Unstoppable: DDR400 vs. Rambus

[fkloster]

The memory is ready...Intel is not ready to implement it...possibly because there is temporarily no need...

 

[Zakule]

<< Not true, as RDRAM and DDR SDRAM scale up, RDRAM's latency (in real-time, not the clockcycle penalty) will be lower than DDR's because of its higher scaling architecture. I can't remember the numbers off hand, but RD won't surpass DDR a while, not until PC1200 RDRAM, although PC1066 is VERY close (but that little boost combined with its already high bandwidth is enough to clean house). So really, the more time passes, the better for RDRAM. DDR's only hope is DDR-II. Well... that and vastly superior marketing and industry support (even RDRAM's big guns, Intel, is waning). >>



You are mistaken. PC1066 RDRAM suffers from the same latency issues and as far PC1200 RDRAM goes, chipset support for it may not ever come to pass. Intel is on the verge of completely phasing out RDRAM because of lack of interest by consumers. It's more likely than not that Rambus will not get the chance to correct the latency issue. I wouldn't hold your breath for faster RDRAM since Intel may not even support PC1066 with the 850E.

 

[Athlon4all]

<< You are mistaken. PC1066 RDRAM suffers from the same latency issues and as far PC1200 RDRAM goes it may not ever come to pass. Intel is on the verge of completely phasing out RDRAM because of lack of interest by consumers. It's more likely than not that Rambus will not get the chance to correct the latency issue. I wouldn't hold your breath for faster RDRAM since Intel may not even support PC1066 with the 850E. >>

I have heard something about 850e having no PC1066 support, but that is not true because in Asus' specs for the 850E based P4T533-E, it has PC800 and PC1066 support. Now on to more important issues.

Take a look at these Cachemem benchmarks on a modded Asus P4T running at 533fsb+PC1066. I know 845-D isn't included but this article was like 6-8 months ago, but these benchmarks are still valid. Look at the numbers, and for those lazy ones, here's the numbers on Cachmem latency:

PC800/400fsb: 270
PC800/533fsb: 247
PC1066/533fsb: 207
And for comparision, 845 PC133/400fsb: 229

So, yes, latency is significantley improved on the 850e+PC1066 platform.

 

[fkloster]

<< You are mistaken. PC1066 RDRAM suffers from the same latency issues and as far PC1200 RDRAM goes, chipset support for it may not ever come to pass. Intel is on the verge of completely phasing out RDRAM because of lack of interest by consumers. It's more likely than not that Rambus will not get the chance to correct the latency issue. I wouldn't hold your breath for faster RDRAM since Intel may not even support PC1066 with the 850E. >>



lol Zakule...for your 5th post @ anands....you sure have a lot of opinion in that statement...would you please supply us with a link to your accusation about Intel likely dropping rambus support all together? As far as latency goes, how on earth did you come up with that conclusion that PC-1066 suffers from same latency that PC-800 suffers from?

 

[oldfart]

I ran cachemem. How the heck do you read this thing?

Cache size/Memory speed info tool 2.65MMX - (c) 1999-2001, LRMS - DJGPP compiled
CPUID support detected... 'GenuineIntel' with FPU TSC MMX
Family=15 Model=2 Step=4 Type=0 Chipset (Vendor/Device ID(Rev)): Intel/1A30(04)
CPU clock: 2400.1 MHz
Using 32MB physical memory block (alignment = 32)
Bandwidth - MMX linear access test... Read/Write/Copy (MB/s)
Block of 1KB: 18899.7 / 8081.1 / 15928.0
Block of 2KB: 18600.1 / 8053.6 / 15558.7
Block of 4KB: 17897.8 / 7994.1 / 15688.2
Block of 8KB: 17640.8 / 8067.4 / 10438.0
Block of 16KB: 9756.1 / 8102.4 / 10440.3
Block of 32KB: 9758.6 / 8119.9 / 10439.0
Block of 64KB: 9761.1 / 8115.5 / 10443.2
Block of 128KB: 9763.1 / 8126.3 / 10433.4
Block of 256KB: 9716.4 / 8056.1 / 10285.7
Block of 512KB: 9352.3 / 7903.6 / 1111.5
Block of 1024KB: 2251.0 / 801.0 / 1111.6
Block of 2048KB: 2077.9 / 800.8 / 1111.6
Block of 4096KB: 2078.0 / 800.6 / 1111.6
Block of 8192KB: 2078.0 / 801.2 / 1111.8
Block of 16384KB: 2078.0 / 801.2 / 1111.7
Block of 32768KB: 2077.9 / 801.1
Latency - Memory walk tests... ("pointer chasing")
Null size: 2 cycles 0 cycles (overhead 376 cycles)
steps: 4 8 16 32 64 128 256 512 1k 2k 4k (bytes)
Block of 1KB: 1 1 1 1 1 1 1 1 - - - cycles
Block of 2KB: 1 1 1 1 2 1 1 1 1 - - cycles
Block of 4KB: 1 1 1 1 1 1 1 1 1 1 - cycles
Block of 8KB: 1 2 2 4 2 2 1 1 1 1 1 cycles
Block of 16KB: 2 3 9 18 18 21 20 22 23 22 2 cycles
Block of 32KB: 2 3 9 18 18 21 21 21 22 23 22 cycles
Block of 64KB: 2 2 9 18 18 21 22 21 21 22 23 cycles
Block of 128KB: 2 2 9 18 20 22 22 21 22 21 22 cycles
Block of 256KB: 2 2 9 18 18 22 22 24 25 26 26 cycles
Block of 512KB: 6 10 7 13 22 24 31 41 45 43 49 cycles
Block of 1024KB: 11 10 18 35 70 226 238 246 251 261 290 cycles
Block of 2048KB: 11 8 17 35 70 232 235 239 245 256 289 cycles
Block of 4096KB: 11 9 17 35 70 226 231 238 244 256 286 cycles
Block of 8192KB: 11 9 18 34 70 227 230 236 243 254 286 cycles
Block of 16384KB: 46 10 50 46 70 236 228 235 242 254 286 cycles
Block of 32768KB: 12 9 18 35 70 227 233 237 243 254 287 cycles
This system appears to have 2 cache levels (enabled).
L1 cache (8KB) speed (MB/s): Read=17897.8, Write=7994.1
L2 cache (512KB) speed (MB/s): Read=9716.4, Write=8056.1
Main memory speed (MB/s): Read=2078.0, Write=801.2

 

[Zakule]

<< lol Zakule...for your 5th post @ anands....you sure have a lot of opinion in that statement...would you please supply us with a link to your accusation about Intel likely dropping rambus support all together? As far as latency goes, how on earth did you come up with that conclusion that PC-1066 suffers from same latency that PC-800 suffers from? >>



Believe what you will, Flokster, but before you attack my credibility because I'm new to this forum you may want to run a few benchmarks yourself. I have no link to give you with reference to PC1066 because I determined this through my own benchmarks with real world apps. I'm new to posting on this forum because I've found many of Anandtech's benchmarks to be biased and not very reliable in general. The same can be said of Tom's Hardware Guide. One of the only places I've found the takes a truly objective look at most technology's is extremetech.com.

In spite of specs given on ASUS i850-E mainboard, Intel is still not sure whether or not they will officially support PC1066 RDRAM on this chipset. This has more to do with consumer disinterest than it does performance.

If the current market for RDRAM is any indication, Intel will probably drop RDRAM support altogether in 2003. As with anything in technology this is by no means a certainty, but it is more likely than not. DDR2 is going to bury RDRAM performance wise from what I've seen.

Perhaps when I've posted over 3,068 posts I might adopt the "wise" approach you have of attacking new forum members instead of staying on topic and giving credence to biased reviews because I do not have the means to benchmark technologys myself. Until then, though, I'm going to continue to rely on what I see with my own eyes vs. bar graphs on a review site's benchmark, which in most instances don't even come close to emulating a real world application environment.

 

[fkloster]

<< DDR2 is going to bury RDRAM performance wise from what I've seen >>



lol DDR2? what the hell? How long has DDR2 been out and can you provide a link...oh....oh ya, I just remembered, you don't have any links to your outlandish statements....how on earth could ANY ONE speculate intelligently that DDR2 (if and when DDR 2 is even developed...lol) would 'bury' 32bit PC-1200 in performance????? It seems very likely that you might now know what the hell you are talking about

 

[Zakule]

<< lol DDR2? what the hell? How long has DDR2 been out and can you provide a link...oh....oh ya, I just remembered, you don't have any links to your outlandish statements....how on earth could ANY ONE speculate intelligently that DDR2 (if and when DDR 2 is even developed...lol) would 'bury' 32bit PC-1200 in performance????? It seems very likely that you might now know what the hell you are talking about >>



At least your consistent. Once again your post does nothing but attack me and offers no relevent information to the discussion, whatsoever.

Yes, DDR2 is still in development and I am fortunate to have gotten a sneak peek at what it can do. That's all I can and will say about it, and again I can't give you link to a review site, which are seemingly your only sources of information - how unfortunate for you.

Believe what you will, but for your own sake I'd advise (as I would to anyone) to perform your own benchmarks so that you have them as a comparison against the more popular review sites. Review sites, in general, are part of a business and every business has their own political agenda and affliliations with manufacturers in the tech industry, which invariably leads to biased benchmarks.

If I continue this exchange with you, Flokster, I will undoubtedly be sucked into a flaming match with you and I have no desire to do that. It's a shame that forums like this often times suffer from the immaturity of individuals like yourself. You could've opposed my view point with at least bogus statistics you obtained from one of your precious reviews, but even that was beyond the scope of your debating skills.

 

[BDSM]

Zakule.. you rule man!.. Very well said!!

I agree with you. I find especially THG extremely biased. I think Anandtech is a li'l bit better but not much.

As usual ppl on this site believe that everyone reasons the way they do themselves, that everyone's a tech junkie that is prepared to spend insane ammounts of money to make an unnoticable performance gain (most ppl do have real lives!).. And to possibly get problems with stabilty, and warranty.

Personally I haven't seen a p4 system equipped with RDRAM in any store ever. They all come with DDR.. or even sdram!
And let's not forget that Intel still sells ALOT of p3's and celeron's! And also remember AMD's increasing market share!
When you think of it.. very few systems use RDRAM! yesyes.. there is the N64 and PS2, right?

Since many in this thread discuss pc1066 and pc1200, both TBA I should remind you that the "superior" RDRAM that manages to outperform DDR is a dual channel solution. Versus DDR that is single channel. Yes, yes.. there is the 7500 chipset that the RDRAM based i860 actually outperforms, however.. that is @ DDR200 speeds! There will surely be a dual channel DDR 266 solution very soon, if not DDR333 which will likely outperform pc1066 with a good margin.

Besides.. there are other factors that decide whether a specific technology will survive. Just look at 3DFX.. They had the best products for a long time.. Then they got greedy, decided to make all the cards themselves.... Before that they had many big name supporters.. and suddenly they were alone.. and look what happened..

The same thing is bound to happen with rambus.

Oh. there is also the issue of pricing.. Again.. most ppl ain't prepared to pay extra for a memory (most ppl hardly know what ram is!!) in order to increase the performance in quake 3 @ 640*480 resulution with ALL details turned off by 15%... And in other apps give no performance boost. Or even give worse performance!!

I can't see how RDRAM would be needed in any other way than when intel wants to squeeze out absolutely all the performance they can out of the p4, comparing with the athlon.

Also.. the Sledgehammer will have a 128 bit memory interface on die which will be extremely low latency and coupled with DDR 333 will yield 5.4 GB/sec.. Yeah.. that's right more than pc1200 dual channel will ever give.

RDRAM has no chance in the long run. It's just for performance hungry geeks that hunger for more frames in quake.

Oh.. checked the article out one again. It's been updated!

Check this out: http://www.tomshardware.com/mainboard/02q2/020501/ddr400vsrambus-06.html

Future chipset 1.. hmmm.. might that be the 845G? it performs better than the current i850 and not too far from 850e with pc1066..

The 845G oc'd with DDR 400 with cas 2 (that isn't too far away) should perform close to pc1200 which isn't available either.

And we are talkign SINGLE channel DDR here ok.. You think about how well a dual channel DDR 333 cl2 setup would perform!..
Are your mouths watering yet?

 

[Zakule]

<< Besides.. there are other factors that decide whether a specific technology will survive. Just look at 3DFX.. They had the best products for a long time.. Then they got greedy, decided to make all the cards themselves.... Before that they had many big name supporters.. and suddenly they were alone.. and look what happened.. >>



You tell it true, my good man. Except that Rambus is actually in a worse position right now than 3dfx ever was. To this day there at still 3dfx loyalists, because 3dfx knew how to present themselves to the industry. Rambus on the otherhand is its own worst enemy. They've done more to damage themselves politically than any of their competitors.

Looking at RDRAM from strictly a performance potential perscpective, it has a lot to offer. I was an avid supporter of RDRAM for P4 systems in the days when PC800 was new to the market. What soured me to them has to do with the shinanegins that they have participated in with Intel. To explain it simply, Intel intentionally crippled it's chipsets to perform below potential with DDR memory. Anyone that's compared SiS's recent P4 DDR chipsets to Intels own P4 DDR chipsets can see that this is the case. After making that determination myself, it was kind of hard to continue to advocate RDRAM. Again, it is a shame that the product is controlled by a company like Rambus because the design ideas behind RDRAM should've earned it a longer stay in the PC market. Latency issues notwithstanding (which worsen when 4 modules are used instead of just 2) which Rambus is working to correct (with only marginal success), RDRAM could've and should've been dominating the memory market at this point.

When you consider how unpopular RDRAM is among both computer enthusiasts and manufacturers and the fact that DDR offers better performance for most existing apps then it's logical to conclude that RDRAM is on its way out. As I stated earlier in this thread, though, this is by no means a sure thing. Many that opposed RDRAM from the start thought that it would be gone before the end of last year, and yet it's still here. I could be eating my words a year from now, but I doubt it.

 

[Athlon4all]

Zakule, I am not certain how you can attack PC1066's performance. You say THG and AT are biased sites ok, now lets take a look at the web. There have been a couple different PC1066 previews running on 533fsb, 2 I know of off the top of my head are from GamePC vs DDR400 and also the Article I linked to from Ace's Hardware(there's a newer article at Ace's as well, I can get a link). Both of them show a pretty noticeable performance increase by using PC1066 and 533fsb, and not just in Quake 3 Arena! Further, I will mention that I have yet to see Extreme Tech have benchmarks of PC1066 so where do you base your conclusion that PC1066 is not a good performer? Or am I mis understnading you and you think PC1066 is actually good stuff?

I do think that it is possible that Intel will let RDRAM die, but we'll see. I sure hope not.

<< Yes, DDR2 is still in development and I am fortunate to have gotten a sneak peak at what it can do. That's all I can and will say about it, and again I can't give you link to a review site, which are seemingly your only sources of information - how unfortunate for you. >>

Hmmmmm. Well, its kind of tough not to question your claim without a website. To be honest, there is so little info on DDR2 that its hard to tell. I suppose we'll have to wait and see.

<< Believe what you will, but for your own sake I'd advise (as I would to anyone) to perform your own benchmarks so that you have them as a comparison against the more popular review sites. Review sites, in general, are part of a business and every business has their own political agenda and affliliations with manufacturers in the tech industry, which invariably leads to biased benchmarks. >>

Well, for people like me, who love to keep up with computer stuff, but don't necessarily want to spend money on everything, its kinda tough to run benchys yourselves. I do disagree about Anandtech's benchmarks. THG, well, they're ify. Let me ask you this? What benchmarks do you use to determine performance? I am curious.

Fkloster, I do agree with you, but I do ask that you don't attack Zekule personally, but keep focused on the topic on hand ok

 

[dullard]

"Low latency memory is what matters most for 3D graphics rendering on today's systems and it is one area where RDRAM falls short. Unless they can improve that part of the technology RDRAM is going to disappear"
Latency for RDRAM is inversly proportional to frequency (just like higher rpm hard drives have lower latency). Thus PC1066 has 33% lower latency than PC800 and PC1200 has 50% lower latency than PC800. DDR will have little latency changes with higher speeds. Athlon4all's article shows PC1066 with lower latency than SDRAM (or DDR since DDR has the same latency as SDRAM)- why won't you believe the facts?

"Believe what you will, Flokster, but before you attack my credibility...Perhaps when I've posted over 3,068 posts..."
The number of posts has nothing to do with knowledge. Fkloster shouldn't have mentioned post count. However fkloster was correct that you were wrong in your post.

"I have no link to give you with reference to PC1066 because I determined this through my own benchmarks with real world apps...I'm going to continue to rely on what I see with my own eyes vs. bar graphs on a review site's benchmark, which in most instances don't even come close to emulating a real world application environment."
Since PC1066 isn't selling yet, how did you get it to benchmark (overclocked PC800 doesn't count)? What benchmarks did you run? What were the results? What would be real-world application environments in your opinion? How do the current office programs, games, and other frequently used benchmarks differ from your benchmarks?

"One of the only places I've found the takes a truly objective look at most technology's is extremetech.com."
I don't read that site, what does it say about PC1066? I do know a few sites have shown RDRAM in the speed lead or have concluded that RDRAM is better for the P4: Tom's Hardware, Anandtech, GamePC, Hardware Analysis, Hardware Central, Aces Hardware, Firing Squad (plus many more exist that I don't read). I realize that you don't trust some of those, but do you have one single link with PC1066 being slower from a site you trust?

"In spite of specs given on ASUS i850-E mainboard, Intel is still not sure whether or not they will officially support PC1066 RDRAM on this chipset."
From Intel itself:
The data bus on all Pentium 4 processors operates at 533 MHz or 400 MHz depending on the model. The 533 MHz system bus on the latest Pentium 4 processor has 4.2 GB/s of system bandwidth...with the Intel 850E chipset.
The 850E is Intel's RDRAM chipset. Only PC1066 RDRAM provides that bandwidth at the moment. Later DDR2 will and 533 MHz DDR will - but it is impossible that these will be available in the next week or so when Intel launches the I850E. Thus it can only be PC1066 RDRAM.

"This has more to do with consumer disinterest than it does performance. If the current market for RDRAM is any indication, Intel will probably drop RDRAM support altogether in 2003."
The consumer was distinterested when RDRAM cost double the price of DDR and had no performance difference. But now RDRAM is equal in price and provides a significant speed boost. RDRAM's marketshare has increased significantly in the last year. (Yes DDR's marketshare has also increased at the expense of SDRAM).

"DDR2 is going to bury RDRAM performance wise from what I've seen."
Can you give a link? When will we see DDR2? As PC1066 will be here in a week, it is silly to compare today's technology to something that won't be released for 1 year or more (if ever). That is like comparing the P3 to the Hammer.

 

[fkloster]

<< ...Once again your post does nothing but attack me and offers no relevent information to the discussion, whatsoever. >>



1) Zakule makes outlandish statement...
2) fkloster asks for link...
2) Zakule asks for link to disprove his outlandish statement but then defends his reasoning for not providing a link of his own...lol





<< Yes, DDR2 is still in development and I am fortunate to have gotten a sneak peek at what it can do. That's all I can and will say about it, and again I can't give you link to a review site, which are seemingly your only sources of information - how unfortunate for you. >>



Ummm, o.k. lets just say I have had a sneak peak @ what PC-1066 can do seeing as how I HAVE BEEN RUNNING IT FOR THE LAST 3 WEEKS!! ...how is that for hands on experience...are you some kind of secret agent man that spied on some lab that tested DDR2 lol how funny...



<< ...but for your own sake I'd advise (as I would to anyone) to perform your own benchmarks so that you have them as a comparison against the more popular review sites. Review sites, in general, are part of a business and every business has their own political agenda and affliliations with manufacturers in the tech industry, which invariably leads to biased benchmarks >>



Agreed, I have done so...



<< If I continue this exchange with you, Flokster, I will undoubtedly be sucked into a flaming match with you and I have no desire to do that. It's a shame that forums like this often times suffer from the immaturity of individuals like yourself. You could've opposed my view point with at least bogus statistics you obtained from one of your precious reviews, but even that was beyond the scope of your debating skills. >>



Why would I refute your unsupported statement with "bogus statistics" from review sites? That would not get through to you anyway because you do not credit those review sites...



<< Also.. the Sledgehammer will have a 128 bit memory interface on die which will be extremely low latency and coupled with DDR 333 will yield 5.4 GB/sec.. Yeah.. that's right more than pc1200 dual channel will ever give. >>



Oh my Lord...how do you consider 5.4GB/s faster than 9.6 GB/s (PC-1200)...?

 

[dullard]

"To explain it simply, Intel intentionally crippled it's chipsets to perform below potential with DDR memory. Anyone that's compared SiS's recent P4 DDR chipsets to Intels own P4 DDR chipsets can see that this is the case."
Tom's hardware (before it was changed) showed Intel's 845G chipset far, far above SiS's recent P4 DDR chipsets. It is now changed to 'future chipset' - probably at Intel's request. So how is SiS ahead in speed?

"Latency issues notwithstanding (which worsen when 4 modules are used instead of just 2)"
Latency differences between 4 modules and 2 is about 1%. That means a latency of 207 with two PC1066, a latency of 209 with four PC1066, compared to a latency of DDR in the 220 range.


"DDR offers better performance for most existing apps"
Please again, tell us what apps these are.

 

[Athlon4all]

<< Since many in this thread discuss pc1066 and pc1200, both TBA I should remind you that the "superior" RDRAM that manages to outperform DDR is a dual channel solution. Versus DDR that is single channel. Yes, yes.. there is the 7500 chipset that the RDRAM based i860 actually outperforms, however.. that is @ DDR200 speeds! There will surely be a dual channel DDR 266 solution very soon, if not DDR333 which will likely outperform pc1066 with a good margin. >>

Yes, but a key factor is that a Dual Channel DDR solution is much more expensive than a DC RDRAM solution due to the width of the buses (2x 16-bit bues for RDRAM where as 2x 64-bit buses for DDR). I will not argue that if a DDR solution has the same bandy as RDRAM, it will likely be faster, but the cost is a factor.

<< Personally I haven't seen a p4 system equipped with RDRAM in any store ever. >>

That's sadly because they have chosen the most cost-effective solution. Cost-effective, doesn't mean fastest.

<< Oh. there is also the issue of pricing.. Again.. most ppl ain't prepared to pay extra for a memory >>

I will point out quickly that high-quality 256Megs of RDRAM (I'd consider High-Quality Samsung) is mainly between $75 and $90, where Crucial, which most consider the best quality, is $88. The price difference is only in terms of mobo's, but that is very small as well. Now as for the price of PC1066, that remains to be seen.

<< I can't see how RDRAM would be needed in any other way than when intel wants to squeeze out absolutely all the performance >>

Take a look at the web. Every single site that has run benchmarks of 533fsb+PC1066 (not just THG and Anandtech)) has shown that there is a significant performance advantage for PC1066 when compared to SiS 645(DX) and DDR333, I just can't see how you can say that.

<< Also.. the Sledgehammer will have a 128 bit memory interface on die which will be extremely low latency and coupled with DDR 333 will yield 5.4 GB/sec.. Yeah.. that's right more than pc1200 dual channel will ever give. >>

Yes, but there's the previously mentioned cost issue for a Dual Channel DDR solution.

 

[dullard]

<< Also.. the Sledgehammer will have a 128 bit memory interface on die which will be extremely low latency and coupled with DDR 333 will yield 5.4 GB/sec.. Yeah.. that's right more than pc1200 dual channel will ever give. >>



<< Oh my Lord...how do you consider 5.4GB/s faster than 9.6 GB/s (PC-1200)...? >>



If PC1200 is 16 bit, it will have 4.6GB/s bandwidth.
If PC1200 is 32 bit, it will have 9.6GB/s bandwidth.
DDR 333 has 2.7 GB/s bandwidth.
Dual channel DDR 333 will have 5.4GB/s bandwidth.

Note that going from DDR 333 to dual channel DDR 333 will double bandwidth. Also note that RDRAM can double bandwidth by going to 32-bit. Thus the effects cancel out.

BDSM was trying to double only one and ignored the fact that the other will also double.

 

[Zakule]

<< Further, I will mention that I have yet to see Extreme Tech have benchmarks of PC1066 so where do you base your conclusion that PC1066 is not a good performer? >>



Extremetech reviewed PC1066 RDRAM back in November of 2001. It compared it to DDR333 P4 setups, as well as AthlonXP machines. Their benchmarks showed PC1066's superiority for app's like Quake 3, but it also pointed out it's weaknesses and briefly discussed latency issues (I think, it's been a while since I read the review). I would never say that Extremetech is without its biases, but to me they tend to take a more objective look at things. I rate them near the top of review sites, Tom's Hardware Guide is at the bottom, and Anandtech is somewhere in the middle.

For the time being, I'm going to have to let the rest of your inquiries go unanswered, and I can't blame you doubtedly my credibility because of this, nor do I particularly care. I'm hear to share information and hopefully, I can be more forthcoming with the details you're requesting in the near future, but if not I won't loose any sleep over it.

Each individual is going to believe or disbelieve the information that I'm posting here based on their own experiences and opinions, and I wouldn't have it any other way. My information is just as colored by my own opinions as anyone elses. I chose to go with my own findings over those of popular reviews if they conflict with one another, but I do understand that most people can't afford to get their hands on every new piece of technology that's released nor do they have the means to effectively test them it environments that will yield accurate results. Not my problem or my concern, though. I just tell it like it is.

During writing this post several others have posted asking me questions, and I'll try to answer at least one of them.

Q: "which apps perform better with DDR than Rambus"

A: Any 3D app with a directx8 engine is going to perform better with DDR than PC800 Rambus, period. This is a general statement, dependent on several other factors, but it is more or less true.

I've noticed that many of you in this thread can't get over quoting these ridiculously high memory bandwidth scores of RDRAM (9 Gb/s - so what?). You are completely ignoring the fact that high overall bandwidth means very little compared with memory latency since only a fraction of total DDR or RDRAM bandwidth is currently used by 3D games. It is going to be years before any game needs more than 2GB/s per second memory bandwidth across the graphics bus, and processor to memory transfers don't necessitate more than that amount either. The raw numbers that benchmarks spit out for performance on P4+1066 mean squat because no apps are using all of that bandwidth. Memory latency comes into play much more for todays 3D apps, and in that department DDR has RDRAM beaten. It is really that simple.

 

[TonyB]

If you already have a RDRAM based P4 system, chances are you wont need to buy the new PC1066 ram because your PC800 can probably overclock to that speed. However, Im really looking forward to the i850E chipset and the 533MHz/PC1066 solution and maybe even try overclocking the PC1066 to PC1200

 

[fkloster]

<< Any 3D app with a directx8 engine is going to perform better with DDR than PC800 Rambus, period. >>



ABSOLUTELY AND UNEQUIVICABY FALSE

3DMark2K1 SE is based on directx8 engine and without a shadow of a doubt yields higher scores for PC-800 rigs than it does for its DDR bretheren. You spew alot of crap & I am getting tired of it...goto madonion and do a search filtered by chipset....

 

[BDSM]

To Athlon4all & Co.

First of all: Dual channel pc1200 does NOT give 9.6 gig/sec in it's present configuration, but rather 4.8 gig/sec! You are talking about dual channel 32 bit RDRAM which is nowhere near to the horizon.

Yes, THG showed you the numbers. He shouldn't have. He could just as well have put on the numbers for DDR 2.. or 3.
Or DDR 87 or something and you would have bought it just as well.

Someone said DDR400 would not decrease latency compared to DDR333 whilst pc1066 RDRAM would compared to pc800 RDRAM.
You ppl must be stupid. Ofcourse it does. However the first modules will most likely carry a higher latency due to higher cas latenices and so forth.
It's gonna be a while before we see modules able to run @ 200 mhz with the agressive timings DDR333 modules can today.

The first DDR200 modules from crucial was cas 2 while the first DDR266 from crucial was not! See my point now??

Also.... I think I'm not the only one that has seen pics of KINGMAX DDR400 cas 2 rated modules.. So they can't be too far off.

Someone talked about the "high cost" of dual channel ddr. I don't believe that. Mobo's based on the n-force dual DDR266 chipset can be had for less than $130 here in Sweden. While the cheapest RDRAM mobo I can find is a good $40 more expensive!


So.. all in all: I'm right and you are wrong


oh.. do forgive me if I forgot to comment on something. I couldn't bother to read everything more than once.

 

[Acanthus]

not to burst anyones bubble, but, like the athlon. The Pentium 4 is going to have a "memory bandwidth ceiling" where if you go higher than a certain amount of bandwidth without increasing the fsb you will see futily small performance gains. take a look at an athlon XP with ddr333 cas2 vs and athlon xp with ddr300 cas2, very small performance increase (unless you increase the athlons FSB)

 

[fkloster]

<< First of all: Dual channel pc1200 does NOT give 9.6 gig/sec in it's present configuration, but rather 4.8 gig/sec! You are talking about dual channel 32 bit RDRAM which is nowhere near to the horizon. >>



1) PC-1200 does NOT have a 'present' configuration because it does not exist in the present yet. When Intel impliments PC-1200, it will come in 32bit & 64bit configurations STARTING @ 9.6gb/s throughput...

 

[Zakule]

<< 3DMark2K1 SE is based on directx8 engine and without a shadow of a doubt yields higher scores for PC-800 rigs than it does for its DDR bretheren. You spew alot of crap & I am getting tired of it...goto madonion and do a search filtered by chipset >>



Relax for a minute, Fkloster. We're not debating Palenstinians vs. Israel here, this a tech forum. Take a deep breath, now, everything is okay

Damn, now you've done it, I'm gonna have to address the ever popular and mostly useless 3DMark2001. I place 3DMark2001 and SiSoft Sandra in the same boat. Good for general performance testing but in no way does it reflect how your equipment will peform in any particular 3D app. I use it myself, but I always recognize it for what it truly is. Ask yourself this, Fkloster, do you have any idea what the end number 3dmark spits out really means? Of course not, because Madonion doesn't tell you. You just know higher is better, right? WRONG. The Radeon8500 has been creaming GeForce3 cards in the 3DMark2001 for over 6 months, yet the Radeon8500 continues to deliver inferior performance to GeForce3 in almost every current game title because of driver issues. Sure, the Radeon8500 kicks but in 3dMark2001, but what about Giants or Sacrifice or Black & White? Depending on your system configuration, you'll probably spend more time tweaking then playing with games like these.

You're placing entirely too much weight on benchmarking utilities, Fkloster. It would behoove you to look past them to what is actually happening on your monitor while you're playing a game.

Sorry to be the one to break this to you. Keep trying, you'll get me eventually.

 

[dullard]

<< The Pentium 4 is going to have a "memory bandwidth ceiling" where if you go higher than a certain amount of bandwidth without increasing the fsb you will see futily small performance gains. >>



Correct - we don't need much more bandwidth at the moment. However along with the higher bandwidth comes significantly lower latency and an increased fsb. These give a significant performance gain. Just look at the 14% faster games when going from PC800 to PC1066. Or 24% faster games when going to PC1200. Neither result is 'a futilly small gain' in my opinion. It isn't the bandwidth giving the boost, it is the lower latency and increased fsb.

 

[fkloster]

<< << Any 3D app with a directx8 engine is going to perform better with DDR than PC800 Rambus, period. >> >>





<< << Any 3D app with a directx8 engine is going to perform better with DDR than PC800 Rambus, period. >> >>





<< << Any 3D app with a directx8 engine is going to perform better with DDR than PC800 Rambus, period. >> >>



Zakule...now would you like to possibly change anything about this false statemnt? 100% of the Dx8 engines better on DDR?....do you want to save face or would you like to come clean now

Answer yes & no to the following questions please...:

1) Is Mad Onion's 3dMark2K1 SE engine directx 8 based? Yes or No
2) Do ddr platforms perform better in 3dmark than PC-800 based platforms? Yes or No
3) Zakule makes statements that are not true? Yes or No