New RDRAM chipset "beats DDR by 50%" claim

[bgeh]

I think that's silly to assume without stressing the chipset with a *FAST* processor. I am NOT stupid enough to believe that this chipset will behave in an identical manner with a 3.06ghz P4 with hyperthreading.

You may choose to BLINDLY believe this. I am not complacent enough to do so. I want to see how fast this chipset performs on Intel's top of the line CPU. Nothing less will convince me of its "superiority" over RDRAM.

in what way does the 3.06GHz processor stress the RAM more than the 2.26GHz?
both use the same FSB, 133MHz, which i believe is the one which decides how much memory bandwidth is being taken up by the processor
the major differences between the 3.06GHz and the 2.26GHz that i know of is HT and higher multipliers

BTW for those who think RDRAM cant oc well, Here is 172 FSB, and Here is 165 FSB both are using the newer 32bit RDRAM as well.

Pics of RDRAM and Samsungs roadmap here note that recently DDR2 has been pushed back to 2005 actually, not really good as RDRAM will have no compatition

have you checked the CPU/Processors and Overclocking Forum recently?
Thugsrook, Mikki and a few others have got their P4-M chips up to 227MHz (effectively DDR 454)

 

[Ice9]

in what way does the 3.06GHz processor stress the RAM more than the 2.26GHz?

I'm surprised you even have to ask. A slower CPU will NOT stress the memory subsystem. We learned that with the i820 chpset that supported single-channel RDRAM. Everyone screamed about how much Rambus stunk because benchmarks were higher on SDRAM.

There was a reason for that. A 133mhz FSB in no way stressed an 800mhz memory technology. And CPU speeds simply weren't fast enough to reveal a bottleneck.

Take a look at synthetic SANDRA benchmarks. Memory throughput scales with CPU speed. However, like most synthetic benchmarks, it only tells half the story. If you notice all the SANDRA benchmarks of dual-channel DDR, they keep up and in some cases BEAT dual channel RDRAM...

However in REAL WORLD benchmarks, E7205 drags its ass. Again, years have passed since PC1066 came around, and DDR, with 256 traces on an expensive PCB, still can't beat this technology.

the major differences between the 3.06GHz and the 2.26GHz that i know of is HT and higher multipliers

Hyperthreading is a HUGE difference. Ask anyone who owns one and multitasks/uses multithreaded apps. I used to be a dual-cpu kinda guy. One of my favorite machines was a dual 933 P3 box I had based on the i840 chipset. No matter what was running, my machine ALWAYS felt responsive thanks to the operating itself supporting multithreading. With the kind of side work I do (Photography) processing images is a DREAM on a 3ghz hyperthreaded box.

Initially, I had a shuttle XPC machine for my 3.06HT, which was based on the 845GE DDR chipset. After upgrading to the 850E, the difference was *actually visible*, just like they're visible in common benchmarks.

 

[FishTankX]

OKay, now that i've read your posts, I realize that RDRAM is artifically expensive. That doesn't change how RDRAM is more expensive now. That doesn't change the fact that at no time in the foreseeable future will RDRAM be cost competitive with mainstream DDR. Not even PC800. That doesn't change the fact that right now the majority of the market is going with DDR.

And most importantly That doesn't change the fact that RAMBUS is a rediclously lititgious company that sues the pants off everyone they can to get their money.. why is why so many people would refuse to buy RDRAM even if it was cheaper and higher performance.

With 64 bit RDRAM (Which I assume will be called RIMM 9600-64, heh, just a wild guess) I find it hard that to believe that the boards will be cheap.

But answer me this, Ice9. My final riddle.

Why would SiS bother the R659 to support 16GB of RAM if it's really using the single module aproach? You would need 32 512MB modules to reach that limit, seeing that RAMBUS isn't avaliable in anything over 512MB right now.

It seems absurd that this would be targeted at anything lower than mid range workstation.

And do you have any explenation for why PC1066 is so darn expensive? I could only gawk in horror at what PC1200 would cost.

 

[FishTankX]

Originally posted by: Ice9

in what way does the 3.06GHz processor stress the RAM more than the 2.26GHz?

I'm surprised you even have to ask. A slower CPU will NOT stress the memory subsystem. We learned that with the i820 chpset that supported single-channel RDRAM. Everyone screamed about how much Rambus stunk because benchmarks were higher on SDRAM.

There was a reason for that. A 133mhz FSB in no way stressed an 800mhz memory technology. And CPU speeds simply weren't fast enough to reveal a bottleneck.

Take a look at synthetic SANDRA benchmarks. Memory throughput scales with CPU speed. However, like most synthetic benchmarks, it only tells half the story. If you notice all the SANDRA benchmarks of dual-channel DDR, they keep up and in some cases BEAT dual channel RDRAM...

However in REAL WORLD benchmarks, E7205 drags its ass. Again, years have passed since PC1066 came around, and DDR, with 256 traces on an expensive PCB, still can't beat this technology.

the major differences between the 3.06GHz and the 2.26GHz that i know of is HT and higher multipliers

Hyperthreading is a HUGE difference. Ask anyone who owns one and multitasks/uses multithreaded apps. I used to be a dual-cpu kinda guy. One of my favorite machines was a dual 933 P3 box I had based on the i840 chipset. No matter what was running, my machine ALWAYS felt responsive thanks to the operating itself supporting multithreading. With the kind of side work I do (Photography) processing images is a DREAM on a 3ghz hyperthreaded box.

Initially, I had a shuttle XPC machine for my 3.06HT, which was based on the 845GE DDR chipset. After upgrading to the 850E, the difference was *actually visible*, just like they're visible in common benchmarks.

EXCUSE ME?! "Again, years have passed since PC1066 came around"? What is that based on? If I remember correctly It's been no longer than a year since the introduction of the first i850E board.

And what does hyperthreading have to do with which one would win out, between the i850E and the SiS655? Lemme guess, more threads = more work done = more memory access.. right? Anyways, it makes sense that the hyperthreading P4 would have less cache per individual thread so with less cache you might have more memory access.. anyways.. The B stepping's are comming out and i'm sure that since the P4 3.06 is the only one that supports hyperthreading, and they'll want to test hyperthreading (Even AT wants to test hyperthreading on the SIS655) you'll have your reviews soon. I predict that the SiS655 will atleast tie the PC1066 system while at the same time offering a problem to the prohibitivley expensive 512MB module problem (The PC1066 modules in the 512MB variety are nearly 2.5X as expensive as even DDR400. Why on earth would that be?! I still have yet to figure that out. Maybe you could enlighten me on why you get a mildly superlinear increase in price when going from 256 to 512 while DDR 400 is
actually a little less than linear..)

However in REAL WORLD benchmarks, E7205 drags its ass. Again, years have passed since PC1066 came around, and DDR, with 256 traces on an expensive PCB, still can't beat this technology.

Excuse me, but did you just say 256 traces? DCDDR is 128 traces. Each DDR module has 64 traces running to the chipset. 2X64=128.

You also have to realize that the cost of more traces is also negated because they don't have to run at such high frequencies, so designing the trace paths for the traces is easier than PC1066 with it's crazy 533MHZ clock, versus PC3200's more modest 200MHZ clock.

 

[bgeh]

I'm surprised you even have to ask. A slower CPU will NOT stress the memory subsystem. We learned that with the i820 chpset that supported single-channel RDRAM. Everyone screamed about how much Rambus stunk because benchmarks were higher on SDRAM.

There was a reason for that. A 133mhz FSB in no way stressed an 800mhz memory technology. And CPU speeds simply weren't fast enough to reveal a bottleneck.

that's because the P3 and the P4 are completely different architectures.
the current P4 actually runs on a 133MHz FSB(4 x 8) = 4.2GB which is what PC1066 and DCDDR 266 provide
and both the 2.26GHz and 3.06GHz use that FSB

 

[Ice9]

And most importantly That doesn't change the fact that RAMBUS is a rediclously lititgious company that sues the pants off everyone they can to get their money.. why is why so many people would refuse to buy RDRAM even if it was cheaper and higher performance.

Ridiculously litigous?

Hmm... Let's see... ONE lawsuit so far, 2 pending (one vs. Micron, one vs. Hynix)... that's 3 total.

Let's look at Intel. How many times have they sued over patent infringement? Let's put this in perspective:

They sued Broadcom
They sued VIA on SEVERAL occasions...
They sued EMI
They sued DEC
They sued AMD
They sued CYRIX
They sued SILICON STORAGE

And you have the balls to say RAMBUS is Litigous? You're a funny guy

Patent lawsuits happen *ALL THE TIME* between companies. It comes with the territory in Engineering.

With 64 bit RDRAM (Which I assume will be called RIMM 9600-64, heh, just a wild guess) I find it hard that to believe that the boards will be cheap.

They are still cheaper to manufacture than 128bit DDR DIMMs. Whether or not they will SELL as such depends on market competition. Rambus does NOT set pricing. They only charge a 1.5% royalty.

Why would SiS bother the R659 to support 16GB of RAM if it's really using the single module aproach? You would need 32 512MB modules to reach that limit, seeing that RAMBUS isn't avaliable in anything over 512MB right now.

You are making the assumption that higher densities aren't possible. There simply hasn't been a demand for anything higher than 512MB in a RIMM formfactor. However, higher densities are available if you're in the market for a Cray X1 or an Alpha EV7.

Just because the R659 CAN support 16GB of RAM, doesn't mean it WILL. Especially when you consider the 4GB limitation of current x86 processors. I think SiS' point might be "We're ready for 64 bit if the need arises" (though I don't think it will).

And do you have any explenation for why PC1066 is so darn expensive? I could only gawk in horror at what PC1200 would cost.

Let's see. $174ish for 512MB for Non-generic. (no such thing as Generic RDRAM thank god)
$84 for a 512MB stick of PC2700 crucial. I mean hey, you can get it for as low as $50 if you don't give a crap about what you're getting. Most enthusiasts DO, right?

Since i'm not of the "younger" generation that wrestles with a $1500 PC purchase, the additional $80 to have the best memory available is INSIGNIFICANT over the cost of my entire 3.06HT system. If you're going "on the cheap", then sure, buy the cheap stuff.

 

[FishTankX]

Anyways, Ice9, you're an intelligent oponent, but i'm arguing for the fact that DDR is the way of the present. Not the way of the future, like you may interpret. I highly doubt there would be any visible difference between a SiS655 B board and an i850E board each running their respective 'Top speed' memories. But 512MB modules for the i850E, and even 256MB modules, are expensive compared to DDR. It just doesn't seem like, with the SiS655 and canterwood on the horizon, DDR will be a part of our normal lives in the near future. By the time the R659 is released, with 64 bit PC1200 in tow, the canterwood would have been released for *atleast* 4 months.

I highly doubt DDR will mean anything 2 years down the road, and I highly doubt even single channel DDR-II will be able to keep up with RDRAM once Intel pulls out the 1200MHZ FSB. But that has no bearing on my next purchase as I plan to get rid of my RAM upon my next motherboard purchase. You may argue for RDRAM superiority. I'm not going to challange that. But it just seems that at present RAMBUS is in a bit of a dryspell. It's not a significant market force.

And where on earth did you get your 'millions of i850E' argument, with the Dell machines? I highly doubt they'd be able to reach that volume, armed with PC1066! And with PC800, the i850E can barley beat out the i845PE, much less fend off the Granite bay and *definatley* not the SiS655.

 

[codehack2]

Originally posted by: DX2Player

Originally posted by: Novgrod
DX2player, last i checked i don't have a problem. i'm just going to hold off judgement on rdram as the be-all end-all of memory solutions until more cpus or gpus than those based on the p4 use it to great effect.

I dont think I ever said it was the end all be all memory choice either but to deny its potential simply because a unjustified bais against the company is rediculas. How happy would you be if they said your beloved DDR solution was able to push 9.6BG per second by the end of the year. The truth is the forums would be lit up and everyone would be extatic about the prospect, including me. But since it isnt DDR its downplayed by the likes of you. Honestly would you respond in the same manner if it was DDR instead of RDRAM?

Neutral on the subject, but last I checked, A DDR equipped Radeon 9700pro is pushing on the upwards of 20gb per second....

CH2

 

[DX2Player]

RDRAM is not expensive you jack asses. Lets just say it cost me $100 more to put my system together than the top end DDR solution 7 months ago. How would that top end DDR system compete now? How would the RDRAM system compete now? Seems that after 7 months the RDRAM system is still compared as the thing to beat. The top of the line single channel DDR 333 needs to be replaced in order to stay competative, now tell me which one costs more. The RDRAM solutions cost more to start but they last longer. Ya I already said that RDRAM has been ousted by the SIS R655 but it took a full year to do so. Instead of crying over the extra cost in making the computer why dont you use some dicression next time you go to the game store youll save a lot more money that way.

 

[dexvx]

Originally posted by: bgeh

I'm surprised you even have to ask. A slower CPU will NOT stress the memory subsystem. We learned that with the i820 chpset that supported single-channel RDRAM. Everyone screamed about how much Rambus stunk because benchmarks were higher on SDRAM.

There was a reason for that. A 133mhz FSB in no way stressed an 800mhz memory technology. And CPU speeds simply weren't fast enough to reveal a bottleneck.

that's because the P3 and the P4 are completely different architectures.
the current P4 actually runs on a 133MHz FSB(4 x 8) = 4.2GB which is what PC1066 and DCDDR 266 provide
and both the 2.26GHz and 3.06GHz use that FSB

The processor is in more demand for the data, as it has the ability to process more data. A P3-533EB doesnt nearly require the same amount of data as a P3-1400 Tualatin. You can stick both those on a Tualatin motherboard, however, the Tualatin would saturate its 133FSB bus far more than the p3-533EB, just like a 3.06HT will saturate the 533FSB far more than a 2.26B.

Also the theortical throughput of DC-DDR is misleading. It has an awesome read speed, which is why its benchmark figures are so bloated. DC-DDR226 for example should have the same bandwidth as DC-PC1066, however, the DC-PC1066 has higher average throughput. Based on Linpack tests from Aceshardware, the DC-PC1066 is more efficient all around, meaning a higher average memory throughput (which is why teh general consensus is that DC-PC1066 > DC-DDR266). RDRam is slower in memory read tests, but it absolutely beats the crap out of DC-DDR in write tests.

About the litigation, every major corporation is always in the middle of some kind of litigation. How else do you think America keeps all the lawyers floating around? At Intel Jones Farm Hillsboro, there is a whole freaking floor thats for lawyers. If anyone thinks their beloved company never sues then they must live in some kind of dreamyland.

The fact of the matter for me was, I built a P4T-E/512MB PC1066 system in March, and its still basically top of the line performance wise. If I went i845D, I would've paid more for PC2700 512MB memory (at the time), which probably doesnt overclock as well anyways, on an outdated i845D chipset.

 

[FishTankX]

Wait a moment. If you start to look at it from the stand point of 1GB of RAM in a 512MBX2 configuration, (For future upgrades or extreme use) you start to have problems.

Name brand 512MB PC3200 (Samsung) costs about 90$. here. Namebrand PC1066 512MB costs 180$. I find that slightly unsettling, don't you? There's no reason why it should be expensive. Not even 'Artifically' expensive means that a namebrand (Samsung is a namebrand, aint it? ) DDR400 module should cost nearly the same as an equivalent PC1066 module at *half* the capacity. Don't you find that highly disturbing?

I fail to see why PC1066 is so expensive when clearly the SiS655 manages to keep within a few percentage points of the i850E (I'll give you the benefit of a doubt, that the i850E beats the SiS655 by a few percent, even when armed with DCDDR400 (Which should reduce latencies significantly over DCDDR266 which the granite bay has, Although you seem very 'intense' to me, ask you say that the Granite bay 'Drags it's ass' when it's only 2-3% slower.) while keeping memory costs at *half*.

Does artifically expensive really have to mean 2X?

If I want to get a a gig of RAM for my DCDDR400 rig I can just pick up 2 samsung DDR400 modules at 180$. A big improvement over the 360$ I would spend on PC1066. Not to mention that there's a gigabyte board that undercuts the nearest i850E board by nearly 50$. It all adds up.

 

[DX2Player]

dude do you think $90 more is worth having the top of the line system for over a year? Ya at the very moment RDRAM might be beat. but only at this moment, the past owned DDR and the future will own DDR. Frankly its a poor time to make a computer anyways with the new Prescott and AMD chips comming but whatever.

Besides this article is about the future not the moment. Im arguing that RDRAM has potential because of its past, and the article points out it will rise again so your flimsy argument that DDR is the best at this moment really does nto belong here.

 

[HokieESM]

Originally posted by: dexvx


The processor is in more demand for the data, as it has the ability to process more data. A P3-533EB doesnt nearly require the same amount of data as a P3-1400 Tualatin. You can stick both those on a Tualatin motherboard, however, the Tualatin would saturate its 133FSB bus far more than the p3-533EB, just like a 3.06HT will saturate the 533FSB far more than a 2.26B.

Also the theortical throughput of DC-DDR is misleading. It has an awesome read speed, which is why its benchmark figures are so bloated. DC-DDR226 for example should have the same bandwidth as DC-PC1066, however, the DC-PC1066 has higher average throughput. Based on Linpack tests from Aceshardware, the DC-PC1066 is more efficient all around, meaning a higher average memory throughput (which is why teh general consensus is that DC-PC1066 > DC-DDR266). RDRam is slower in memory read tests, but it absolutely beats the crap out of DC-DDR in write tests.

I see a LOT of comparisons with DCDDR and RDRAM in computational mechanics. Lots of calculations.... and in many tests, the amount of memory isn't that critical, but the memory bandwidth is. That's why almost every "workstation" around here is a P4 with PC1066.... I can physically SEE the difference (when you're doing things that takes DAYS.... a few % makes a noticeable difference) between i850E and Granite Bay workstations. PC1066 is wonderful for the kind of computational work I do... evidently its VERY well matched to the 533fsb P4s. Now would any of this matter to the average user? No. Even the power-hungry user? maybe not--depends on quantity versus bandwidth needs. Granite Bay allows addition of quantity at lower cost (and this is important if you're going to the swap file)... but, in my experience, if you need bandwidth for extended periods of time, PC1066 is hard to beat.

I don't know why everyone is so anti-development in reference to Rambus. Sure, I hate the company... but their technology is good. And it is PROBABLY the way of the future. For EVERYONE but the enthusiast, this argument is moot anyway--PC2700 is a much better standard for "mom and pop" people... not PC1066, DCDDR or whatever. Memory bandwidth just isn't critical to them.

Just as a note: one should promote ANY kind of research/advancement. Even if it is counter to what you LIKE. Because (as the AMD people like to point out): competition is the ONLY way to make things succeed. Encourage Rambus. Encourage DDR makers. Let them battle.... let them fight. Because, basically, all us consumers will be the ones who benefit.

 

[FishTankX]

DX2 player: I really doubt resorting to name calling is going to improve anyone's impression of you. For the record books, I no longer believe in 'Top of the line', and i'm now strongly considering AMD.

I used PC800 9 months ago when I built my system. I paid a 100% royalty to do so.

P4T-E (Back then) 190$
128X2 PC800 (Back then) 120$

MSI i845D system (Back then) 120$
1X256MB module 50$

It wasn't worthit..

RDRAM was and is expensive, if you want to compare on equal grounds. If I want to arm my rig with a gig of RAM (Which i'm about to be upgrading to soon, so that's my benchmark) i'd be paying...

2X512MB Samsung DDR400 180$
Gigabyte SiS655 minimalist board 120$
That's 300$

Versus..

Iwill i850E board 125$
2X512MB PC1066 360$
That's 480$

50% more expensive.

That's enough to upgrade your videocard from a Radeon 9500 to a 9700 pro, or a 17 incher to a 19 incher.

Somehow I doubt you're 'RAMBUS isn't expensive' argument works. Now, if you're arguing 'DDR266 is outdated', it can be used to great punch in either a granite bay board or a SiS655 board. If I had gone with the i845D board I could have carried my RAM over to my new system and still have had the new system with the old RAM pack more of a punch than my poor old P4T-E.

Also..

Ice9: Is the reason why the i850E boards, with 1/4th the memory traces, are equally expensive with the SiS655 boards, perhaps because memory traces don't significantly impact board price, maybe? HMM!! I wonder. Well, they do, but not on the scale that you would believe. Maybe 15$ max.

And you also have to remember that frequency impacts the cost of designes almost as much as trace counts. It'd probably be equally difficult to produce (I know, not design) a DDR 400 (200MHZ) DDR interface that ran 128 bits wide and a PC1200 (600MHZ) interface that ran 32 bits wide. (Fair comparison, I believe.)

 

[DX2Player]

LOL you buildt a PC800 RDRAM system 9 months ago, now thats not to bright, thats like making a single channel PC2100 system now and wondering why its not top of the line in 9 month, especially considering that PC1066 and PC800 cost the same you ripped yourself off and trying to blame RDRAM for it

AND FOR THE LOVE OF GOD ALMOST NOBODY WHO IS CONCERNED ABOUT $$$ USIS A GIG OF RAM, man your complaining about $ when your willing to buy a whole system for a few fps in 1-2 games, LOL if you care about $ so much the cheapest thing would be to just add more ram to the system you have.

 

[codehack2]

Originally posted by: DX2Player
AND FOR THE LOVE OF GOD ALMOST NOBODY WHO IS CONCERNED ABOUT $$$ USIS A GIG OF RAM

Your painting with a pretty broad brush here buddy... when you can pick up a 1GB of PC2700 for less than $130, I'd hardly lump it into the $ is no object category.

CH2

 

[Ice9]

Name brand 512MB PC3200 (Samsung) costs about 90$. here. Namebrand PC1066 512MB costs 180$. I find that slightly unsettling, don't you? There's no reason why it should be expensive. Not even 'Artifically' expensive means that a namebrand (Samsung is a namebrand, aint it? ) DDR400 module should cost nearly the same as an equivalent PC1066 module at *half* the capacity. Don't you find that highly disturbing?

Samsung is the #1 manufacturer of RDRAM. They have been able to sell RDRAM at a profit.

I'll tell you what's unsettling:

Hynix, the #3 memory manufacturer employing thousands, posting HUGE LOSSES:
"The chip maker reported a net loss of 917 billion won ($788 million) in the three months ending December, compared with a shortfall of 1.37 trillion won a year ago and 617 billion in the third quarter. Analysts were expecting a loss of around 500 billion won."

Micron, the #2 memory manufacturer, employing thousands, LAYING OFF HUNDREDS, after posting HUGE LOSSES to the tune of $2B over the past 2 years alone...

Samsung, Rambus' biggest partner, happily makes a profit, but still has lower revenues as a result of the market downturn. Their profits go up and down based on normal market factors. Normal market factors that Micron and Hynix, the #2 memory manufacturers in the world, CANNOT DEAL WITH because they have NO profitable technology to manufacture.

This is where your DDR prince is today. Driving companies into the GROUND over a 3% royalty. Rambus is PRICED APPROPRIATELY. DDR is NOT. It is SELLING AT A HUGE LOSS.

Can you not comprehend that? Or are you so blinded by its cheapness that you don't CARE if the people working for these companies LOSE THEIR JOBS because you *gotta have your cheap DDR?*

I fail to see why PC1066 is so expensive when clearly the SiS655 manages to keep within a few percentage points of the i850E (I'll give you the benefit of a doubt, that the i850E beats the SiS655 by a few percent, even when armed with DCDDR400 (Which should reduce latencies significantly over DCDDR266 which the granite bay has, Although you seem very 'intense' to me, ask you say that the Granite bay 'Drags it's ass' when it's only 2-3% slower.) while keeping memory costs at *half*.

It had 2 years to beat RDRAM. It hasn't done so. Case closed.

Does artifically expensive really have to mean 2X?

Again, see DDR's losses above.

If I want to get a a gig of RAM for my DCDDR400 rig I can just pick up 2 samsung DDR400 modules at 180$. A big improvement over the 360$ I would spend on PC1066.

And another nice little nail in a person's career coffin. All praise cheap DDR! Let the companies DIE!

This is why RDRAM will survive, and DDR will not.

 

[FishTankX]

Oops.

Miscounted.

I created a PC800 system 13 months ago. It was designed 14 months ago. Sorry for the confusion.

To clear up all confusion, it was designed Janurary 2002 and assembled Feburary 2002. At the time, the northwood was unreleased, the i850E didn't exist, the i845D was fledgling, and I was convinced that PC800 would give me the performance edge. That's why I bought it.

Same thing with my Radeon 8500 64MB. It costs a shocking 300$. Remember those days?



For you're information, PC800 and PC1066 don't cost the same. The average 512MB PC800 module costs 135$. The average PC1066 module costs 175. For the 256MB variety, it's 55/80. It's not the same. And *I* care about the price of a gig of RAM. Espically when in any given system designs excluding granite bay and alot of price points you can stuff in double the amount of DDR you can RDRAM, up to 2GB. I'd feel much happier running a gig of DCDDR 400 than running 512MB PC1066, and they cost the same.

 

[DX2Player]

If your uptight about spending any cash your not gonna waste $65 for no reason, why not just get 500W PCU and 200GIG HD its overkill and unless your not uptight about $ your not gonna do it.

Excuse me I thought we were talking 9 months ago, 9 months ago they were the same.

You dont seem to understand Im not trying to get you to buy RDRAM right now, I already said 2 times that if I were to build a system right now it would be the R655, the article is talking about RDRAMs future! The argument should be 2 channel DDR vs 4 channel RDRAM, what would you choose then?

 

[kuk]

I'm surprised no one has said this ... but ...

GUTB??? :Q

 

[FishTankX]

Well, DX2 player, I have, for now.. decided that the price of RDRAM doesn't justify it's performance benefits for gaming. So for me it would still have to be the DDR sollutions, assuming current pricing trends (RAMBUS atlesat 50% more expensive than DDR) I would pick up the R655.

If I were a video editor, or someone who was into heavy duty work, I would probably pick Quad channel RDRAM. Assuming it was reasonably priced and had a significant performance benefit over DDR. That's the part i'm not sure about.

We have all seen what happens when you pair up excessive bandwidth with a FSB that cannot take it. (Nforce 1/2) At the time that chipset is released, the P4 FSB will still be at 800MHZ FSB, thus a max of 6.4GB/s which is satasfied with DCDDDR. The only gains will be in ultra heavy 3d applications. When the P4 hits 1066/1200MHZ FSB then Quad channel RDRAM will truly be able to 'Stretch it's wings' and beat the living daylights out of DDR. That I can whole heartedly believe. I really doubt, when the R659 is released though,

If the performance benefit is minimal (Less than 3%) I would still pick DDR even if QC RDRAM did provide a performance benefit. The performance benefit doesn't justify the cost, for me. Anymore. Used to think that $$$ spent on high performance hardware was $$$ well spent. Now I see that it's more economical to stay middle end and upgrade twice as often.

Let me emphasize, In terms of which one is better, i'd have to say that the R659 will probably beat the 655 and the canterwood. But first of all, by how much? Second of all.. unlike Ice9 I don't have that much money to throw around. I'm not even in college yet. This computer was a gift from my parents. I won't own one for the same reason I haven't bought a i850E board. If I had lots of money to burn on nothing, sure, i'd go for it.

If I was grownup, had a well paying job, and had money to splash around, I *might* give the R659 consideration. But as it stands it's just a dream for me. I make 175$ a month. Even though I don't have to pay for food or clothing or housing, that's still not enough to just blow on high priced RDRAM or even DDR400. I guess i'll have to take a backseat. Right now my priorities are..

1.Reach over 512MB of RAM. (It involves SiS 655)
2. Get bigger HD.
3. Upgrade CPU.

I do admit though, once Rambus puts out a product it tends to keep it's performance crown for a very long time, in the computer wold. PC800 and PC1066 both held their ground despite the onslaught by DDR.

 

[FishTankX]

There's a posibility, though, that the low latency of the PC1200 will help it beat DDR400 by a significant margin. Just don't know, if it does, by how much.

 

[DX2Player]

Ya but think about how long the 64bit RDRAM will last, I can use the same ram till end of 2005 and still taking advantage of the CPUs fsb while you will yet again have to upgrade to keep up.

The difference I upgrade less often but pay more, you upgrade often and pay less

Think about how much sooner you would have felt it necisary to upgrade if you went the other route 14 months ago

 

[Sunner]

IMO what will make or break RDRAM on the PC side is Intel's support, or lack of support.

If Intel comes out with a brand new RDRAM chipset to replace i850E, it might just have a very bright future.
If Intel sticks to DDR, RDRAM will never find it's way out of the niche it's in now.

I have a hard time seeing which market this new SiS chipset will target.
The power user/enthusiast market is very small, and many people in this market would probably rather buy DDR simply cause of the bad rep Rambus has, no matter if it's well deserved or not.
Also, many enthusiasts look for price/performance, and if the 659 turns out to be expensive, that's not good.

The market that is truely rather insensitive to prices is the workstation market, and does anyone really think Dell/HPaq/IBM will start putting SiS chipsets in their high end workstations?
Sooner or later i850 will either have to get a refresh, no for the sake of performance, but for the sake of features, or it will be replaced by Granite Bay and it's follow ups.
If the latter happens, where does that leave RDRAM in the x86 market?

Servers? Nope.
Workstations? Maybe, but highly doubtful since SiS ain't exactly a well established brand in that part of the market.
Desktops? Nope, they don't require the highest performance, and i845ZXY/DDR will continue to dominate here.
Laptops? Nope.
Enthusiasts? Sure, but how large is this market? And how large a part of this market will go for RDRAM?

Not saying RDRAM is bad, Im just saying their success in the x86 space is at the mercy of Intel, and so far I haven't seen anything to indicate Intel will produce any more RDRAM chipsets for the i786/P7 family.
Maybe this will all change with i886/P8, and DDR will finally be pushed out of the market?

As for financial factors, such as Micron going out of biz, I leave that to the fellas over at Intel, they didn't get where they are by being stupid, I trust them to do what it takes to ensure their CPU's aren't left without the proper memory technology, and I trust them far more than anyone on this BBS, Citibank employee or not(no offence intended Ice9, Im sure you understand what I mean).

 

[FishTankX]

Oh yes..

To Ice9

I'm all for DDR price hikes, really, as long as they're not severe (40%+).

Now, i'm not being sarcastic here. I'm as worried about hynix (Which produces some of the best graphics card memory on the planet) and Micron as you are. But why don't they all raise prices together? It's really their fault that they got into this mess, and I don't think us consumers are gonna get them out of it. What should we do? Ask them to raise prices?

Personally I think it's not nearly as bad as theh time where you could get 512MB crucial DDR2100 for 30$. Micron, in the financial reports you gave me, is posting losses yes. But minor ones, recently. To the tune of 30 million. Not too bad compared to their 650 million income. (Correct me if I read the charts wrong, I'm really bad at financial charts). A few price tweaks in DDR and they should be in the green. I wouldn't mind paying 15-20% more for DDR if it meant the price would be more stable.

One thing I don't understand though. If the i850E is priced around the same as the SiS655 and was priced way below the granite bay, why couldn't samsung just drop the prices a little, so that they could be more competitive with DDR? Then it wouldn't be relegated to all but the ultra high end. PC1066 that is. It certainly wouldn't have adversley affected it's profitability. Is it possible that they were already producing at maximum capacity, so that they couldn't fulfill any more orders?

And as for my statment on RAMBUS being a very litigious company, I was under the impression that they got their SDRAM and DDR patents shadily anyways, and not just that. It also sees that they sued alot of people in a very short amount of time. But i'm uneducated on the details on this issue. Care to explain exactly how RAMBUS got the patents for SDRAM and DDR? I still don't understand that too well. Some people say they stole it from a JEDEC confrence. Others say they got it legit. What's your take on it?