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

[fkloster]

Just because I am lazy does not mean I don't pay attention.
1) First Lie: you said you didn't lie
2) Second Lie: you said rimm terminators cost $50
3) Third Lie: you said there is instability with different rimms & latencys (????)

not necessarily in that order

Thats why I got involved! (oh & thnks for the vote, if girls like you had seeds enough to put your rate in your profile we would all be able to rate you too!)

The reason I can't respond to Zippy is he will get me banned I am a reformed rambus zealot, true, but I don't lie. The technology is great, the company....jury is still out... Rambus has a bad enough rap around this BBS without people like you spewing crap that isn't true.

 

[superbaby]

I never lied. You want me to lie but good luck trying.

This is where I got the RIMM placeholder information:


<< FYI I got my information from my local DBA who administers 2 test servers which use RDRAM >>


Our supplier is http://www.beamscope.com/, feel free to give them a call to confirm the price.

The instabilities of RDRAM w/ different latencies:
http://www6.tomshardware.com/mainboard/00q1/000315/rambus-02.html

Any other lies you want me to say? Bring it on, you're looking dumber all the time. You should see the grin on my face.

 

[fkloster]

More Tomshardware quotes? All right read on my friend. (two can play your qouting game) You need a fresh prospective. &amp; enough with the insults. I grow tired of this.



<< Although component latency is an important factor in system performance, system latency is even more important, since it is system latency that reduces overall performance. System latency is determined by adding external address and data delays to the component latency. For PCs, the system latency is measured as the time to return 32-bytes of data, also referred to as the 'cache line fill' data, to the CPU.

In a system, SDRAMs suffer from what is known as the two-cycle addressing problem. The address must be driven for two clock cycles (20ns at 100 MHz) in order to provide time for the signals to settle on the SDRAM's highly loaded address bus. After the two-cycle address delay and the component delay, three more clocks are required to return the 32 bytes of data. The system latency of PC100 and PC133 SDRAM add five clocks to the component latency. The total SDRAM system latency is:

40 + (2 x 10) + (3 x 10) = 90ns for PC100 SDRAM
45 + (2 x 7.5) + (3 x 7.5) = 82.5ns for PC133 SDRAM

The superior electrical characteristics of a RDRAM eliminate the two-cycle addressing problem, requiring only 10ns to drive the address to the RDRAM. The 32 bytes of data are transferred back to the CPU at 1.6 GB/second, which works out to be 18.75ns. Adding in the component latency, the RDRAM system latency is:

38.75 + 10 + 18.75 = 67.5ns for PC800 RDRAM

Measured at either the component or system level, RDRAMs have the fastest latency. Surprisingly, due to the mismatch between its interface and core timing, the PC133 SDRAM latency is significantly higher than the PC100 SDRAM. The RDRAM's low latency coupled with its 1.6 gigabyte per second bandwidth provides the highest possible sustained system performance.

From a performance point of view we must note that L1 and L2 cache hits and misses contribute greatly to memory architecture performance. Also, individual programs vary in memory use and so have different impacts on its performance. For example, a program that uses random database search using a large chunk of memory will 'thrash' the caches, and the memory architecture having the lowest latency will have the advantage. On the other hand, large sequential memory transfers with little requirement for CPU processing can easily saturate SDRAM bandwidth. RDRAM will have an advantage here with its higher bandwidth. For code that fits nicely within the L1/L2 caches, memory type will have virtually no impact at all. >>



...take the blinders off for a while. Oh, and just because one person wants to pork people on rimm terminators, does not mean that person sets the market...even though there is no market, they can be had for cents...

 

[superbaby]

flokster,

You didn't even read the continuing part of this article:
http://www.hardwarecentral.com/hardwarecentral/reports/1686/5/

Nor did you bother to visit their forums and see how badly the author got flamed for failing to include other information.

The thread with the discussion:
http://discussions.hardwarecentral.com/Forum15/HTML/000045.html

Read that and tell me if you can still call this article objective. The article is also very superficial.

Here are some DDR-SDRAM vs. RDRAM articles that might be beyond your understanding but at least they aren't generalized:

http://www.inqst.com/ddrv840.htm

http://www.dccworkstation.com/htm/articles_0400/DDRvsDC_RDRAM_2.htm



<< In the overwhelming majority of cases, DDR exceeds the performance of dual channel RDRAM, at times by a very substantial margin. There are several cases where there is very little difference, and finally a few where the 840 pulls ahead by a small margin.

Based on this rather broad mix of applications and benchmarks, Micron?s single channel (64-bit) DDR implementation must be declared the performances winner over Intel?s dual channel RDRAM platform. >>




Keep it coming.

 

[superbaby]

Oh so now I'm not really a liar, I just happen to have a different supplier that charges for placeholders?

Your integrity is so weak, so very very weak. I can't keep this smirk off my face.

 

[fkloster]

I have integrity &amp; alot of it. I pride myself on it. Everyone on this BBS knows that above all else, I have integrity and am committed to my beliefs &amp; your insults &amp; presumptions will not deter (sp) me. Say what you will. I have read that article LONG before you were a member here.

 

[superbaby]

You keep referring to this discussion as some sort of public announcement center. This is between you and me, I don't know a lot about the other forum members and I highly doubt they know you well enough either to confirm your character.

You said yourself you were a &quot;reformed Rambus zealot&quot;. That is the anti-christ of integrity if there ever were such a thing.

 

[fkloster]

They know me. Ask around....

 

[superbaby]

Aside from posting a useless question to the forum members asking about your lack of integrity, am I still a liar or have I proved that what I have said is correct?

And back to the RDRAM subject I believed I have provided sufficient hard data to back up every one of my arguments. This discussion is over.

 

[fkloster]

You have proved nothing.
I take back my statement about you being a liar (although you may very well be one) No one I know sells rimm terminators for $50 bones &amp; that latency stability line you pulled from Mr. Pabst is crap. I should know. I have used rambus much longer than the doctor. It is possible that you interpreted these things in your own special way. This is over.

 

[superbaby]

You are as stubborn as a rock. Have fun with your stable RDRAM setup, knowing that you paid so much more for it makes spending $50 for placeholders seem like pennies - don't give me that Pricewatch crap because you bought it when RDRAM was at it's premium. Actually I want to wave that price in your face... HEY $115 for 128MB PC700 RDRAM, how much did you pay for yours? haha... How do you like running an overpriced snail of a system? Regarding the latency I took the quote off the site word for word, how did I interpret it differently? I speak English, and you? And you obviously have two sticks of the same RDRAM with the same latency, you cannot use your system as a comparison at all.

I know I said it was over but you asked for it.

 

[fkloster]

1) I am very stubborn.
2) You are correct, I paid a huge premium back in january for my rig and I'm sorry for it. I had the cash to blow and I knew I could carry over my memory to my future i850 platform without penalty. I believed rdram was the future &amp;, perhaps, I was wrong. But I love the technology. I am a very strong believer in serial interfaces, much like you. I took it to the next level. I should have waited.
3) My system is far from being a snail. i820 w/MTH=BAD i820 w/RIMMS=Stable platform that does not allow PC-800 to stretch its legs. I own the flagship of all i820 motherboards. No, it does not take advantage of rambus very well &amp; I blew it by buying this sucker back in january, but, it is a very good OCing mobo and it is rock stable and performs well.
4) Stop insulting me.
5) How is it possible to equip a platform with Rambus of two different latency's? (seriously?)

 

[superbaby]

I didn't start the insults but I will gladly stop if you will.

To equip a system with different latencies - buy two different sticks of the same type (i.e. PC700) but from different manufacturers and put it in the same system. Because the manufacturering process will be different these chips will have timings off by several nanoseconds. In dealing with this, the memory hub when it intializes the RDRAMs on bootup, will have to compensate for the highest latency RDRAM BGA chip. When doing this, it could cause other the other RIMM of lower latency to be out of sync and thus causing system instability. However this does not happen in every case of mixed RDRAM but issue persists.

 

[PCResources]

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



I suggest you go over to aces forums and discuss these matters about the A21364 and RB with Paul DeMone and Aaron Spinks, these guys are involved in the design of the 21364.

I used to have the same ignorant attitude towards RB as you do until i discussed the matter with them (and what a heated discussion that was).

Patrick Palm

Am speaking for PC Resources

 

[fkloster]

You did start the insulting. (it does not matter who started it, lets stop)

Your saying that the timings of PC-800 from one company is different from the timings of PC-800 from another? Or is that what the Doctor says? If true, would that situation not be indicative of all memory yields including SDRAMS? Why or why not?

 

[Fozzie]

Alright this will probably be my last reply, you obviously believe what you want to believe and have an arrogance beyond your technical knowledge. This discussion has been in many forms in many places and does not need to be dragged on here.

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

And you do not know what I do or do not know so I advise you to not make such RASH assumptions. Firstly you didn't make enough sense to understand whether you were talking about 128bit 200MHz DDR for main memory or at all.
And second you need to take a look at the upcoming XBox which has 64MB of 128bit 200MHz DDR in a UMA architecture. The NV2A chip from nVidia has a integrated North Bridge and memory controller. The PIII is limited to 64bits @ 133MHz, however that is a limitation of the CPU design not a problem with the memory. Most of the bandwidth would be used by the GPU in anyway.

Yes 64MB of DDR soldered onto the mainboard is different from 2 to 4 DIMM slots. However it is not a unsermountable engineering issue, if you had evidence to the contrary I would gladly listen, I'm not holding my breath though.

&quot;As for being rude to Superbaby with that comment, that was my answer to his post about me contradicting myself&quot;

Excuse me but I was the person who said you contradicted yourself. :disgust: Contradict may have been the wrong word, but you were certainly giving two stories for trying to explain away Rambus performance problems. You can go and re-read your post if you wanted to, in fact I'd suggest it.

&quot;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.&quot;

And I am saying that Rambus in like configurations is inferior to SDRAM for PC usage in most cases and inferior in virtually all situations with DDR. EG PC800 vs PC133 or PC2100. I'll explain why.


&quot;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.&quot;

:disgust: You don't understand what I was saying apparently, I guess you don't know how main memory works(). Again what I was saying is that in the majority of cases SDRAM does not provide consistant memory bandwidth above 500MB in good implimentations, less in bad ones(VIA). THIS IS WITH 133MHz FSB. If Rambus was being bottlenecked by the FSB then it should still be faster then SDRAM since SDRAM has shown itself to be incapable of saturating the FSB.

STREAMS memory benchmark results

My point about the i840 is that if Rambus was bottlenecked by the FSB in singlechannel form then how can the i840 with dual channel out perform the i820? This is simple logic to understand.

&quot;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.&quot;

Funny that if it is so expensive that even the upcoming XBox, which is estimated to cost less then $500 for MS to build, will have it. I don't really think either of us are knowledgable enough about motherboard design to claim one way or another how many layers it would require.

And to integrate the 128bit controller on the CPU chip? Well, NO, that will not be done.&quot;

Funny how you seem to know what AMD will or won't be doing. Its amazing what new die processes and packaging technology brings.

*edit*

SGI Zx10 VE Visual Workstation Technical Specifications



<< Processor

Intel® Pentium® III processor; 1 GHz, 933 MHz, or 866 MHz, single or dual; 32KB Level 1 cache, 256KB Advanced Transfer Cache
Memory

256MB-6GB; 133 MHz ECC SDRAM DIMM, three banks, two DIMMs per bank; 128 bits wide; industry-standard 168-pin, synchronous >>



I'm sure this is expensive, but it does show 128bit DRAM interfaces for main memory is possible in PC configs.

Rgrds,

 

[Mr.IncrediblyBored]

Neither. I like 'em both. (I like 3dfx more than Intel though)

 

[pm]

Patrick, Paul Demone works for Mosel-Vitelic (sp?) - which is a Canadian memory manufacturer - or at least he did last time that I corresponded with him - how could he be involved in the development of the 21364?

 

[PCResources]

<< Patrick, Paul Demone works for Mosel-Vitelic (sp?) - which is a Canadian memory manufacturer - or at least he did last time that I corresponded with him - how could he be involved in the development of the 21364? >>



You are correct, i had the discussion with Aaron Spinks and Paul Demone, but only Aaron is involved in the work with the 21364.

Paul Demone knows a lot about the RB technology and also about SDRAM, and he backed up the RB technology in the 21364 design.

He also explained why a 128bit DDR SDRAM solution would not be viable, and why an integrated memory controller would be hard to make with SDRAM.

PM, i know you know a lot about these things too, what would you say, is a 128bit, on-chip DDR SDRAM interface a viable solution?

Patrick Palm

Am speaking for PC Resources

 

[PCResources]

<< I'm sure this is expensive, but it does show 128bit DRAM interfaces for main memory is possible in PC configs. >>



Ok, so this is a very expensive system with the DRAM bandwidth of DDR-266. This is SDRAM, not DDR SDRAM, and certainly NOT 400mhz DDR SDRAM.

Patrick Palm

Am speaking for PC Resources

 

[Fozzie]

Alright one last reply. Explain to me the XBox then please. And lastly an onboard 8 channel Rambus controller and an onboard 128bit DDR controller would both need the same amount of data lines.

And again though, it does show 128bit DRAM interfaces are possible on PC style motherboards. Switching that over to DDR is not trivial, but not impossible. 128bit PC2100 would be pretty fast yes?

 

[PCResources]

<< Alright one last reply. Explain to me the XBox then please. And lastly an onboard 8 channel Rambus controller and an onboard 128bit DDR controller would both need the same amount of data lines. >>



The X-box is not a regular PC.

There are to many differences between RB and DDR SDRAM to make that kind of comparison.

Patrick Palm

Am speaking for PC Resources

 

[pm]

<< PM, i know you know a lot about these things too, what would you say, is a 128bit, on-chip DDR SDRAM interface a viable solution? >>



To be honest, I really don't know that much about DDR and Direct RDRAM. I, of course, know what they are and have a basic understanding of how they work. I've read a lot of expert opinions on the subject of the two and I've heard quite a bit about the difficulty of implementing RDRAM on a PCB motherboard. But I really don't know that much about the detailed performance differences between the two - aside from reading reviews on the web.

[In an uneditted post, I decided that I couldn't answer this, but then changed my mind]

Now that I've attempted to cover my rear by claiming ignorance, to actually answer the question: yes, I think it is a viable solution from an engineering standpoint, but probably not from a commercial standpoint. 64-bit SDRAM memory controllers have been done on a CPU. DDR is an extension of that involving signalling changes and tighter timing. A 128-bit SDRAM controller could be done if you combined two separate SDRAM controllers driving separate busses. But you'd practically double your CPU's pin-out (well, not quite, but close enough) by doubling the address bits and the data bits. This would increase cost substantially and dramatically increase the cost of the motherboard due to routing changes. I really don't think that this would be a commercially viable outside of a server implementation.

 

[ktchong]

Ah, the ironic love-hate sentiment the American public has toward the rich, famous, and successful is always amusing... On one hand, we admire people and companies that have achieve successes (i.e. Bill Gates, Microsoft, Intel, AOL, Hollywood celebs, etc.) They are, after all, the ultimate embodiments of the capitalistic American dream... On the other hand, the regular john's and joe's (i.e. we) are just plain jealous of their wealths and fames and secretly can't stand the fact that they are not willingly sharing their successes with the rest of us. How dare them!?! The socialist side of us (ok, or maybe just the democrats) want to punish the wealthy and successful for being capitalistically better than the rest of us. And we desperately want to believe that we must be better than them in some ways, morally if noting else. That's why we have to to bash 'em.

And since Intel is more successful that 3dfx. I think Intel would certainly make a better public target.

-----

&quot;Money can't buy happiness!&quot;

&quot;I'm glad that I'm not a Hollywood celebrity. Their lives are such a disgusted mess and they have no morals to speak of!&quot;

&quot;At least I have MORALS!!!&quot;

-----

Self-contradictory comment of the day: &quot;Ugh! I CAN'T stand close-minded people!&quot;

 

[HaVoC]

Wow...quite a discussion this poll thread has turned into. All three major participants would look a lot more intelligent if they weren't trading insults in every reply to the thread.

Rambus certainly has its technical merits and demerits. However, the standard I judge it buy as a fan of open technology is whether this Rambus is good for the industry. With all of the flimsy patent-based lawsuits and license issues that seem to revolve around Rambus, I must say they are an ENEMY to the memory industry as a whole. It's too bad a promising technology (serial memory) was implemented poorly and managed by such a bad company.