFTER the issues with RDRAM occurred, which is why there are problems with it. There are problems with and without RDRAM on the i820, same with SDRAM. It was added because RDRAM at the time caused the system to be unstable blah blah blah. As a fix Intel released the MTH to support SDRAM which actually DID work after several updates. >>



I stand corrected. There was a problem with RB and i820, but why was there a problem? Because of the chipset or was it the RB that was flawed?

So Intel didn't know how to implement RB correctly, well that's what i said, wasn't it?

Intel made a mistake with RB, i agree, BUT Intel made the mistake, not RB. Making a half-assed attempt to design a high speed memory bus is a mistake, and Intel DID make it. You simply cannot take an expensive, high speed memory, run it with a CPU that does not take advantage of the speed and expect to get a good performer at a reasonable price.

Patrick Palm

Am speaking for PC Resources

 

[superbaby]

<< Making a half-assed attempt to design a high speed memory bus is a mistake, and Intel DID make it. >>



If you read the article more carefully you will see *why* it was not implemented properly. It was the strict RDRAM design that caused Intel's engineers to be unable to develop a solid chipset. It's like developing an application that say uses TCP/IP connections which involves a third party package, BUT it only accepts IPX/SPX connections. So you now have to somehow make the TCP/IP talk to the third party package that can only understand IPX/SPX. This is probably an oversimplification of the conflict that the Intel engineers faced, but you get the idea.

I don't believe it was a half-assed attempt. The MTH probably has some interesting implementation details to make SDRAM communicate to a chipset that only speaks Rambus.

It's not a *mistake* on the implementation of the i820 chipset as much as it was a mistake on the Intel execs, asking the Intel engineers for something unfeasible. The fault also can be traced back to Rambus for giving such a rigid spec and for being assholes.



<< You simply cannot take an expensive, high speed memory, run it with a CPU that does not take advantage of the speed and expect to get a good performer at a reasonable price. >>



It certainly is expensive. High-speed? That still remains to be seen, I have yet to see any benchmarks on any platform that make Rambus a clear performance winner.

 

[PCResources]

<< If you read the article more carefully you will see *why* it was not implemented properly. It was the strict RDRAM design that caused Intel's engineers to be unable to develop a solid chipset. It's like developing an application that say uses TCP/IP connections which involves a third party package, BUT it only accepts IPX/SPX connections. So you now have to somehow make the TCP/IP talk to the third party package that can only understand IPX/SPX. This is probably an oversimplification of the conflict that the Intel engineers faced, but you get the idea. >>



Yes, i do know what you are talking about, i am an EE you know... But if Intel saw this coming, they knew they couldn't make a decent chipset for RB (this is what you are saying, right?) then why didn't they just can it?

Compare it to the application you mentioned, who would be dumb enough to create that application? So what you are saying is that Intel's engineers are real stupid?

Please understand this: Intel's CPU PIII doesn't have the performance to saturate the memory bus of the RB mem, i do not care what you think about this, this is not an opinion, it is a FACT, just stating anything else would be real stupid. The implementation of RB that Intel did was never good, can you honestly say that you believe that putting an 800Mhz memory with pretty high latency on a 133Mhz FSB is a good choice??



<< It's not a *mistake* on the implementation of the i820 chipset as much as it was a mistake on the Intel execs, asking the Intel engineers for something unfeasible >>



Finally we agree on something, it was stupid of Intels execs to think that the RB would ever come to any real use with the PIII.





<< It certainly is expensive. High-speed? That still remains to be seen, I have yet to see any benchmarks on any platform that make Rambus a clear performance winner. >>



You will, you will.



<< The fault also can be traced back to Rambus for giving such a rigid spec and for being assholes. >>



Sure, it is all RB's fault that Intel couldn't make a decent chipset to take advantage of the high speed bus of the RB, isn't it?

About RB being assholes, well, i guess that is what it all comes down to, it is RB's fault, not because their technology is bad, but because everyone thinks that they are assholes. If you think RB are assholes because of their lawsuit, then what do you think of Intel after their lawsuit against VIA?

Just because RB is behaving like a spoilt child (they are not alone about this), it doesn't mean that their tech isn't valid at all, now does it?

Patrick Palm

Am speaking for PC Resources

 

[superbaby]

<< But if Intel saw this coming, they knew they couldn't make a decent chipset for RB (this is what you are saying, right?) then why didn't they just can it? >>


Again refer to the article... the Intel execs didn't listen to their engineers and pushed for the RDRAM implementation despite warnings that there would be problems.


<< Compare it to the application you mentioned, who would be dumb enough to create that application? So what you are saying is that Intel's engineers are real stupid? >>



No they aren't stupid, I highly doubt they are, but they *were* silenced by upper management. &quot;Disagree and conform&quot; I believe it was. As I said... that example was just to illustrate my point. And I agree, it's a pretty dumb application in concept (and sigh I've had to deal with crap like that at the office).


<< The implementation of RB that Intel did was never good, can you honestly say that you believe that putting an 800Mhz memory with pretty high latency on a 133Mhz FSB is a good choice?? >>


The &quot;bad&quot; implementation has nothing to do with matching the memory with the FSB. It's not a performance &quot;mismatch&quot;, regardless of the bandwidth - Rambus is inherently slow. The fact that RDRAMs have high latency speaks for itself. Since the CPU can't take advantage of the high bandwidth there is NO advantage to RDRAM at all. For example PC800 RDRAM actually runs on a 400hz bus and with a DDR approach, and the bus width has to be reduced to 16 bits, which means incoming data has to be packeted and serialized. This further adds to the latency of RDRAM.

While you keep maintaining that RDRAM is good for server configurations I have to disagree. Rambus technology is nothing more than complex tweaks to the DDR SDRAM concept, and the company also holds the patents to keep manufacturers from producing lower latency DDR SDRAM which has equivalent bandwidth to RDRAM *on paper*. Why do you think Rambus is so scared of DDR SDRAM?


<< Just because RB is behaving like a spoilt child (they are not alone about this), it doesn't mean that their tech isn't valid at all, now does it? >>


I'm not pro-Intel and I agree that their decision to move against VIA was pretty dumb. While VIA has some pretty nifty chipsets who can dispute that they have much more stability problems then say, the 440BX and i815? However the difference between Intel and Rambus is that Intel does creates products that are still competitive, whereas Rambus released a horrible memory spec and instead of improving on it they go around collecting royalties and suing manufacturers. Don't tell me RDRAM memory is worth salvaging, because it's cleverly disguised and crippled DDR SDRAM.

Man that post took awhile

 

[PCResources]

<< Again refer to the article... the Intel execs didn't listen to their engineers and pushed for the RDRAM implementation despite warnings that there would be problems. >>



Ok, but what does this mean? it means that Intel has people deciding about things they really have NO clue about. It doesn't matter how many ways you try to put it, you are just saying the same thing. Intel made a bad implementation, in a system that was in no way ready for it. I don't know wich is worst, EE's that don't know their job, or EE's that are not allowed to do their job.



<< Since the CPU can't take advantage of the high bandwidth there is NO advantage to RDRAM at all >>



Two questions:

1. Why can't the CPU take advantage of the high bandwidth, could it be because the FSB is to slow for the CPU you are thinking about?

2. Wich CPU are you talking about?? The CPU that i already stated was no good with the RB?

You keep saying that the way that RB was used by Intel is no good, and i keep agreeing.



<< While you keep maintaining that RDRAM is good for server configurations I have to disagree. Rambus technology is nothing more than complex tweaks to the DDR SDRAM concept, and the company also holds the patents to keep manufacturers from producing lower latency DDR SDRAM which has equivalent bandwidth to RDRAM *on paper*. Why do you think Rambus is so scared of DDR SDRAM? >>



&quot;complex tweaks to the DDR SDRAM concept&quot;, oh christ, i was going to give a long answer to that one, but as you obviously do not know anything about these techs so i'll save my breath.

&quot;keep manufacturers from producing lower latency DDR SDRAM&quot;, no, no, no. how does the patents Rambus have prevent Manufacturers from producing DDR SDRAM with lower latency than 2*CAS2?

And even if they could make the DDR SDRAM a lot faster, i doubt they could reach 12.6 - 16GB transfer rates, wich RB does in Alphas configuration.



<< Don't tell me RDRAM memory is worth salvaging, because it's cleverly disguised and crippled DDR SDRAM. >>



Ok, so you do not know the difference between RBDRAM and DDR SDRAM? well, that does it for me.

Just stop your bashing for a while and think about this: I will build a server based on the A21364 with RBDRAM, it will use 8 channels and it will have an on-chip memory controller.

The on-chip memory controller helps keep the latency to a minimum, and the 8 channels would give me 12.6-16GB transfer rates (depending on wether i use 800Mhz or 1Ghz memory).

Now i really need that extra memory speed, because the system used today is way to slow, all because of the memory performance, i upgraded the cpu from 500-1Ghz and noticed a 2% difference, so i checked it out, and it clearly shows that the memory usage is high, and that the CPU has to wait for the memory to deliver all the time.

The RB solution will solve my problems, and for server that need high memory bandwidth it will be a great solution. No matter how much you dislike it, that is the way it is.


Patrick Palm

Am speaking for PC Resources

 

[Fozzie]

Well I'm coming in a little late, didn't see this topic was still going. I don't really understand you &quot;PCResources&quot;, you act arrogant and attagonistic. With that attitude you should be posting in the Ars forums(J/K for you Ars-rs out there).

&quot;Next time try to understand what i write instead of just trying to bash me.&quot;

Please. Don't flatter yourself, you should get a shrink to treat you for that superiority complex.

I do not think that the FSB is not a limiter in the majority of cases with the current Rambus OR SDRAM implimentation. And you certainly haven't shown any evidence to the contrary. First off benchmarks show memory performance well under the theoretical limit with both Rambus and PC133 in the vast majority of situations. If PC133 averages only 500MB/s tops in most situations shouldn't Rambus, even if it is limited by the &quot;slow&quot; 1GB FSB still outperform the SDR? Not only that but how do you explain the performance gain of the i840 over the i820?

&quot;with a built in memory controller the latency would be lower, and with eight channels? well SDRAM wouldn?t stand a chance, this is the configuration I am talking about.&quot;

The only thing you brought up was a EV7 implimentation. Which means squat in the PC world, 128bit DDR @ 166 or 200MHz is far more likely to be seen in the near future(which means little). So don't try to limit it back down to 64bits when your pulling Alpha numbers out of the air. Shall we start including IBM and others mainframe SDRAM implimentations?


&quot;Nope, i don't think so. But from where did you get the extra 64bits? This is NOT an easy thing to do, the pin count would be VERY high, so don't count on getting that 128bit DDR (at 400Mhz it would be DDR-II).&quot;

What are you saying here? That 128bit DDR is impossible? Or that it isn't likely to ever be PC system ram? Surely you know that 128bit DDR implimentations already exist for graphics cards?

&quot;And even if they could make the DDR SDRAM a lot faster, i doubt they could reach 12.6 - 16GB transfer rates, wich RB does in Alphas configuration.&quot;

Again why bring up the Alpha? If someone like IBM wanted to design a 256bit 500MHz DDR setup they most certainly could. And if you wanted to check the bandwidth figures on that you'd see that it's quite competitive. Let alone look at their current workstation, or the upcoming Power4. But what does this have to do with the current PC world? If I was a Car enthusiast why shouldn't I disagree/bash someone who recommended prop engines?

What is there to argue here? I think that in time we'll see proper implimentations of both Rambus and DDR. Which will be the victor? Who knows? But your trumpeting of the EV7 really means squat. I think most people bash Rambus, INC MAINLY, and Rambus the technology second. The reasons most people disliked Rambus - the Ram, is because it was slower and more expensive in the currently available form. What was there to like?

Sure the technology may have promise. But its certainly not the only fish in the lake. And if Rambus, Inc stays its current course I hope both the company and the technology die a horrible death.

 

[PCResources]

<< What are you saying here? That 128bit DDR is impossible? Or that it isn't likely to ever be PC system ram? Surely you know that 128bit DDR implimentations already exist for graphics cards? >>



The graphic cards memory subsystem and the CPU memory subsystem are vastly different, you obviously do not know this, but it is.

As for being rude to Superbaby with that comment, that was my answer to his post about me contradicting myself, he obviously hadn't read the post properly, i don't think he took offense as he never mentioned it again, he probably just went back and read the post over again and understood what i meant, i suggest you do the same.

Alpha is not your average desktop CPU, and i have never ever ever stated that i think so, and i do not believe that RB is for your average desktop computer either, but i do believe that when i can get my hands on one of the Alpha 21364's my clients memory performance problem will be solved.



<< The reasons most people disliked Rambus - the Ram, is because it was slower and more expensive in the currently available form. What was there to like? >>



Again, i am agreeing, the way that RB was implemented by Intel (the only implementation sofar) really sucks, it WAS slower than SDRAM in intels implementation, i have never stated anything else.



<< If PC133 averages only 500MB/s tops in most situations shouldn't Rambus, even if it is limited by the &quot;slow&quot; 1GB FSB still outperform the SDR? Not only that but how do you explain the performance gain of the i840 over the i820? >>



If your system tops 500MB/s SDRAM would be a better choice, but that is NOT the situation that i am talking about, i speak of high-end servers with the craving for higher memory bandwidth than DDR can offer.

The performance gain of the i840 would be because it has two memory channels, so adding channels does improve performance, just as i have stated before.

Now if you were to add channels with DDR, and integrating the controller on-chip, then wouldn't you get a performance gain? Sure, but adding channels with DDR will result in two things, a very high pin-count and it would be very sensitive, requiring an eight layer mobo, that would indeed be an expensive solution. And to integrate the 128bit controller on the CPU chip? Well, NO, that will not be done.

Patrick Palm

Am speaking for PC Resources

 

[superbaby]

PC Resources,

Please read this article:
http://www6.tomshardware.com/mainboard/00q1/000315/rambus-02.html

It contradicts everything you have posted.

 

[PCResources]

Oh, christ, not again.

The article states that if you implement RB the way it has been done, you would be better off with SDRAM or DDR SDRAM... I totally agree with this, there is a solution to every problem about RB mentioned in Tom's article, and the A21364 will use everyone of these solutions.

Once upon a time, a company called intel released it's first SDRAM chipset (it was called VX). Now this chipset with this new fantastic memory should be real fast, but when they tested it and compared it to another chipset using EDO ram (HX), it wasn't faster. So should we have ditched SDRAM then? Nope, and we shouldn't ditch RBDRAM now either.

Patrick Palm

Am speaking for PC Resources

 

[Priit]

Um, I think 3Dfx is pretty O.K., Intel isn't fully crap, just too expensive. Company, who IMHO deservs most bashing is Microsoft. Last thing they made almost O.K. was DOS 6.22 (and even that was mostly re-written CP/M &amp; UNIX).

 

[superbaby]

Yes SDRAM at that time was slower than EDO DRAM at the time. However SDRAM, as it was then, is open source, as it is now. Rambus isn't and if it were priced the same as SDRAM then this dicussion would be moot.

Regardless of the so-called bandwidth advantage of RDRAM, have you considered the many other factors that plague Rambus? That you need to have place holders in empty RIMM slots because the RIMM array bank runs SERIAL (you know how much these place holders cost? like $50 for a stupid PCB without the RDRAM BGA packages)? Did you know that different latencies of RDRAM can cause the system instabilities? PC800 RDRAM has a bandwidth of something around 1.6GB/s while DDR PC266 SDRAM has a bandwith of > 2GB/s.

Given that Rambus Inc. is also NOT in that Future Memory Technologies Alliance thingy, their technology will never be salvaged.

Your Alpha EV7 server was supposed to be released Q3 2000, we're in Q4 2000 and no EV7. It's also going to be initialliy released with SDRAM/DDR SDRAM configurations, NOT Direct Rambus. I figure you have a long time to wait before you can even put a deposit down on one of those overpriced snails.

 

[fkloster]

<< Regardless of the so-called bandwidth advantage of RDRAM, have you considered the many other factors that plague Rambus? That you need to have place holders in empty RIMM slots because the RIMM array bank runs SERIAL (you know how much these place holders cost? like $50 for a stupid PCB without the RDRAM BGA packages)? Did you know that different latencies of RDRAM can cause the system instabilities? PC800 RDRAM has a bandwidth of something around 1.6GB/s while DDR PC266 SDRAM has a bandwith of > 2GB/ >>



1)Factors that plague rambus? Its serial nature is NOT one of them. &quot;Place Holders?&quot; No one complains about termination in SCSI technology. Termination is no different in serial memory architecture... $50 bucks my arse!! Asus gave me as many as I wanted for free! Give me a call I got 6 of those suckers. That is crap!!

2) Instabilities because of different latency's is crap as well. I have been running 2-128mb rimms stable for 10 months with 0% stablity problems &amp; I should know! I made the mistake of buying this expensive crap back when it first came out!

3) Superbaby, your camparison of bandwidth between PC-800 on a single channel compared to PC-266 DDR is like comparing apples &amp; oranges. A level comparison is using rambus on the i840 w/3.2gb of bandwidth compared to DDR's 2.1 GB(?)



<< Yes SDRAM at that time was slower than EDO DRAM at the time. However SDRAM, as it was then, is open source, as it is now. Rambus isn't and if it were priced the same as SDRAM then this dicussion would be moot. >>



Superbaby, impress me by going to pricewatch and inform the public of just how much a premium rambus is commanding @ this very moment...



Superbaby you have really been getting on my nerves lately. Chill out on serial memory until you get your facts straight.

 

[superbaby]

What did I say wrong about the serial technology? And why would Asus give you free RIMM placeholders, and why would you need 6? You don't have 8 banks of RIMM slots.

<< Instabilities because of different latency's is crap as well. I have been running 2-128mb rimms stable for 10 months with 0% stablity problems >>


You obviously don't have RIMMs with different latencies so this would not apply to you.


<< Superbaby, your camparison of bandwidth between PC-800 on a single channel compared to PC-266 DDR is like comparing apples &amp; oranges. A level comparison is using rambus on the i840 w/3.2gb of bandwidth compared to DDR's 2.1 GB(?) >>


You want comparisons?

Sysmark 2000:
http://www1.tomshardware.com/mainboard/00q4/001030/athlon-15.html

DivX4 MPEG encoding (something RDRAM should do very well *considering* that big big bandwidth):
http://www1.tomshardware.com/mainboard/00q4/001030/athlon-18.html

In both cases RDRAM comes out performing worse than DDR SDRAM. And once again all that touted memory bandwidth is just vaporware (or vaporspecs in this case).

Lowest memory price on pricewatch:
$115 - 128MB PC700 RDRAM

Specs of PC700 RDRAM:
RDRAM PC700 356 MHz (External Clock-speed) 712 MHzm (Effective Clock-speed) 1.424 GB/sec bandwidth

(Oh BTW I wouldn't jump at this PC 700 price if you were gonna add this to your system, it would cause instabilities! PC800 RDRAM for you system, 128MB costs $299, but you still have to be careful... if it has different latencies you'll crash! And since you have SIX place holders... man if you were to buy 4 more RDRAMs at 128mb each... that would be like $1200...)

In a dual channel setup of PC700 you would have an effective bandwidth of 2.848GB/s, compared to 2.128 GB/s to single channel PC266 DDR SDRAM. A 0.72GB/s gain over DDR SDRAM. If bandwidth is all important as you like to point out, how come this ~33% gain does NOT show up in real world tests (refer to links above)? 33%! I mean, that's a LOT more bandwidth!


My facts are straight and your nerves don't have anything to do with this discussion.

 

[fkloster]

TOMS HARDWARE IS A BIASED/AMD SUCKING/PROPAGANDA SPEWING SITE!

Superbaby, anytime you want to outfit a ref. i850 board with 2-PC 800 rimms striped to a max/bandwidth of 3.2 Gig's on a 400mhz FSB and compare it to an AMD 760 chipset w/266mhz DDR memory then I will be all ears. But please don't try to prove anything to people by pulling links from that bozo's site...

 

[superbaby]

i850? I thought we were talking about the i840 here. Is the i850 even released? And where are the 400 mhz FSB processors? The latest P4's have 400 mhz FSB but if you haven't heard, there's no motherboards to support them, haha So your request for a performance comparison is rather uneducated.

I didn't find Tom's articles biased, which is why I posted links to them. I mean, they explained everything and backed everything up with results, instead of of spewing PR specs like you have. Please point me to where they are biased, other than just calling them biased... that's just dismissing them without even looking. I doubt you read those articles that fast.

 

[fkloster]

Is the 760 even released? (i840 would be faster than any chipset Via has currently out, assuming it is paired with PC-800 striped)

BTW, when are you going to inform the public where you heard that outlandish rumor about the $50 rimm termination parts? &amp; this crap about different latency's with different rimms? All BS!

 

[PCResources]

<< TOMS HARDWARE IS A BIASED/AMD SUCKING/PROPAGANDA SPEWING SITE! >>



No, i don't think so, but by now everyone has understood that i really don't like RB with Intel. And that is what Tom Pabst is saying too.

I think he might just change his tune once the A21364 is out.

Superbaby: you are making comparisons you cannot possibly win, you are doing everything you can to discredit the use of RB in Intels systems, and i fully agree with you, the RB is not a bad tech, it just needs the right implementation.

But to think that RB is all bad, i think you will se a serial packeted interface in the future, it might or might not be RB tech, but right now RB is the only manufacturer of this type of memory, usable in the REALLY high end.

Patrick Palm

Am speaking for PC Resources

 

[superbaby]

<< when are you going to inform the public where you heard that outlandish rumor about the $50 rimm termination parts? &amp; this crap about different latency's with different rimms >>


You make it sound like I'm preaching the gospel or something. I'm simply trying to have a discussion and you are turning it into something personal. FYI I got my information from my local DBA who administers 2 test servers which use RDRAM. And the different latency issue can be found in the links I posted above. Please, just read before you reply.


<< s the 760 even released? >>


And yes, the 760 is released. You can't buy it now because the motherboards are in production, it is NOT vaporware, that much is certain. You didn't read the articles I linked, otherwise you wouldn't sound to dumb right now. Also it's not VIA's chipset, it is AMD's. Simply dismissing them because you *think* Tom is biased then you are no better than him.


<< (i840 would be faster than any chipset Via has currently out, assuming it is paired with PC-800 striped) >>


Sigh, please read... read... stop posting ignorant slander.


<< I think he might just change his tune once the A21364 is out. >>


Yes, if it does turn out to be a performer then of course, but all available current evidence points in the other direction.

And no way am I saying serial technology is bad. Have you heard of Serial ATA? That technology is serialized and I can't wait to get that in my computer - imagine, the end of inflexible, space wasting IDE cables. Serial memory probably will take off in the future too, just not Rambus'.

 

[fkloster]

<< You make it sound like I'm preaching the gospel or something >>



Oh, so I guess its o.k. to lie. Cool! (idiot) BTW Last time I checked, i850 boards are also in production &amp; will be available in a couple of weeks. P4 chips are IN STOCK / 1.2 tbirds ar NOT! And because you don't no how to read, I will reiterate: Do not post any links to that site if you want any credibility w/me...

 

[superbaby]

Where did I lie? Name one thing that I posted was wrong. And when did I say anything about AMD Thunderbirds? Where did THAT come from? I never questioned the availbility of the i850, I merely asked if they were out.

And to calm your &quot;nerves&quot; I don't care your credibility because you have none. I would never want to be associated with you.

http://forums.anandtech.com/messageview.cfm?catid=27&amp;threadid=265811

The last person who posted in reply to you laid out a very organized argument which you conveniently ignored. And he has said exactly the same things I have said.

In regards to Tom's Hardware... what exactly is he so biased against? Last I heard it was towards Nvidia, and he was PRO Intel, not AMD, he even has articles with are PRO Rambus and AGAINST Rambus. I don't see how he is biased on this subject, unless you would like to point me to any evidence of it.

If you would like a link, which outlines my entire argument, but not from Tom's, check it out from our very own AnandTech...

http://www.anandtech.com/showdoc.html?i=1239&amp;p=1

Read it and weep...