Why do the majority of people here like DDR better than RDRAM??

[yourharddrive]

It's always been in my experience that RDRAM is faster/better...

UPDATE: Just build new system and PC800 RDRAM is beauifully working at PC1033

 

[Mingon]

Its more that people dislike rambus rather than rdram.

 

[sparks]

Originally there were 2 reasons, first was the cost, but now there is almost parity. The second reason is the company behind the technology. Rambus has always been anti-competative and deceptive in their business practices, as evident by the recent FTC inquirys. The dislike of the technology comes from the general dislike of the company.

 

[busmaster11]

It's only marginally faster, due to higher speeds, but its got longer latencies and shorter bandwidth. Plus historically their first chipset support, the 840 camino, never worked well at all.

DDR worked well from the getgo. Plus its cheaper, and if you get standard PC2100 and up, it'll work on just about any platform you chose.

 

[fkloster]

Originally posted by: busmaster11
It's only marginally faster, due to higher speeds, but its got longer latencies and shorter bandwidth. Plus historically their first chipset support, the 840 camino, never worked well at all.

DDR worked well from the getgo. Plus its cheaper, and if you get standard PC2100 and up, it'll work on just about any platform you chose.

Longer latency? please explain how current i850e chipsets are hindered from "Longer latency" as opposed to ...shorter latency of ddr?

Shorter bandwidth? Ummm, shorter? ...you mean much MORE bandwidth for rambus right? More bandwidth = better?

840 first rambus chipset? I thought i820 was first rambus chipset....

 

[AndyHui]

i820 was the first RDRAM chipset, codenamed Camino.

i840 was the first dual channel RDRAM chipset, released after the i820, codenamed Carmel.

 

[Mavrick007]

it's all about the MAN.. fight the power

 

[SexyK]

I've never really had a problem with RDRAM. I mean, I would never have bought it for a PIII, when it was pointless and $800 a stick, but I never thought twice about going RDRAM when I first built my P4 box. <sarcasm> I can't understand all the anger either, maybe it's because the forum is overrun with AMD zealots who are jealous of P4's massive bandwidth. </sarcasm>

Kramer

 

[busmaster11]

Originally posted by: fkloster

Originally posted by: busmaster11
It's only marginally faster, due to higher speeds, but its got longer latencies and shorter bandwidth. Plus historically their first chipset support, the 840 camino, never worked well at all.

DDR worked well from the getgo. Plus its cheaper, and if you get standard PC2100 and up, it'll work on just about any platform you chose.

Longer latency? please explain how current i850e chipsets are hindered from "Longer latency" as opposed to ...shorter latency of ddr?

Shorter bandwidth? Ummm, shorter? ...you mean much MORE bandwidth for rambus right? More bandwidth = better?

840 first rambus chipset? I thought i820 was first rambus chipset....

Right. I meant 820.

Dual channel PC800 RDRAM handles data at 4 bytes per cycle, as opposed to 8 for DDR. But its running at higher speeds, 400MHZ as oppsed to 133 for DDR.

As for latency, here's a quote from overclockers.com:

However (always a but) RDRAM also has a few significant latency disadvantages due to its serial nature. Since it only transmits 16 bytes a cycle, it takes two cycles for each 32-bit WORD, as opposed to SDRAM's one. This can be remedied by using two channels, but of course, then you need two RDRAM RIMMs.

More importantly, since it is a serial device rather than a parallel device like SDRAM, if there are multiple modules being used, all requests and responses must wend their way through all the modules in between it and the CPU before getting to its destination. This slows things down more than the direct, parallel connection SDRAM has.

To keep things from getting chaotic, RDRAM deliberately slows down the responses of all the modules to the length of time it takes to communicate with the chip furthest away.

Practically speaking, the more modules you use, the more latency you have. If you have something like a server, this is the last thing you want. This is a big reason why even Intel is using DDR for its Willamette-class server processor Foster.

 

[Thor86]

I think higher overclocks in terms of cpu speed is the reason why DDR is so much more popular over RDram. It's also a personal preference thing more than anything. Both platforms offer high performance for relative prices. I just think quad-pumped FSBs of RDram platforms sound better!

 

[busmaster11]

I've noticed that AMD has generally preferred the more conservative route. While Intel goes for brute force architectures that enable higher clock speed, ie P4 / RDRAM, AMD has gone for more intelligent, yet evolutionary designs.

All things being equal, increasing performance through higher clock speeds will generate instability more often than increasing performnce by making a more intelligent chip, which is what AMD has done. While DDR and QDR, if and when it arrives, will always be able to trace its roots back to SDRAM, RDRAM is more revolutionary. But its key negative, that it does so little per clock cycle, bugs me. That plus the fact that its been years since its introduction, and its prices are still double that of DDR.

Intel has seen it, and its no coincidence that they immediately surfaced with i845DDR immediately after their contract with rambust expired.

I hope the bastards at rambust are forced to pay back every penny they conned out of the Jedec organizations.

 

[Soccerman]

Longer latency? please explain how current i850e chipsets are hindered from "Longer latency" as opposed to ...shorter latency of ddr?

Shorter bandwidth? Ummm, shorter? ...you mean much MORE bandwidth for rambus right? More bandwidth = better?

are you kidding me? RDRAM has been known to have worse latency than DDR SDRAM ever since the two started competing.

i850 chipsets are hindered by latency simply because it's there. with a higher latency comes a lower overall bandwidth due to the RAM not transferring data right away.

which chipset was the i840 again.. wasn't it another P3 chipset? what was the difference between it and the i820?

 

[SteelCityFan]

RDRAM technology has proven over and over again to be the fastest solution for the P4. First with PC800 vs DDR200-DDR333 and now with PC1066 vs DDR333-DDR400. I think it is time people start accepting it. PC1066 is without a doubt the fastest and best memory available.. and with each speed boost for RDRAM, the latency goes down.

The price is the same and often cheaper than the latest (DDR333/DDR400) DDR memory. I put 512MB's of Samsung PC800 in my computer for $140. At the time of purchase, DDR333 with a CL2.5 clock was priced around $80-$90 for 256MB. Furthermore, most name brand PC800 produced in the last few months can easily reach PC1066 speeds.

Now if you are using an Athlon system, you don't value bandwidth since they can't use it anyway. What good is DDR333/400 if the FSB of the processor is only running at 266Mhz.

Soccerman

820 was single channel RDRAM, the 840 was dual channel AFAIK.

 

[fkloster]

...i850 chipsets are hindered by latency simply because it's there. with a higher latency comes a lower overall bandwidth due to the RAM not transferring data right away.

Huh? HIGH LATENCY DOES NOT EQUAL LOW BANDWIDTH!!!!! Latency affects memory access....Bandwidth is how much juice can go through the pipe @ one time. Soccerman what is up with your statement? Your smarter than that I know it...

Simply put...:

System memory latency is NOT hindered OR slower in i850e + PC-1066 when compared too i845 + DDR 333.

 

[busmaster11]

Originally posted by: SteelCityFan
RDRAM technology has proven over and over again to be the fastest solution for the P4. First with PC800 vs DDR200-DDR333 and now with PC1066 vs DDR333-DDR400. I think it is time people start accepting it. PC1066 is without a doubt the fastest and best memory available.. and with each speed boost for RDRAM, the latency goes down.

The latency doesn't go down because part of the latency is inherent to the architecture. All the data goes through all modules before it reaches the cpu.

The price is the same and often cheaper than the latest (DDR333/DDR400) DDR memory. I put 512MB's of Samsung PC800 in my computer for $140. At the time of purchase, DDR333 with a CL2.5 clock was priced around $80-$90 for 256MB. Furthermore, most name brand PC800 produced in the last few months can easily reach PC1066 speeds.

Wrong. Only the misinformed (or those AMD zealots wishing to spend every last penny) would run 333mhz memory because the speed boost is minimal and there is a nice increase in price between DDR266 and DDR333. Most people are still getting PC2100 and PC800 RDRAM so you're comparing apples to oranges. With this comparison, RDRAM is atleast twice as expensive. Couple that with a more expensive chipset and minimal performance advantages, and you end up with not a particularly enticing option for people looking for bang for the buck.

In fact, since AMD has announced that there will not be official support for Athlon CPU FSBs over 266DDR, and Intel's only supporting PC2100 on the 845DDR as well, I'd say the price descrepancy is only increasing with PC1066 on the way.

 

[Soccerman]

RDRAM technology has proven over and over again to be the fastest solution for the P4. First with PC800 vs DDR200-DDR333 and now with PC1066 vs DDR333-DDR400. I think it is time people start accepting it. PC1066 is without a doubt the fastest and best memory available.. and with each speed boost for RDRAM, the latency goes down.

how about we compare DDR to RDRAM when they have equal theoretical bandwidths? that would show how much better DDR CAN be. unfortunately DDR will most likely top out at 200mhz, which means it would only be able to beat dual channel PC800.

I wonder what DDR II is going to be like.. will it be 64 bits wide? will it transfer twice per clock? how will they increase the frequency?

Now if you are using an Athlon system, you don't value bandwidth since they can't use it anyway. What good is DDR333/400 if the FSB of the processor is only running at 266Mhz.

well DUH. that's why people don't buy KT333 and KT400 motherboards and RAM simply to run their CPU's with a 266mhz FSB. what we do is overclock the FSB in synch with the RAM to hit 333mhz and if you're very lucky 400mhz (which happens to be where the original P4 started from), and the Athlon obviously takes use of this bandwidth because the performance gains are noticeable.

are you kidding me with those prices? right now IN CANADA I can get a stick of 512 megs of DDR333 CL2.5 DDR SDRAM for ~$160. if that was in the US, it would be around $100. but if you want to get DDR333 @ CL2, you'll have to have something slightly better than 6ns DDR SDRAM (say, 5.5). as for actual DDR400, it doesn't really exist, because no-one has a stick with 5ns or lower chips on it.

 

[busmaster11]

Originally posted by: fkloster

...i850 chipsets are hindered by latency simply because it's there. with a higher latency comes a lower overall bandwidth due to the RAM not transferring data right away.

Huh? HIGH LATENCY DOES NOT EQUAL LOW BANDWIDTH!!!!! Latency affects memory access....Bandwidth is how much juice can go through the pipe @ one time. Soccerman what is up with your statement? Your smarter than that I know it...

Uh, did you decide to completely avoid the truth here? Nevermind latency. This is what you're asking for...

Dual channel PC800 RDRAM handles data at 4 bytes per cycle, as opposed to 8 for DDR. But its running at higher speeds, 400MHZ as oppsed to 133 for DDR.

 

[Soccerman]

Huh? HIGH LATENCY DOES NOT EQUAL LOW BANDWIDTH!!!!! Latency affects memory access....Bandwidth is how much juice can go through the pipe @ one time. Soccerman what is up with your statement? Your smarter than that I know it... .

actually I haven't been all that well informed in this matter, but I do know that the more latency that a bus has, the less actual throughput of data goes through it. it might burst at 3.2 gigs/second or 4.2 gigs/second, but due to latency it isn't accomplishing that all the time is it?

I don't know for sure how DDR SDRAM is compared to RDRAM, though I do know that DDR SDRAM latency ALSO goes down with mhz increase like RDRAM. I just don't know how they are relative to each other.

 

[fkloster]

Uh, did you decide to completely avoid the truth here?

Here is the truth:

1) Current P4 implementations (i850) need more bandwidth out of system memory than current ddr implementations can give...
2) Serial rimms deliver that bandwidth, & coupled with hardware prefetch and i850, are NOT held back by poor latency perf...
3) DDR memory is a PERFECT solution for the 'elderly' Athlon platforms that are currently available as they are not as bandwidth hungry as the P4 is...
4) P4 + i845 works 'good' with DDR memory, but, rambus offers 'better' overall performance for the P4...that is fact.

enough of this...

 

[busmaster11]

Originally posted by: fkloster

Uh, did you decide to completely avoid the truth here?

Here is the truth:

1) Current P4 implementations (i850) need more bandwidth out of system memory than current ddr implementations can give...
2) Serial rimms deliver that bandwidth, & coupled with hardware prefetch and i850, are NOT held back by poor latency perf...
3) DDR memory is a PERFECT solution for the 'elderly' Athlon platforms that are currently available as they are not as bandwidth hungry as the P4 is

enough of this...

1. Is correct.
2. needs some serious sources to back it up.
3. is spewage...

Since you've lost the bandwidth argument from the architectural standpoint, you've decided to avoid that part and continue on to looking at AMD vs Intel, which was not part of the original argument.

If you want to look at the whole picture and say that RDRAM on the 850 provides more bandwidth than DDR - fine. No argument here. The 845DDR platform gives it a run for its money for much less, and because of this Intel will soon relegate RDRAM to nothing more than a niche market, and possibly not even in server roles.

Finally, how do you explain the "elderly" Athlon XP DDR platform outperforming any P4 implementation clock for clock?

 

[Insane3D]

"as for actual DDR400, it doesn't really exist, because no-one has a stick with 5ns or lower chips on it."

Actually, Kingmax has 256mb sticks of PC3200 out with 5ns modules. They are pretty cheap too...I just picked up a couple @ $61 each. The Samsung PC2700 with 6ns modules actually runs up to DDR400 just fine most of the time.

 

[SteelCityFan]

Wrong. Only the misinformed (or those AMD zealots wishing to spend every last penny) would run 333mhz memory because the speed boost is minimal and there is a nice increase in price between DDR266 and DDR333. Most people are still getting PC2100 and PC800 RDRAM so you're comparing apples to oranges. With this comparison, RDRAM is atleast twice as expensive. Couple that with a more expensive chipset and minimal performance advantages, and you end up with not a particularly enticing option for people looking for bang for the buck.

In fact, since AMD has announced that there will not be official support for Athlon CPU FSBs over 266DDR, and Intel's only supporting PC2100 on the 845DDR as well, I'd say the price descrepancy is only increasing with PC1066 on the way.

Ummm. When did I mention AMD in the sentence you quoted? Most Intel buyers are not getting PC2100.. at least not the enthusiast/performance group. PC2100 is a handycap on the memory bandwidth hungry P4.

When I was shopping for a P4 system, I was not going to cripple my computer using slow memory. I wanted the fastest. You cannot compare an outdated technology/speed with a current one when comparing prices, and then slam one for being more expensive. The older one (PC2100 in this case) will of course be cheaper. I wanted fast DDR or fast RDRAM. The fastest RDRAM was cheaper than the fastest DDR and outperformed it.

I also was not aware that Intel and Intel alone was making chipsets for the P4. The "Intel only supports PC2100 on the 845DDR" holds absolutely no water because there are obviously plenty of DDR333 and higher boards available for the P4.

 

[Soccerman]

the Athlon IS as bandwidth hungry as the P4. that explains why you see increases in performance as the fsb and RAM increase in speed.

now, I'm not saying that PC2700/DDR333 are capable of catching PC800 RDRAM, becuase even IF it has latency problems, it still manages to get a higher overall throughput to the CPU.

yes it is true that RDRAM offers more overall throughput. I don't deny that. there are other things holding me back from getting a P4 + PC1066 RDRAM, such as price, and the fact that I don't like RAMBUS the company (which is the same reason I won't buy from Creative Labs if I can). oh and I tend to go for the underdog (in this case AMD). I'm hoping to be able to take a KT400 chipset to 200mhz fsb with some DDR SDRAM @ 200mhz with a Barton I'm just hoping that AMD impliments a heat-spreader on the Barton..

 

[apoppin]

Let me step in here with a question . . .

Evidently I am going to get a BUDGET P4 system - a 1.6A and OverClock it as high as it will go.

Isn't it true that a DDR solution is the best (budget) choice for me? By increasing the FSB to (hopefully) 160 on an Epox 4G4A MB, I get some serious memory bandwith (at a really cheap price).

Correct me if I am wrong, but IF I chose NOT to O/C - but buy a P4@2.53 and run RDRAM - then bandwith would be better than non-o/c'd DDR RAM? Much like fkloster - a more elegant solution, but also more expensive?

 

[fkloster]

Since you've lost the bandwidth argument from the architectural standpoint,....

Ok, lets break this down so people can understand this bandwidth thing.....I like the restaurant analogy:

Parallel memory interface....(ddr)

A McDonalds franchise. You drive your car up, park it, walk inside and order your burger. 5 or 6 people waiting to take your order but you have to park you car and walk inside.

Serial memory interface....(rdram)

A McDonalds franchise. You pull up your car to the drive through. 1 person taking your order but you don't have to park and walk inside.

Now....

As far as loosing an arguement? Current rdram implementations have a 16 bit data path and is quad pumped with 533mhz frequency
ddr has a 64 bit data path and is double pumped with a 200mhz frequency (DDR 400)

rdram has higher bandwidth w/no noticable latency penalties (benchmarks proof this) Now what arguement did I lose?