DDR SDRAM More Bandwidth Efficient Then RDRAM

noko

Distinguished
Jan 22, 2001
2,414
1
19,785
Interesting change in SiSoft Saundra :
<font color=purple><b>A new version of a popular independent benchmark shows that DDR SDRAM may be considerably more efficient when transferring data to the CPU than RDRAM.

British company SiSoftware recently enhanced its widely used Sandra utility suite to include processor specific data streaming optimizations in the utility's bandwidth benchmarks. Bandwidth is the rate of flow of data, usually expressed in megabytes per second or gigabytes per second. In a note passed to beta testers, Adrian Silasi, CTO of SiSoftware stated:

<i>This minor beta marks one of the most important changes in the benchmarks so far; it is time to replace the ALU/FPU STREAM tests that have been around since the first release way back in 1997 with modern equivalents that take advantage of the EMMX/SSE/SSE2 streaming/prefetching instructions in order to show the real bandwidth of the current platforms.

While there have been SSE2 STREAM versions in Sandra for the P4 since Sandra 2001, these were disabled as there was no equivalent for the PIII/Athlon/Duron. With AMD, SGI and other entities/interested people help we have developed a common version that supports all common platforms, i.e. PIII/P4/Athlon/Duron/VIA C3.

Briefly (http://www.sgi.com/developers/library/resources/asc_cpu.html), the new tests work by streaming the data into the L1 cache while working on the previous data and posting the results directly to memory without "polluting" the caches. This way the true limit of useable memory bandwidth is approached....
</b></i></font color=purple>

Read the whole interesting analysis at Van's Hardware:
<A HREF="http://216.194.77.198/articles/2001/july/010720_Sandra_Bandwidth/010720_Sandra_Bandwidth.htm" target="_new">http://216.194.77.198/articles/2001/july/010720_Sandra_Bandwidth/010720_Sandra_Bandwidth.htm</A>

Northwood and DDRam may give Intel an edge, lets hope Intel ups the speed wide open.

<b><font color=blue>1.5</b></font color=blue> T-Bird
<b><font color=red>2.1</b></font color=red> P4 Speed
 

Crashman

Polypheme
Former Staff
Intel originally planned to find venders to provide DDR system ram at 400DDR, bypassing QDR. It does not look like they will stick to those plans, which really stinks. Intel will be limmiting themselves to 266DDR instead. Nobody but Intel has the power to push for 400DDR, and it looks like theyv'e given up.

Video killed my Radio Card!
 

Crashman

Polypheme
Former Staff
Info on which? Intel's 400DDR inquiry were mentioned in Tom's Hard News or one of his columns several months ago. Certainly you read his articles including news and do not want me to search all those?
Or the specualtion that they are dropping the 400DDR idea? That one would be based on the fact that the i845 is designed for PC133 or PC2100. Since nobody mentioned 400DDR (PC3200) on the i845, we have to assume it is either not supported by the chipset or not supported by industry.

Video killed my Radio Card!
 

zengeos

Distinguished
Jul 3, 2001
921
0
18,980
Interesting post. As far as the Intel P4 DDR Ram i845 chipset goes, however, it's doubtful that it will enhance P4's performance much. Yes, RDRam has high latency issues, which has caused P4 to underperform in some respects compared to Athlon. However, the differences between the two chips run deeper than that. Intel will be releasing an unproven chipset that will take time to mature, while nVidia will also be releasing their nForce chipset. It's doubtful i845/P4 will outperform Athlon/nForce.

Mark-

When all else fails, throw your computer out the window!!!
 

noko

Distinguished
Jan 22, 2001
2,414
1
19,785
Yes I found the article very interesting and it gives me some hope for the Northwood and DDR ram available chipsets. In the last two years Intel has not shown conclusively that they are a Desktop CPU leader in performance. Hopefully the .13micron Northwood with a redesign cache and hopefully other features will lift the stalemate bar between Intel and AMD. Frankly I don't think Intel is giveing AMD enough competition causing AMD to slow down their developement of top notch CPU's.

<b><font color=blue>1.5</b></font color=blue> T-Bird
<b><font color=red>2.1</b></font color=red> P4 Speed
 

flamethrower205

Illustrious
Jun 26, 2001
13,105
0
40,780
Hmm, Noko, about that 1.5 Ghz T-Bird, 2.1 Ghz P4, is that what acoprding to your calculations a P4 would need to be to measure up to a 1.5 Ghz T-Bird, or is it what you have OCed both to?

When I rule the world, Apple will only mean the fruit.
 

noko

Distinguished
Jan 22, 2001
2,414
1
19,785
Based on SiSoft Saundra data. For real programs I doubt I could make an equivalent comparison, because each program would behave somewhat different relative to each cpu. Does that answer you question?

<b><font color=blue>1.5</b></font color=blue> T-Bird
<b><font color=red>2.1</b></font color=red> P4 Speed
 

OmegaX

Distinguished
Jun 23, 2001
75
0
18,630
Since nobody mentioned 400DDR (PC3200) on the i845
Not to put you down or anything, I just wanted to point out that DDR Ram with a bus speed of 400 Mhz is not PC3200, rather it is PC6400.

Bus Speed * Bandwidth * Bits/Clock / 8 (to convert to Bytes)

PC1600 = 100 Mhz * 64-bit * 2 / 8 = 1600 Mb/s
PC2100 = 133 Mhz * 64-bit * 2 / 8 = 2128 Mb/s
PC2400 = 150 Mhz * 64-bit * 2 / 8 = 2400 Mb/s
PC6400 = 400 Mhz * 64-bit * 2 / 8 = 6400 Mb/s

Just wanted to make sure that everyone uderstands how the transfer rates works, hope I helped. Btw...that was a very informative post and appreciated it.

My System: <A HREF="http://www.anandtech.com/mysystemrig.html?rigid=9417" target="_new">http://www.anandtech.com/mysystemrig.html?rigid=9417</A>
 

OmegaX

Distinguished
Jun 23, 2001
75
0
18,630
I was correcting the statement of someone else above, the technology of PC6400 is non-existant for consumers yet. So, there is no point in answering your question.

My System: <A HREF="http://www.anandtech.com/mysystemrig.html?rigid=9417" target="_new">http://www.anandtech.com/mysystemrig.html?rigid=9417</A>
 
G

Guest

Guest
And who could afford this?
As far as real life apps please explain what you use as multiple programs and what everone else uses that gives the benchmark, then tell me whats real life senario? And how does one big app relate to another being run@the same time as many others in real time performance, this is what real users are interested in!
I have seen many bench marks, non are what relates to my needs. Lets get a real perspective on the whole thing.



<font color=blue>Is there</font color=blue><font color=orange> a forum</font color=orange><font color=purple> for forum addiction?</font color=purple><P ID="edit"><FONT SIZE=-1><EM>Edited by scotty3303 on 07/23/01 08:34 PM.</EM></FONT></P>
 
G

Guest

Guest
Sorry dude!
Wrong person.lol

<font color=blue>Is there</font color=blue><font color=orange> a forum</font color=orange><font color=purple> for forum addiction?</font color=purple>
 

Crashman

Polypheme
Former Staff
You misundertood me to mean 400MHz x DDR, instead of 200MHz x DDR = 400(at DDR), abreviated 400DDR. I definately did NOT say 400xDDR, so you can't correct me on this, unless you can come up with a better name. 400DDR is 200MHz memory times DDR which is 200x2x64/8=3200. And 400DDR, being only 200MHz actual clock rate, would be fine with 5 or 5.5ns memory. When companies come out with fast DDR memory, they always misname it at twice the speed.

Video killed my Radio Card!
 

Crashman

Polypheme
Former Staff
OH, so as I was saying, I never said it was 400MHz, just to further clarify, I said it was 400DDR, which is 200MHz. DDR rates are meant to compair the theoretical transfer rate to that of SDR, hence AMD's 266 DDR bus, which runs at 133.

Video killed my Radio Card!
 

slvr_phoenix

Splendid
Dec 31, 2007
6,223
1
25,780
[shudder]
Ugh. I hate how marketing for the computer industry has taken perfectly understandable scientific labeling and ruined it entirely. I miss the days when a measurement of the FSB's MHz actually meant frequency.

On another note though, isn't it true that RDRAM is only using something like a 16-bit bus, and if they upped it to the same 64-bit bus that DDR SDRAM uses, RDRAM's bandwidth would jump by 4 times as much?

<pre><font color=orange>Sunnova</font color=orange> <font color=blue>Beach</font color=blue>, <i>ain't life a beach?</i></pre><p>
 

OmegaX

Distinguished
Jun 23, 2001
75
0
18,630
Sorry, I miss read your post...I thought you said 400Mhz.
It was early in the morning, I got 4 hours sleep and did not have time to go out and get my coffee this morning. Anyway, I appologize for miss reading your statement.

My System: <A HREF="http://www.anandtech.com/mysystemrig.html?rigid=9417" target="_new">http://www.anandtech.com/mysystemrig.html?rigid=9417</A>
 

noko

Distinguished
Jan 22, 2001
2,414
1
19,785
The cost for the T-Bird 1.5 is very affordable. As for the P4 at 2.1 I have no idea.

Follow the following thread to see the data indicating this speed rating for the T-Bird compared to a P4. If you can bring to light any false data from SiSoft then please let me know.

<A HREF="http://forumz.tomshardware.com/modules.php?name=Forums&file=faq&notfound=1&code=1" target="_new">1.5 T-Bird = 2.1 P4 in speed or does it?</A>

<b><font color=blue>1.5</b></font color=blue> T-Bird
<b><font color=red>2.1</b></font color=red> P4 Speed
 

FatBurger

Illustrious
isn't it true that RDRAM is only using something like a 16-bit bus

Exactly correct, which is why it's so much faster (in MHz), but doesn't have the same overall bandwidth. Smaller datapath.

-----------------
Whoever thinks up a good sig for me gets a prize :wink:
 

noko

Distinguished
Jan 22, 2001
2,414
1
19,785
Seems like due to the limited serial 16bit bus RDRAM will hit a wall sooner then DDR ram. Like RDRAM went the wrong way to achieve its higher frequency but somewhat faster transfer rate. Maybe a reduce pin count but you can only push so much through a 16 bit path. Looking at the GF3 using 3.8ns with its cross-bar 256bit technology makes RDRAM look utterly anemic and antiquidated. DDR ram has already surpassed RDRAM transfer rates, maybe in the video card industry but it is already proven.

<b><font color=blue>1.5</b></font color=blue> T-Bird
<b><font color=red>2.1</b></font color=red> P4 Speed
 

Crashman

Polypheme
Former Staff
That's why RDRAM is dual channel in newer systems. What RDRAM needs is Quad channel. But Intel would do better with 400DDR or 400QDR (200x2 or 100x4) than with RDRAM simply because of lower latency. But wait, wouldn't PC3200 be faster as DDR than QDR simply because the latencies are also doubled or quadrupled? And why didn't anyone just use dual-channel SDRAM?

Video killed my Radio Card!
 

noko

Distinguished
Jan 22, 2001
2,414
1
19,785
Dual-channel SDRAM?

Excellent question, lower latency then DDR ram. ATI actually did this in their chipset but it was tied to the integrated graphics. A duel channel pc133 would have the same bandwidth as DDR2100 ram of 2.1gb/sec but in reality it would probably be about 20-30% faster in actual transfer rates.

Now you got me wondering. HMMMMMMMMMMMMMMMMMMMM

<b><font color=blue>1.5</b></font color=blue> T-Bird
<b><font color=red>2.1</b></font color=red> P4 Speed
 

lhgpoobaa

Illustrious
Dec 31, 2007
14,462
1
40,780
lol anemic

i also read that the p4 with ddr ram is initially only comming out with PC1600. yuck! (i suppose it matches the native 100Mhz bus speed of the P4)
i can see the evil hand of rambust in here again.

"i love the smell of Overclocking in the morning!" Says my Hamster.
 

Crashman

Polypheme
Former Staff
For most overclockers PC1600 support would be fine, because overclocking by 33% would not push PC2100 over it's intended speed, or PC2400 at 50% overclocking.

Video killed my Radio Card!
 

TRENDING THREADS