Why isn't there a backlash against DDR SDRAM like there was with RDRAM?...RE: VIA KT133A

[Dulanic]

And I still stand behind the facts that DDR will outperform RDRAM when a system is built for and around DDR... The fact still stands that RDRAM 800=1.6GB/sec and DDR PC2100=2.1GB/sec.... is that shown in the current implmentation? No and like I said before, my arguement has NOTHING to do with the current implemtation... it has to do with the memory and the memory ONLY.

 

[OneEng]

To clarify my position, I don't believe that serial RAM can ever be as effective in a PC as parallel RAM. I believe that latency distroys processor performance. This is true today when the latency differential between CPU cache and main memory is ONLY one order of magnitude (10). By the end of next year processors will have reached speeds of around 2Ghz making the differential more like 20!

RAMBUS RDRAM isn't all bad. They do multi-pump their bus. PC800 memory does not run a clock at 800, only 400 and double pump. Still at 400 Mhz, RDRAM has much more difficulty being produced than PC2100 which clocks at only 133Mhz. RDRAM still runs signifigantly hotter to the point of requiring the use of heat spreaders in the design.

Serial protocol RAM does perform well under situations where large amounts of information in one continous stream are needed frequently. Under these conditions the CPU simply asks for 500Kb of stuff from address X and then receives it all in one nice package.

Unfortunately for serial RAM, this isn't the most common memory request in most applications! It isn't that the high bandwidth of multi-channel serial RAM isn't good, it is just that the latency problem IMHO overshadows it.

From an engineering standpoint, it is better to simply quad pump a parallel SDRAM. Any increase in clock speed is then rewarded 4 fold in bandwidth while the latency of the part decreases.

Soccerman,
How can a serial protocol driven memory ever achieve the latency values of a parallel memory at the same clock? My arguement is based on an assumption that simply raising the clock speed is a bad idea. It makes more heat, and lowers the yields, and ultimately costs more to produce and purchase while suffering in performance.

It would be interesting to see how P4 performs on a quad pumped SDRAM. With a system already designed to minimise system RAM latency issues, I would think it would fly. The quad pumped FSB would mate pretty seamlessly to a quad pumped memory bus making the chipset pretty effecient.

 

[NOX]

<< For the 5th time ive said this IM TALKING ABOUT PURELY DDR AND RDRAM... NOT the chipsets or CPUs... this is PURELY about the memory ONLY. >>

Sure that?s fine, now does that change anything? No! DDR still performs the same, which is not much more then SDR, great, glad we got that out of the way.

<< So you can sit here and show me benchmarks... I KNOW in the current implentation of DDR with the Athlon there is only a small performance difference. But that is NOT what I was talking about. I was talking about purely the memory not their CURRENT implementation. >>

Great, I?m glad you agree that DDR offers only a ?small performance difference?. There is nothing other then current benchmarks that suggest DDRAM offers little performance over SDR. So you don?t even know for sure if in fact a future implementation will allow for a performance increase right?

<< And I still stand behind the facts that DDR will outperform RDRAM when a system is built for and around DDR... The fact still stands that RDRAM 800=1.6GB/sec and DDR PC2100=2.1GB/sec.... is that shown in the current implmentation? No and like I said before, my arguement has NOTHING to do with the current implemtation... it has to do with the memory and the memory ONLY. >>

Hahaha did your crystal ball tell you that? Again, those numbers don?t mean jack, they are theoretical at best. Like I said, if you?re only talking about the memory then fine, but as of current, DDR is not much faster then SDR, and that?s fact! Until otherwise seen I would not have the audacity to speculate the way you have.

 

[rmarango]

reply for Nox:

The mainboards that support both memory types (SDR and DDR) do exist, so keep you better informed going here [L=]http://www.socketa.com/cgi-bin/search_form.pl[/L]:
Then select for the memory search both SDR and DDR, for your information these are some of the mainboards that I retrieved from the database:


Arima
KUAIA
ALi MAGiK 1
ATX


Chaintech
A7A266
ALi MAGiK 1
ATX


PC Chips
M816
ALi MAGiK 1
micro ATX

 

[Sohcan]

nm

 

[Sephiroth_IX]

Yet another annoying topic changing blurb:


Dual Channel DDR on the NVIDIA Crush chipset. Meeeoow.

 

[ElFenix]

DDR should be really important for the amd 760mp chipset though. dual ev6 buses fed by ddr... *drool*

 

[viper007]

It is easy to say that Dual channel DDR is going to be better then Dual channel RDRAM!

But What about the price of this! Its going to be huge!!

Also what about the future, RDRAM has a might brighter future. Why, its scalibiltiy is far better then DDR..

Dual channel RDRAM is able to deliver a bandwidth around the same as DDR @ 128bit, but its only running at 32bit! Imagine how much it can scale! 64bit RDRAM would be unbeatable, while DDR would cost more, because it requires a 256bit data pathway to keep up...So if DDR was able to stay with RDRAM in bandwidth, it would actually cost quite a lot more...

Cheers..

 

[fkloster]

One thing we should all agree on:

1) Comparing a product that is not available, Dual Channel DDR, to an existing product, Dual Channel Rambus, is pointless. Might as well compare future Dual Channel DDR to future rambus solutions...

 

[Sohcan]

You've got a good point viper, but I don't think it's feasible for desktop systems. Alpha's EV8 is going to have eight RDRAM channels for a total bandwidth of over 10 GB/sec, but it is going to be really expensive. Even four channel Rambus would be cost prohibitive for desktop systems, as they would require four RIMMs. Very few mainstream desktop PCs have four memory slots (discounting the higher-end retail Asus, Abit, MSI, etc. boards that DIY builders use), and I would hate to upgrade all four RIMMs if I wanted to have more memory.

 

[Dulanic]

<< Sure that?s fine, now does that change anything? No! DDR still performs the same, which is not much more then SDR, great, glad we got that out of the way. >>



DDR technology has been used in lots of applications and it DOES double the performance.... You KEEP going back to current implementation because its the ONLY backup you have and it means jack.... DDR has doubled the performance of memory on NVidia cards... it doubles the FSB bandwith on the AMD EV6 bus... you can keep using your very uninformed arguement of current implementation... anyways its obvious there is no point in argueing with you.

 

[fkloster]

<< DDR has doubled the performance of memory on NVidia cards. >>



Yup, sure has! &amp; everyone knows that video memory architecture implementation is &quot;One-In-The-Same&quot; as mainboard memory architecture implementation.

 

[Dulanic]

LOL you guys are hilarious (not to add complete morons) you totally warp what everyone says and ignore the truth but thats ok.... When DDR is correctly implemented on a platform Ill come back and laugh right in your face.

 

[etech]

I look at it this way. Higher memory bandwidth does not seem to show much improvement in actual applications at this time. The KT133A chipset will probably the one in my next motherboard.

By the time the higher bandwidth is needed the memory manufactuers will have come up with a better solution then Rambus that they will not have to pay royalties on.

 

[Dulanic]

BTW it can be proven that DDR CANT be using its potential also. If raising the FSB but leaving the memory speed the same (KT133A) increases SDRAM performance to about the same level as DDR it is obvious there is a bottleneck. How can you fit all the memory bandwith of DDR through the FSB if it cant even push the full potential of SDRAM through that same FSB? Obviously it cant.

 

[Rigoletto]

Hmm, yes DDR has potential. It is an evolutionary progression of SDR, and perhaps other alternatives would have been better (though I don't know what SLDRAM is).
But all that we need to know is that there is no need to wait for DDR boards or memory instead of building SDR rigs now. The rest is just snore

 

[NOX]

Dulanic,

You keep comparing video cards to system ram why? They have nothing in common as far as SYSTEM instructions!

You?re right about one thing, there is no sense in arguing, DDR just doesn?t live up to the hype, and you need to accept that. Currently there is nothing to suggest otherwise. Maybe VIA has something better.

<< LOL you guys are hilarious (not to add complete morons) you totally warp what everyone says and ignore the truth but thats ok.... When DDR is correctly implemented on a platform Ill come back and laugh right in your face. >>

You also suggest that I continue to talk about current implementation, yet you bring up future implementation in your response! Don?t you think that?s hypocritical of you to suggest currently implementation means jack, but future implementation will correctly implement DDR? That may be true, but you don?t know that right? Until that time comes we?ll just have to wait and see, you suggest it like no one I?ve ever seen before.

Rigoletto,

Agreed, it?s really speculation at this point whether DDR will produce greater yields. To compare anything else but what?s currently available is guessing, and some of us seem to be experts at that.

rmarango,

You have any idea if ASUS, MSI, ABIT, Tyan, will produce boards that support both DDR/SDR? I?ve never heard of those companies you posted except for Chaintech, which that board it not even available.

 

[Dulanic]

<< You also suggest that I continue to talk about current implementation, yet you bring up future implementation in your response! Don?t you think that?s hypocritical of you to suggest currently implementation means jack, but future implementation will correctly implement DDR? That may be true, but you don?t know that right? Until that time comes we?ll just have to wait and see, you suggest it like no one I?ve ever seen before. >>



Whats even more funny is that isnt it hypocritical of you to say DDR is no better than SDRAM? I mean RDRAM was in the same position with the P3 was it no?



<< You?re right about one thing, there is no sense in arguing, DDR just doesn?t live up to the hype, and you need to accept that. Currently there is nothing to suggest otherwise. Maybe VIA has something better. >>



You can keep telling yourself that DDR sucks and that it offers no advantage over SDRAM... but when a platform is really built around DDR I will come back and laugh at you. Just remeber how hypocritical you are being yourself.... because as you know RDRAM was in the EXACT same position as DDR with the P3.

 

[OneEng]

NOX,
Agreed, it?s really speculation at this point whether DDR will produce greater yields

It is statistical fact that higher frequency parts necessarily have lower yields. They also consume more power, produce more heat, and require more stringent design rules when routing the PCB.

 

[NOX]

<< Whats even more funny is that isnt it hypocritical of you to say DDR is no better than SDRAM? I mean RDRAM was in the same position with the P3 was it no? >>

Do you know the meaning of hypocritical, I don?t think so? It is no better then SDR, there is nothing hypocritical about that. Yes, RDRAM was and is still in some areas in the same position. Why did RDRAM show more of what it could achieve, well because the P4 and i850 was designed to provide better support for RDRAM, can you not understand this? It doesn?t matter if DDR is better or if RDRAM is better, if you don?t have the proper support it?s all for nothing (i.e. chipset, CPU).

<< You can keep telling yourself that DDR sucks and that it offers no advantage over SDRAM... but when a platform is really built around DDR I will come back and laugh at you. Just remeber how hypocritical you are being yourself.... because as you know RDRAM was in the EXACT same position as DDR with the P3. >>

Get real man, please, I never said DDR sucks! All I said is what Anands review suggest, no big advantage over SDR at this present time. Please don?t be so dumb! You can come back and laugh if in fact DDR is improved with better support, I never said it wouldn?t. In another thread I spacifically ?VIA?s DDR solution should/will bring fort an extra performance boost to DDRAM. We need to give it time?.

 

[rmarango]

Reply to Nox:

There is also a mainboard from MSI model 6375 chipset ALi MAGiK 1 that supports 3 DDR dimms + 2 SDR dimms.

For more please go here :
Anand Article

In the link above Mr. Kuo of the Anand Team will show you the flexible solutions of Asus,Microstar and PC Chips.
The entire article is named ALI in 2001 : DDR SDRAM for all.

Then please go here :
Anand article 2

In the link above Mr. Lal Shimpi in person will show you the benefits of ALi Magik 1 , here is an abstract :
&quot;While DDR is definitely the name of the game when it comes to high performing Athlon solutions, we have pointed out time and time again that the Athlon platform in general and the Duron in particular isn?t being taken seriously enough in the value market. This is unfortunate because there is such a large market for AMD here, and it is even more unfortunate because it isn?t directly AMD?s fault that they are in this position, rather it is the lack of any third party chipsets that promote a value platform that is holding AMD back here.

Although the MAGiK 1 does not feature any sort of integrated video as you would expect to find in a value solution, the M1647?s DRAM controller features SDRAM support in addition to the DDR SDRAM support we just talked about. This is a very strong feature for system builders and OEMs since they can implement one system design, based around a single motherboard, and simply switch out the type of memory they are planning on using according to the needs of their customers. Using DDR for the high end and performance market segments, while resorting to conventional single data rate SDRAM for the rest of the market&quot;


Greetings

 

[Jhhnn]

Why No Backlash For DDR?

Simple, really. The market has changed.

When Intel introduced Rambus memory, they adopted the brute force method to attempt to make the market accept something that offered no real advantage at that time, at about 4X the price. Their stated intention at the time was to drop the BX entirely in favor of the 820. The VIA chipsets and AMD socketed processors were in their infancy, and not yet ready for primetime.

All different today. AMD and VIA offer real alternatives, the BX isn't going away just yet, plus there is the 815 family. Nobody is shoving DDR into anybody's orifices. It actually offers some small performance increases as well, which will likely improve with bios updates.

The P4 is optimised for Rambus, for sure. Given the nature of Intel's agreements with Rambus, it has to be. They are stuck supporting it for a while yet. They tried to make a killing without offering the market anything in return, and got burned.

 

[Dulanic]

<< Do you know the meaning of hypocritical, I don?t think so? It is no better then SDR, there is nothing hypocritical about that. >>





<< You?re right about one thing, there is no sense in arguing, DDR just doesn?t live up to the hype, and you need to accept that. Currently there is nothing to suggest otherwise. Maybe VIA has something better. >>





<< You can come back and laugh if in fact DDR is improved with better support, I never said it wouldn?t. >>



Wow didnt you just contradict yourself? One minute your saying its no better than SDR, then the next minute your saying that you never said that and that it could be better with the right support.

 

[Leo V]

Just adding oil to the fire:

The effectiveness of DDR memory could be adversely affected by AMD's chipset design. NVIDIA's Crush chipset could change that. Of course, even then applications that really use the extra bandwidth (dynamic geometry over AGP, software image rendering, etc) would really benefit.

 

[NOX]

Dulanic,

Please man; you need to stop being dumb.

FACTS:

1. DDR has no real advantage over SDR at this present TIME (for a number of reasons, reasons which you don?t seem to understand or want to understand)!
2. I never said better DDR support was not obtainable.
3. I?ve stated from the START current support is lacking, and to give it time.
4. DDR is not better then RDRAM, not the way you are suggesting. Seeing Linkpack, MBP, and a few Q3 benchmarks, it is very simple to say RDRAM is better then DDR also. But to do so would be ill informed.
5. You can?t make a better statement once a P4 DDR solution is here because the P4 wasn?t designed for DDR.
6. The numbers you posted are theoretical and nothing more.
7. The latency issue of RDRAM is not even an issue in real world operations.

There is nothing contradicting about what I have said.

ASSUMPTIONS:

You on the other hand have suggested erroneous things. You have nothing to offer in terms of facts, rather your own assumptions and half-truths.

Anyway, it?s a waste of time to argue with you, you have obviously made up your mind that DDR is paramount to everything else, and you continue to twist my words.