rambus will kill ddr

[Alex]

see this article

ive always had my doubt but after reading this article its pretty clear whats going on...

its just a shame that intel and rambus have teamed up (= the crappy processor with the good memory)

perhaps the thoroughbreds or bartons or hammers or whatever will be made to use more memory bandwidth, justifying the use of rdram for end-users... because athlonXP does perform way better than northwood and its a shame that it has the worst memory...

 

[Pariah]

With Intel dropping them and AMD not using them, it's going to be a tough task without a platform to run on no matter how superior the technology is.

 

[hoihtah]

why do you say that athlon xp performs way better than northwood?

some proof?

before it was all about money.

for the same amount of money... amd did out perform pentium.
but the story changed once northwood came out

i can get a mobo and p4 1.6a for the price of xp 2000+
and with most 1.6a's o/cing easily to 2.4ghz... i don't think you can say that 2000+ is faster than 2.4ghz p4.

ok... so the argument is a bit lobsided... with only p4's being o/c'ed.

so in reality... how much do you think you can push that xp2000+ to?
with air cooling...

let's say hypothetically... you can get it to run at xp2300+'s speed.
according to toms review... it still falls below p4 2400.

and the last time i checked...
amd xp2000+ was around $250-$260 range.
and very unlikely to do xp2300+'s speed with air cooling.

once thoroughbreds come out... i'm sure the story will change.
but as of now... i'd say p4's are far from being crappy processors...
neither does it underperform
neither does it cost more.

i'm not one to bash one cpu over another.
just so you know... i do have athlon and duron running at home along side with p4 1.8a doing 2700.

 

[Whitedog]

Rambus WILL NOT kill DDR... Bah! get over it.

 

[formulav8]

<< why do you say that athlon xp performs way better than northwood?

some proof?

before it was all about money.

for the same amount of money... amd did out perform pentium.
but the story changed once northwood came out

i can get a mobo and p4 1.6a for the price of xp 2000+
and with most 1.6a's o/cing easily to 2.4ghz... i don't think you can say that 2000+ is faster than 2.4ghz p4.

ok... so the argument is a bit lobsided... with only p4's being o/c'ed.

so in reality... how much do you think you can push that xp2000+ to?
with air cooling...

let's say hypothetically... you can get it to run at xp2300+'s speed.
according to toms review... it still falls below p4 2400.

and the last time i checked...
amd xp2000+ was around $250-$260 range.
and very unlikely to do xp2300+'s speed with air cooling.

once thoroughbreds come out... i'm sure the story will change.
but as of now... i'd say p4's are far from being crappy processors...
neither does it underperform
neither does it cost more.

i'm not one to bash one cpu over another.
just so you know... i do have athlon and duron running at home along side with p4 1.8a doing 2700. >>




I heard something about Intel dropping Rambus. So, if this is true there is no way Rambus will kill DDR.

 

[Oyeve]

DDR2. 'nuf said.

 

[dullard]

<< I heard something about Intel dropping Rambus. So, if this is true there is no way Rambus will kill DDR. >>



Intel dropped Rambus on their Xeon MP system released Monday. This is since it is meant for machines with gigabytes of memory and was designed at the time RDRAM was sky high in price. So this step was understandable. I haven't heard anything else about Intel and Rambus - so do you have a link?

 

[SteelCityFan]

As long as DDR2 comes out in a hurry. Current DDR333 simply can't keep pace as Processor speed increases... and we all know processor speed increases a lot faster than DDR/SDRAM speeds do. With RDRAM going to 1066 Very soon, the gap between performance will widen.

 

[hoihtah]

i realize that in my previous post... i forgot to mention about the whole rambus vs ddr debate.

as with lot of you guys...
i'm with ddr
never like rambus.

 

[ToBeMe]

<< With Intel dropping them and AMD not using them, it's going to be a tough task without a platform to run on no matter how superior the technology is. >>


Someone please come up with some proof as to this......................with the next gen. RD boards for the 533FSB chips already made, and future chipsets & boards already in the roadmap for companies like Asus, Abit, and even Intel itself, why do people suddenly think Intel wil be "dumping" RDRAM & Rambus anytime in the near future???? If you recall, Intel recently signed a new 3 year contract with Rambus..................

 

[Pariah]

"Someone please come up with some proof as to this"

http://content.techweb.com/wire/story/TWB20001030S0016

 

[dullard]

<< http://content.techweb.com/wire/story/TWB20001030S0016
Intel dropping Direct Rambus DRAM from every computing platform but high-end workstations by mid-2001...
while the yet to be introduced Intel 850 chipset will be dropped in the middle of the third quarter [2001]...
Intel's sole remaining Rambus chipset will be an enhanced 850 device code-named Tehama-E... >>



It is almost mid-2002 and Intel is still releasing its top of the line P4 and its non-server Xeons with RDRAM. Wasn't Tehama cancelled a long time ago? Try a more up-to-date link please.

 

[ToBeMe]

<< "Someone please come up with some proof as to this"

http://content.techweb.com/wire/story/TWB20001030S0016 >>


Just one question for ya'...................Did ya' happen to notice the DATE on that page............:Q LOL!


<< Intel Roadmap Shows Little Rambus Support October 30, 2000 (2:40 p.m. EST) >>

 

[BD231]

Currently DDR is the best bang for the buck and its obvious RAMBUS will always be faster, hell, thats what it was specificlly made for right?. I personaly dont see DDR being as fast as rambus anytime soon, but I also dont see anything else than DDR being widely available. So either way we're stuck with DDR.

 

[Def]

Is anybody forgetting that EVERYONE here is talking about DUAL CHANNEL RDRAM and comparing it to SINGLE CHANNEL DDR?!?!?!

Once dual channel DDR is implemented, RDRAM will be begging for mercy. Dual channel DDR33 = 5.4Gbps of memory bandwidth, which is MORE than dual channel PC1200 RDRAM puts out(4.8Gbps). Keep in mind this speed DDR memory is already out, while PC1066 RDRAM hasn't even been released.

RDRAM would have to move to two 32bit channels to keep up, and the cost of those modules would be absolutely insane.

DDR is the way of the future. RDRAM finally "caught up" with DDR and is about to be left in the dust again.

Defster

 

[majewski9]

I heard that SIS is trying to come up with RDRAM chipsets!!!!!!!!!!

I also heard that Hammer might use RDRAM in the future, but for now will relay on DDR333! It is a possibility that SIS will release a Hammer + Rambus chipset however I am inclined to believe that they'll probaly a P4 RDRAM chipset only. I dont ever think that VIA or AMD will ever develop Rambus chipsets.

DDR is going to be the main memory until DDR II comes out!

 

[dullard]

<< Keep in mind this speed DDR memory is already out, while PC1066 RDRAM hasn't even been released. >>


333 MHz DDR hasn't been officially designed yet. Sure some places sell good performing 266MHz and call it 333 MHz - but this stuff might not meet the final specifications and you are screwed if you bought it and it doesn't meet them... PC1066 RDRAM specifications have been finalize a long time ago. Samsung and other big manufacturers claimed that they will start mass producing it this month. This makes your statement pointless.


<< RDRAM would have to move to two 32bit channels to keep up, and the cost of those modules would be absolutely insane. >>


According to Rambus the cost is about the same. Moving to 32bits is much easier than trying to keep bumping up the speed of DDR.

 

[ToBeMe]

<< Is anybody forgetting that EVERYONE here is talking about DUAL CHANNEL RDRAM and comparing it to SINGLE CHANNEL DDR?!?!?!

Once dual channel DDR is implemented, RDRAM will be begging for mercy. Dual channel DDR33 = 5.4Gbps of memory bandwidth, which is MORE than dual channel PC1200 RDRAM puts out(4.8Gbps). Keep in mind this speed DDR memory is already out, while PC1066 RDRAM hasn't even been released.

RDRAM would have to move to two 32bit channels to keep up, and the cost of those modules would be absolutely insane.

DDR is the way of the future. RDRAM finally "caught up" with DDR and is about to be left in the dust again.

Defster >>


Um, sorry, but PC1066 & PC1200 are going to be 32 bit...........From Rambus itself....................




<< PC-1066 and PC-1200 is what we can expect to see within the next year, whereas they'll also be moving from a 16-bit module to a 32-bits module. Which means that in case of the i850 chipset you'll only have to stick in one module instead of two. >>



As for DDR333................that's some great stuff..........where can you actually find it????? I run an Sis 645 board with the Corsair XMS PC2700 as well as two i850 boards here................thing is, while the XMS PC2700 cuts the lead..............it still can't match the PC800 RDRAM we have right now and PC1066 RD will be available along side the 533Mhz FSB Northwoods on May 20th.............

 

[fkloster]

there is a lot of nonsense floatin' around this thread...I'll stay clear.....



<< ...its just a shame that intel and rambus have teamed up (= the crappy processor with the good memory) >>




lol foolish statement... A P4 2.2 northwood .13u is a technological marvel...

 

[rockhard]

hmmm?
P4 has longer branch prediction pipe resulting in lower IPC, yeah?
This might be fine and dandy with a game or a few apps running simultaneously which spend some of their life's eating zero clocks, but what happens when you run a dozen apps, all competing for processor time?
Cache is not going to be able to mask this kind of load and the TBird/XP should walk all over the P4 IMO.
Ive recently come from an AMD XP 1600+ to a P4 1.6A overclock and i can tell you that when heavily loaded with clock crunching apps the P4 is not as good as the lower IPC XP or TBird.
So much so im going back to the XP as im not happy with my lower system performance/less responsive nature.

Benches are good and fine but DO NOT reflect real life load/useage of a CPU IMO.

 

[JellyBaby]

<< I heard that SIS is trying to come up with RDRAM chipsets >>

SiS has a new license but who knows what they'll do with it.

Both memory technologies have a future. To me, so long as DDR is a viable solution I will use it and leave Rambutts, Inc. out on the porch like the naughty red-headed step child it is.

 

[ToBeMe]

<< hmmm?
P4 has longer branch prediction pipe resulting in lower IPC, yeah?
This might be fine and dandy with a game or a few apps running simultaneously which spend some of their life's eating zero clocks, but what happens when you run a dozen apps, all competing for processor time?
Cache is not going to be able to mask this kind of load and the TBird/XP should walk all over the P4 IMO.
Ive recently come from an AMD XP 1600+ to a P4 1.6A overclock and i can tell you that when heavily loaded with clock crunching apps the P4 is not as good as the lower IPC XP or TBird.
So much so im going back to the XP as im not happy with my lower system performance/less responsive nature.

Benches are good and fine but DO NOT reflect real life load/useage of a CPU IMO. >>


Well..........I personally have an XP1900+ right along side my P4 2.2........both are O/C'ed and I'm wondering how you are telling the difference or if your P4 setup has a problem or bottleneck of some sort because I'm constantly multi-tasking with many different apps............the Xp's in no way "walk all over" the P4's in any way. In fact, as far as multimedia editing or proccessing, I'd definately favor the P4's I run over the XP's. As for "clock crunching"......my P4's crunch the hell out of UD and while all of the P4's are clocked higher than the XP's I have (so you have to consider this), the P4's are the fastest crunching procs. I've ever used...........

 

[Pabster]

Wrong. DDR has already made RAMBUSt defunct, and will continue to walk all over it.

Dual-Channel DDR chipsets are just around the corner. Even with today's basic PC2100 modules, PC1066 RDRAM has met its match. DDR-II will be a reality sooner than most people realize, rendering PC1200 obsolete before it ever enters mass production.

RAMBUSt simply makes no sense. Besides the lawyers masquerading as IP specialists, let's look at the technology. PC1066 modules run so friggin' hot you literally need an 80mm cooling fan placed squarely over them to keep them from melting. And that is with the factory installed heat spreaders, which all RDRAM requires due to the ridiculous thermal dissipation of the modules.

I don't know if a select group of people spent a little too much for RDRAM, or have some other vested interest in the success (or lack of) RAMBUSt, but it is kinda hilarious to see topics like this popping up. I do believe Intel is shifting away from RDRAM; perhaps not entirely, but for the most part. And with good reason.

 

[rockhard]

Heh, im thinking your idea of load and mine are different
I have 3 boxes that are running VMWare GSX Server, 1.5gb ram and 10k raid 0 SCSI in each.
When i install the distributed clients onto the Virtual Clients/Servers (can be as high as 5 or 6 Clients on each GSX Server host) the rigs with the XP's/Tbirds fair a lot better than the P4 1.6A that i have tried extending my Virtual LAN onto
I really noticed this when i had to hookup one of these headless servers to a monitor and tried to connect to the internet on the Host while it was fully loaded with clients and i was stunned at how unresponsive it was
Just to make sure i retried the same thing with an XP rig and noticed it too was sluggish, but not as bad as the P4 :/
This i could only put down to the lower IPC/longer branch prediction pipe of the P4.


This kind of load is more applicable to a heavily loaded server environment than Desktop, so may be why you feel differntly maybe?

 

[fkloster]

<< Wrong. DDR has already made RAMBUSt defunct, and will continue to walk all over it. >>



Ummm, Pabster, care to explain to the reluctantly listening population @ Anandtech how DDR memory is 'walking all over' current RDRAM implemintations?



<< Dual-Channel DDR chipsets are just around the corner. Even with today's basic PC2100 modules, PC1066 RDRAM has met its match. DDR-II will be a reality sooner than most people realize, rendering PC1200 obsolete before it ever enters mass production. >>



Dual-Channel DDR is 'just-around-the-corner' of what year? PC2100 is equal (met its match?) to PC-1066 in what way? DDR-II will be a reality and make 64bit PC-1200 obsolete? Please elaborate with hard facts and less 'running of the mouth' ...perhaps you could educate some people instead of looking vague and desperate to get a point across...



<< RAMBUSt simply makes no sense. Besides the lawyers masquerading as IP specialists, let's look at the technology. PC1066 modules run so friggin' hot you literally need an 80mm cooling fan placed squarely over them to keep them from melting. And that is with the factory installed heat spreaders, which all RDRAM requires due to the ridiculous thermal dissipation of the modules. >>



Serial memory occupies todays most advanced personal pcs and it makes no sense? Could you please provide some facts about your findings that PC-1066 melts when not actively cooled? What do you consider 'ridiculous' thermal dissapation when comparing to parrallel memory heat dissapation?