Hey Sander(et al) .. you wanted some URL's etc. for some other criticisms of Rambus and/or regarding this article... heres one for you if you havent already been there:
JC linked directly to your article and asked "technologically-oriented" people to comment on it.. I think you'll find some of those rebuttals to your article here that you can address in your next article if you so choose to do it.
Now I agree with all that was stated by Moraelin.
1. - First of all, CAS2 would have been a much better choicethan CAS3. If you are going to compare PC800 RDRAM you should use the best in SDRAM, PC133 CAS2.
2. - The Peak Bandwith is not good for a comparison. "In Theory" and "In Reality" you will have two totally different results. First of all, most people do not buy computers with PC800 RDRAM. PC700 or PC600 but not PC800. PC800 has a "Theoretical" bandwith of 1.6GB/sec, PC700 1.424GB/Sec, and PC600 1.064GB/sec; PC133 is 1.064GB/sec; same as PC600 RDRAM. This based on the simple calculation ("X" cycles/s x "Y" bits/cycle x 1 B/8 bits = Bandwidth "B/sec"). DDR is coming out shortly. PC266 with a peak bandwith of 2 x 1.064GB/sec = 2.128GB/sec or 2 times PC133's Bandwith. Peak is only what it could do, not what it actually does. So peak means nothing here. Show reality!!! Set some comparisons utilizing mutiple benchmark tests. (I.E. Threadsort, QMC, Steam, and multiple others.) Let us see the effective components of bandwith and see how SDRAM compares to RDRAM. I think you'll see a different story.
3. The statement "The high peak bandwidth of RDRAMs (1.6 GB/sec) is complemented by an efficient packet-based protocol and a high bank count that achieves high effective bandwidth." is just not true. How can you say that the protocol is efficient? RDRAM works by sending 16-bit wide data packets from one RDRAM device to another until the packets are put togeher into 64-bit chunks, then the processor can use it. The data may have to go through all of the RIMMs in a system. That is totally different from the grid used in SDRAM. Its protocol adds to RDRAM's complexity and causing several problems; especially increases in latencies. So if I don't want my memory to be a even bigger bottleneck than it already is, I can only us ONE RIMM, then it will only have one to go through. That is rediculous!! Better protocol??????? I don't think so, and in the famous words of NHL Goalie Bob Frose, "Can't Go Jack!!" Latency just ate your bandwith, Chomp! and there are more Chomps to come.
4. I know this one is only a technicality, but, on the DRDAM Protocol Diagram it states "Smaller Write-Read bubble, Increases Bandwith" That's incorrect. It only means that there will be a lesser of and deficit. It is like the Federal Government. Instead of being in the red $10 Billion, they are only in the red $5 Billion. The bubble is still a negative effect, how big the bubbles are only determine how bad the effect is. It won't improve bandwith. So there is another latency, and that to the protocol latency. Chomp! Chomp! Now there's two Bites.
5. I'll write that one tomorrow, I just have a little research on the power consumption - modes of the RDRAM. I'll be back later. Until than please respond to this. You've got two strikes against you.
First I would like to apologize if for not acting in a polite fashion. I do not mean to degrade or belittle you. You have more of an education than I do. Myself, I only have a BA Economics; but reading and research make up the rest. Well my parents both were IBM'ers, but I doubt that they could pass anything on geneticlly. But seriouslly, I did like the article in reference to the direction that RDRAM could take.
I just don't feel that article put RDRAM in the correct light. I mean you stated that you used all of the "reommended" drivers and try to be as objective as possible. But if you do that with the 440BX, the recommended is PC100; you have, http://www.hardwarecentral.com/hardw...views/1787/14/
"System Configuration Intel 440BX
CPU : Intel Pentium III ES at 600 or 800 MHz with a 100 MHz FSB
Motherboard : Soyo SY-6BA+IV Intel i440BX BIOS v2BA1
Memory : Samsung 128 MB GA PC133 SDRAM CAS3
Videocard : Elsa ERAZOR X2 GeForce DDR 32 MB BIOS v7.05.00
Harddisk : Quantum Fireball Plus KA 13.6GB ATA-66
CDROM : Samsung SCR-3231 32X IDE-ATAPI
Floppy : Generic 1.44 MB
Case : Generic Midi-ATX
Powersupply : Enlight 250W AMD-Approved
Operating System : Windows98 SE, typical installation
DirectX Drivers : DirectX 7.0
Drivers Video: nVidia Detonator v3.68
Drivers ATA : HighPoint Technology HPT366 v1.22
Drivers Chipset : Intel Chipset Inf Update v2.2"
I mean you are just opening up yourself for an attack. It is not stated if it is overclocked or not. But if you use Intel's Bios as is, you won't take advantage of PC133.
Personally I think that you have a lot of guts taking the abuse of so many people that have posted to this discussion board. Just be careful.
Sander, as a graduate of MIT with a Doctorate in Electrical Engineering and a Masters in Computer Science, I have a great deal of empathy for you and what you were trying to achieve.
These nonsensical flame wars do nothing as there's nothing to back them up. Of all of the comments posted here, 90% arose from people who do nothing more than read the work product of others and lack the ability to perform the testing on their own at their own expense as well as the understanding of benchmarking in and of itself.
Simply put, if you don't have an operating system that is optimized for the new technology, and the bench mark software cannot identify the new technology, or identify it correctly and it too is not optimized for the new technology, what ever the charts indicate is largely irrelevant other than to show a comparison based upon a theoretical standard.
In this particular instance, the only information that can be derived is how well the bench mark software ran on the various hardware with Windows 98 and Office 2000, not how well the hardware itself performed.
Now given that Windows 98 and Office 2000 are somewhat optimized around earlier x86 models utlizing X chipsets, throwing the new VIA and Intel 8XX chipsets into the fray to try and prove or disprove the advantages of DDRAM over SDRAM really develops nothing other than to say that if your using the BX chipset your software will probably perform better. But this is a function of the OS, not the hardware.
As for the remaining 10%, at least some of these demonstrated the ability to remain open minded and objective towards new emerging technologies.
After reading this article on Nov 12/00, I cannot believe the points brought out in said article. Not only is most of the information misleading, it is completely wrong. To try and say that RD-RAM has lower latency than SD-RAM, is absolutly ludicrous. It is now common knowledge that Rambus is a farce, and is only an "intellectual property" company. RD-RAM is still far more expensive than SD-RAM, and DDR-RAM has now proven itself. I am thoroughly disgusted with the quality of reporting and will make sure this site is kept away from.
i belive that the via board was crippled i use rdram but i notice a performence increase in my multiprocessor systems
it makes sence the higher bandwidth would do better with multiple request
even the dual channel single processor boards arnt that impressive
but i belive that in high end systems rdram is worth the coin but for games and stuff
get yourself some ddr 2100 thats what i use for my consoles.
im open to oppinions if anyone has them
Thansk for the continung article on RDRAM vs SDRAM.
I'm not a techno type. I am a consumer and one who happens to be in the middle of researching the best buy and performance for my money in the current technology and economy arena.
The info in the articles clearly states what is going on.
omg, i can't believe how may people make their first posts in the article forum. and how long this thread has been up!
I just saw a P4 system using DDR PC2100 RAM. Is this the compromise between SD and RD on P4 systems? Does anyone know the PROS or CONS of this? Or even seen this before? It seems that DDR SDRAM has all the benefits of SDRAM and is a good alternative to RDRAM for the P4 systems. I don't claim to much about this subject, but I saw this configuration and I thought I would ask. Any replies would be very helpful.