Sign in with
Sign up | Sign in
Your question

Yet another setback for AMD

Tags:
  • CPUs
  • RAM
  • AMD
Last response: in CPUs
Share
October 22, 2001 10:30:05 PM

rambus is going to release 3.2ghz rdram, compared to the current 266mhz ddr-ram standard, this is 12x faster. Unless amd somehow signs a deal with rambus, it looks like their cpu's will be beat even worse in the latest benchmarks, because they will have 2 things against them: sse-2 and gigaherz speed ram. Then again, you can always try and support the weaker underdog.

More about : setback amd

October 22, 2001 10:51:01 PM

That RDRAM won't be released until 2004. By then, DDR will be old news and a new memory technology will probably emerge (QDR RAM perhaps)? Why is this a set back for AMD? AMD processors aren't the only processors that use DDR. P4 will use them too. SSE2 means nothing right now, and by the time it is used in many apps, the Hammer will be out to counteract Intel. AMD is, in every case, keeping up with and even exceeding Intel. I'm not AMD biased, I'm just conveying the truth. Intel_inside, do you even know what SSE2 is and what cache is? Do you understand the architectural differences are between Intel processors and AMD processors? If you do not, then I suggest you do some research before you make some wild claims about AMD. We need AMD as competition. If we don't have competition, then CPUs would still cost a fortune like they did a couple of years ago.

AMD technology + Intel technology = Intel/AMD Pentathlon IV; the <b>ULTIMATE</b> PC processor
a b à CPUs
October 22, 2001 10:56:14 PM

What a mature response from a self-proclaimed AMD man!

Back to you Tom...
Related resources
October 22, 2001 11:02:30 PM

Hehe, I couldn't think of a better name at the time. Ohh well, I just don't like uneducated claims.

AMD technology + Intel technology = Intel/AMD Pentathlon IV; the <b>ULTIMATE</b> PC processor
October 22, 2001 11:18:49 PM

O, u didnt hear the news?
there is now ZDRAM-ZiloinDram, will only work on AMD coz for some reson it wont boot when INTEL is in..there have been some reports of a script that detacts on start up if there is an intel part in the mechine, after which it boots only if there is none.

while back on earth..can some 1 plz kill that moron, this aint getting any smarter..

<font color=green>
*******
*K.I.S.S*
*(k)eep (I)t (S)imple (S)tupid*
*******
</font color=green>
October 22, 2001 11:41:17 PM

3.2 ghz is its bandwith not its operating frequency you dolt, thus making your claim of it being 12 times faster completly baseless. And what, you don't think DDR will evolve or improve at all between now and 2004 when your Rdram is released?

Video editing?? Ha, I don't even own a camera!
October 22, 2001 11:42:47 PM

12x faster my ass....RDRAM is 16But, ddr is 64Bit. In a year we'll have QDR at 550MHz. Geuss what'll happen by 2004. Intel uses DDR also (I say this anyway, although it was mentioned).

U got a problem?! Then dial 1800-328-7448!
October 22, 2001 11:50:18 PM

Hmmmm.... i was hearing about MRAM.... Some technology that will replace the harddisk and RAM.... making it super fast and no more bottlenecks.... DRAM is going to die sometime i hope...

Nice Nvidia and ATi users get a Cookie.... :smile: Yummy :smile:
October 23, 2001 12:45:50 AM

I believe MRAM's potential is in replacing flash memory, not necessarily good old DRAM. Current bottlenecks will be around for little while longer.



-----
It's a shame we have only one soul to sell...
a b à CPUs
October 23, 2001 2:38:23 AM

Sorry but MRAM drives are available today. Although they are still small (3.2gig is the largest I've read about so far)and expensive($1200usd) As the technology evolves they will get bigger and faster. And one day will probably replace hard drive and memory as one unit.

I aint signing nothing!!!
October 23, 2001 6:08:54 AM

i personally dont care much for rambus and their "lets sue everone alive on the planet makeing ram" attitude.

i wonder how long it will be before we see QDR DRAM in our graphics cards and put into a Nforce like mobo?

lets see...
133mhz qdr = 533mhz = PC4266 x 2 = 8.5mb/sec.


Religious wars are 2 groups of people fighting over who has the best imaginary friend.
October 23, 2001 6:23:28 AM

Quote:

i wonder how long it will be before we see QDR DRAM in our graphics cards and put into a Nforce like mobo?

lets see...
133mhz qdr = 533mhz = PC4266 x 2 = 8.5mb/sec.


Are there even plans for this? RDRAM should hit this level of bandwidth in the second half of next year with dual channel 32-bit-per-channel PC1066. Who is developing QDR? Regardless of what kind of memory technology it is, we do desperately need more bandwidth.

-Raystonn


= The views stated herein are my personal views, and not necessarily the views of my employer. =
October 23, 2001 6:55:07 AM

"Then again, you can always try and support the weaker underdog."

So your saying that somewhere there's a third player out there who is beating AMD **AND** Intel which makes Intel and AMD both underdogs of which, according to you, AMD is the weaker. I apoligise for my concise English which may land beyond the understanding of some PC forum cyborg. (well he does unabashedly admit to being "Intel Inside", frankly I'm all organic)

also, LHGpoobaa, you meant 8.5GIGAbytes didnt' you.
How come that doesn't sound very high, what do you get these days??
and hte x2 in your equation, was that referring to dual channel on the nForce.

Balzi


"I spilled coffee all over my wife's nighty... ...serves me right for wearing it?!?"
October 23, 2001 11:34:03 AM

your intel_inside arent u?

<font color=green>
*******
*K.I.S.S*
*(k)eep (I)t (S)imple (S)tupid*
*******
</font color=green>
October 23, 2001 12:28:18 PM

Post deleted by LoveGuRu
October 23, 2001 12:28:43 PM

rayson RDRAM is not same as DDR "just higher clocked", DDR is faster then RDRAM and costs less, wake up and smell the cofee stop being such an ass, your not convinsing any one here..

<font color=green>
*******
*K.I.S.S*
*(k)eep (I)t (S)imple (S)tupid*
*******
</font color=green>
October 23, 2001 12:48:06 PM

Guru,

why are you attacking Ray?

I have never seen Ray be anything but polite on the CPU message boards and he has always been very accurate and fair, with a slight Intel bent (while I try to be fair with an AMD bent.

I don't see where Ray said anything that contradicts what you just said.

The main difference between the two memory types id RDRam offers more memory bandwidth, while DDR Ram offsets the higher bandwidth of RDRam with much lower latency.

Now we see DDR Ram coming out with more bandwiddth, RDRam with more bandwidth and improved latency (supposedly) so, both memory types are becoming more competetive with each other.

There's really no need to flame Ray over your differing opinions.

Mark-

When all else fails, throw your computer out the window!!!
October 23, 2001 12:58:53 PM

That number has the unit Hertz after it. What 3.2 Ghz RDRAM means that the memory is transferring data at 400MHz but 8 times per clock cycle or the same as if it were single data rate and was running a frequeny of 3.2 Ghz. The bandwidth of this memory is 6.4 Gbytes per second. (presuming it still uses the 16 bit interface, so it would be 4 times faster.

Democracy Bernad, it must be stopped!
October 23, 2001 1:06:18 PM

he is trying to make the RDRAM as an advanced technology over the DDR and QDR and one who wouldnt know any better would take it as a fact.

even thogh im a big fan for AMD i dont go and trash INTEL 24/7, and i dont like ppl doing just the oposite about AMD.

<font color=green>
*******
*K.I.S.S*
*(k)eep (I)t (S)imple (S)tupid*
*******
</font color=green>
October 23, 2001 1:22:16 PM

I have not seen Ray trash AMD or AMD technology, other than presenting facts as they are available. Yes, Ray is an Intel advocate and thru that likely an RDRam advocate since that is what P4 uses for max performance.

But I don't see him trashing AMD or DDR or QDR at all. Asking who is working on it so he can get more information about it is NOT trashing it. Stating that RDRam is expected to have a particular bandwidth by a certain timeframe is not trashing AMD or non RDRam. It's just offering information as projected and publicly available.

Anyhow, no need to continue this I guess. It serves no purpose to argue over our mutual misunderstandings, ehh? :) 

Mark-

When all else fails, throw your computer out the window!!!
a b à CPUs
October 23, 2001 2:20:32 PM

QDR sounds good, especially in video cards. In systems it presents a problem as to how many modules can be supported.

Back to you Tom...
October 23, 2001 3:21:13 PM

i dont usually argue about these things but what the...
in response to Raysonn's post:

"Are there even plans for this?"

Yes.

"RDRAM should hit this level of bandwidth in the second half of next year with dual channel 32-bit-per-channel PC1066. Who is developing QDR? Regardless of what kind of memory technology it is, we do desperately need more bandwidth."

So what?who the hell would buy it then?no one is buying it now, Intel should say thanks to the P4x266 and the SDRAM supportive chipset, in my saying "wake uo and smell the cofee" i wasnt trying to trash you but to make you face the reallity, you dont see any OEM trying to make AMD support RDRAM do u?


<font color=green>
*******
*K.I.S.S*
*(k)eep (I)t (S)imple (S)tupid*
*******
</font color=green>
October 23, 2001 3:33:43 PM

poobahaha, I hear AMD is looking for rambus engineers, I say rambus is coming full circle! rambus for you and I ahahahahahahaha!

"<b>AMD/VIA!</b>...you are <i>still</i> the weakest link, good bye!"
October 23, 2001 3:57:38 PM

Quote:
rayson RDRAM is not same as DDR "just higher clocked", DDR is faster then RDRAM and costs less, wake up and smell the cofee stop being such an ass, your not convinsing any one here..


Ok, let's compare current platforms. PC800 vs. PC2100
Oops, RDRAM wins, to the tune of 3.2 vs. 2.1

Alright then, let's compare current technology. Hah! DDR is only one channel, which means it's faster! No, wait...RDRAM is only using a 16-bit channel, which means it's faster again.

Ok, let's compare future plans. Nah, it would just come to the same result.


Face it, (as much as I don't want to admit it, since I've been arguing the other side for months), RDRAM is in position to go significantly past SDRAM.

<font color=green>I post so you don't have to!
9/11 - RIP</font color=green>
October 23, 2001 4:54:47 PM

damn it burger almost had him there!?:p 

do u think i would argue by-"NO!DDR IS FASTER THEN RDRAM!"
bah, so sad *snif* and yet i still prefer DDR coz it has a nicer icon..
RDRAM on the other hand..


<font color=green>
*******
*K.I.S.S*
*(k)eep (I)t (S)imple (S)tupid*
*******
</font color=green>
October 23, 2001 5:16:27 PM

I think that is the problem, "should."

That is a very scary word.

<b>Egon, Remember that time you tried to drill a hole through your head? - Ghostbusters</b> :lol: 
October 23, 2001 5:22:32 PM

I love this discussion on Faster. RDRAM vs. DDRSDRAM.

Faster is not the issue. Performance, Performance, and again Perfomance.

Peak bandwith and actual are two totally different issues. Latencies, latencies, and latencies.(Man - Deja-Vu)

The Colorado river has huge bandwith at the rim or the grand canyon but it runs along the bottom. Latency!

Theoretical numbers look good on benchmarks but real world playability is all that matters. Who cares if the peak, rising edge and/or falling edge is high. Where does it run most of the time? No where near those hypo numbers.

Comments?

<b>Egon, Remember that time you tried to drill a hole through your head? - Ghostbusters</b> :lol: 
October 23, 2001 11:32:01 PM

hey guru...
if ya wanna make your links clickable put:
[ u r l ]
in front of the link, minus the spaces of course, and get rid of the http://
and at the end
[ / u r l ]

Religious wars are 2 groups of people fighting over who has the best imaginary friend.
October 23, 2001 11:44:55 PM

Yes raystonn... we need to kill latencies and boost bandwith...
cant say ive cared much for rdram thus far, but i suspect thats mostly due to its initial pricing and the way its parent company, rambust, has conducted its affairs.

hopefully gone are the days of the celleron 766 and its 11.5x multiplier... or the athlon 1400B and its 14x...


Religious wars are 2 groups of people fighting over who has the best imaginary friend.
October 24, 2001 12:37:32 AM

I think that Intel decision to use RDRAM with the P3 for initial Coppermine support may have colored the RDRAM introduction for some people.
October 24, 2001 1:15:29 AM

OK, someone on here needs to outline the MHz vs. Gbytes for RDRAM and DDR SDRAM.
what i have seen is people throwing around numbers without attaching any meaning to them??
it like arguing that a PII-400 is better than an Athlon with DDR266 cause 400 is better than 266. or my 10Gbyte HD beats your 2GHz P4.

it's all ludicrous. I would like to point out that charlie's quote of "3.2GHz RDRAM" is way out.

firstly, don't confuse real bit rates with manufactured bit-rates. that said, Charlie was on the right track breaking down the 3.2GHz to real figures.

let's start some real figures.
DDR266 (PC2100) is really running at 133MHz. but the RAM output lines change on up AND down edges of the strobe. now there's 64bits per edge, and 2 edges per cycle and 133million edges per second.
64 bits equals 8 bytes.

so 8 x 2 x 133 = 2128 (million) which is approx 2.1GB per second.

now I don't know as much about RDRAM.
btu here's a guess. RDRAM has 16bit data channels. FSB is 400MHz. rising/falling edges again.. but there's 2 channels or something .. not sure here..

but I think it's 400 million cycles times 4 bytes times 2 channels equals total *theoretical* bandwdith of 3.2Gbytes per second.

Q: why is it called PC800?? I think I missed something.

anyway, someone verify the RDRAM then we can start talking sensibly about the numbers and RDRAM vs DDR. and what's comign up.. how much a highe clock rate will help out both RAM types.. etc

Balzi

"I spilled coffee all over my wife's nighty... ...serves me right for wearing it?!?"
October 24, 2001 1:57:35 AM

Quote:
DDR266 (PC2100) is really running at 133MHz. but the RAM output lines change on up AND down edges of the strobe. now there's 64bits per edge, and 2 edges per cycle and 133million edges per second.
64 bits equals 8 bytes.

so 8 x 2 x 133 = 2128 (million) which is approx 2.1GB per second.


My description would have been:

PC2100 DDR SDRAM runs off a 133MHz external clock. The DDR (Double Data Rate) pushes this to an 'effective' 266MHz by transmitting twice per actual clock. The SDRAM channel is 64-bits wide, and so transmits 64-bits at a rate of 266MHz for a bandwidth of 2128MB/s per channel. One channel is provided for a total system memory bandwidth of 2128MB/s (2.128GB/s.)


PC800 RDRAM runs off a 400MHz clock. The DRCGs (Direct Rambus Clock Generators) on the motherboard provide this clock by quad-pumping off a 100MHz external clock. The DDR nature of RDRAM pushes this to an 'effective' 800MHz by transmitting twice per actual clock. The RDRAM channel is 16-bits wide, and so transmits 16-bits at a rate of 800MHz for a bandwidth of 1600MB/s per channel. Two channels are provided for a total system memory bandwidth of 3200MB/s (3.2GB/s.)

In the second half of next year, RDRAM will move to 32-bit channels and will run off a 533MHz (PC1066) DRCG clock, providing approximately 2.666 times current bandwidth levels (8.533GB/s.)

-Raystonn


= The views stated herein are my personal views, and not necessarily the views of my employer. =
October 24, 2001 7:55:53 AM

Hey, I do actually know what I am talking about. If you look further up the thread you will find that I have explained just what you have said in your post. I can;t remember who it was. Anyway you will find if you check that I have actually made the distinction between frequncey and band width quite clear.

The reason why PC800 RDRDAM is called PC800 is because the effective FSB frequency to the memory banks is 400 MHz. Because PC800 is actually double data rate, the marketing men have labelled it PC800. Returing to the reference of memory at 3.2 Ghz, the debate was about the developement of Octal Data rate memory or memory transferring data 8 times per clock cycle. So with the same quad pumped FSB, with effective clock frequency of 400 MHz, the memmory is essentially behaving as if it were SDR and seeing a FSB frequency of 3.2 GHz. Finaly band width is quoted with the unit bytes per second or bits per second (which is the same thing just divide bps by eight), frequency is measured in hertz.

So now you know, read all the posts, not just the last few.

Democracy Bernad, it must be stopped!
October 24, 2001 8:51:35 AM

Ok so we have had a little refresher on what the current types of ram actually are and do. So my question is what does this got to do with the original post? How does any of this amount to a setback for AMD? AMD or Intel could use DDR or RDRAM. Intel does not have any patent on RD and DDR does not = AMD. Both companies can use either type so what does it matter if RD is gonna be a lot better in the future?

Remember if you ain't Muslim you ain't Shiite.
Anonymous
a b à CPUs
October 24, 2001 9:51:10 AM

In the second half of next year, RDRAM will move to 32-bit >channels and will run off a 533MHz (PC1066) DRCG clock,
>providing approximately 2.666 times current bandwidth
>levels (8.533GB/s.)

Yep; however currently dual 266 DDR ram is already available giving us 4,2 GB/s low latency bandwith. Single DDR 333 is also on the market and combining both is no rocket science and could be implemented any day. There is just no cpu take advantage of all this bandwith just yet.

---- "Creative soundcards are excellent, <snip> It's worth the low price." (c) intel_inside
October 24, 2001 10:13:22 AM

RDRAM is a technology developed by intel, amd cant use it.

<font color=green>
*******
*K.I.S.S*
*(k)eep (I)t (S)imple (S)tupid*
*******
</font color=green>
Anonymous
a b à CPUs
October 24, 2001 11:21:59 AM

Get a clue. RDRAM is developped by Rambus Inc. Nothing to do with intel. AMD even has Rambus licences, but chooses not to use them.

---- "Creative soundcards are excellent, <snip> It's worth the low price." (c) intel_inside
October 24, 2001 4:48:36 PM

Quote:
Get a clue. RDRAM is developped by Rambus Inc. Nothing to do with intel. AMD even has Rambus licences, but chooses not to use them.


Yeup, exactly right. The question is, do Via and SiS have the rights to use RDRAM?

<font color=green>I post so you don't have to!
9/11 - RIP</font color=green>
October 25, 2001 4:02:31 AM

apologies for mistaking you. but you were following a forgotten train of thought is all.
as for the 3.2Ghz NEW memory.. well it seems that processors aren't the only thing with a name that could mean anything.
isn't it true that 3.2GHz will not have any signals that even come close to that mark.. it's all base values (real signal speeds) with all these manufactured mulitplying figures.

ah well, let 'em fight it out.. see who can stretch the customers imagination the furthest.

bazil (add some spice to your name)

"I spilled coffee all over my wife's nighty... ...serves me right for wearing it?!?"
October 25, 2001 4:14:24 AM

3.2GHz memory means the core of the memory module is working at 3.2GHz. The interface to the memory bus may be operating at a slower clockspeed with multiple bytes being transmitted per clock.

-Raystonn


= The views stated herein are my personal views, and not necessarily the views of my employer. =
October 25, 2001 7:09:32 AM

Yes, you are right. The crux of this demabte hinges on the concept of Octal Data Rates. So while the clock frequency of the FSB is still set at a relatively sedate 100 Mhz (then quad pumped to the equivalent of 400 Mhz). The combination of all these things makes it look like the memory is transferring data at a frequency of 3.2 GHz. But you are correct in thinking this is not the actuall speed but the effect if lots of mulitpliers.

Democracy Bernad, it must be stopped!
!