Why are so many people purchasing DDR SDRAM based P4 boards when RDRAM costs the same as DDR SDRAM now?
An Open Question Regarding Socket 478 boards
- Thread starter Daovonnaex
- Start date
Two reasons that I can think of:
1) Many people (like me) already have DDR RAM--either from an earlier motherboard or from when DDR was going for $25 per 256MB--that they can use so it may not be worth the hassle to sell it and buy new RDRAM.
2) Supposedly DDR RAM is generally more overclockable than RDRAM, which clearly comes in handy with these new Northwoods.
1) Many people (like me) already have DDR RAM--either from an earlier motherboard or from when DDR was going for $25 per 256MB--that they can use so it may not be worth the hassle to sell it and buy new RDRAM.
2) Supposedly DDR RAM is generally more overclockable than RDRAM, which clearly comes in handy with these new Northwoods.
<< Two reasons that I can think of:
1) Many people (like me) already have DDR RAM--either from an earlier motherboard or from when DDR was going for $25 per 256MB--that they can use so it may not be worth the hassle to sell it and buy new RDRAM.
2) Supposedly DDR RAM is generally more overclockable than RDRAM, which clearly comes in handy with these new Northwoods. >>
1) Definitely understandable.
2) A myth. DDR SDRAM has more problems when overclocking (stability and latency--latency goes up as frequency increases). RDRAM overclocks very well (no stability issues, and latency goes down), and evidence of this could be me (P4 2000A @ 3.2GHz) or the Japanese (P4 2200A @ 3.8GHz).
<< A myth. DDR SDRAM has more problems when overclocking (stability and latency--latency goes up as frequency increases). RDRAM overclocks very well (no stability issues, and latency goes down), and evidence of this could be me (P4 2000A @ 3.2GHz) or the Japanese (P4 2200A @ 3.8GHz). >>
True, the whole poor RDRAM overclocking is a moot point by now, but I remember last year when it was a more confusing situation. Perception is reality in a lot of ways, and I know for a fact that the stigma of performance questions like that scared some folks away from RDRAM back then, especially when combined with its then-high price. Heck, I was still under that impression myself up until I started researching motherboard reviews just a few weeks ago. Remember how many people there were ripping RDRAM as overly sensitive to FSB changes (*cough*Tom'sHardware*cough*)? LOL It is funny now to see Tom & co. having to eat there words on that a little, though.
BTW, those are some very impressive overclocks you have there.
What sort of cooling are you using?<< A myth. DDR SDRAM has more problems when overclocking (stability and latency--latency goes up as frequency increases). RDRAM overclocks very well (no stability issues, and latency goes down), and evidence of this could be me (P4 2000A @ 3.2GHz) or the Japanese (P4 2200A @ 3.8GHz). >>
"True, the whole poor RDRAM overclocking is a moot point by now, but I remember last year when it was a more confusing situation. Perception is reality in a lot of ways, and I know for a fact that the stigma of performance questions like that scared some folks away from RDRAM back then, especially when combined with its then-high price. Heck, I was still under that impression myself up until I started researching motherboard reviews just a few weeks ago. Remember how many people there were ripping RDRAM as overly sensitive to FSB changes (*cough*Tom'sHardware*cough*)? LOL It is funny now to see Tom & co. having to eat there words on that a little, though.
Tom needs to come back and beat the crap out of his worthless staff.
"BTW, those are some very impressive overclocks you have there.
What sort of cooling are you using?For the P4 2A, evaporative water/methanol cooling, and for the Athlon MP 1900, a 350W cryogenic peltier with evaporative water/methanol cooling.
I asked in another thread, maybe won't have read it yet, but please point to proof that the technology of SDRAM/DDR automatically gets higher latencies with higher frequencies. This is NOT the same as saying that DDR266 modules have to go to CL3 in order to operate at DDR333 speeds, which is simply due to them not being rated for the lower latency at higher speeds. THAT will be changing as DDR333 becomes the norm, and more modules will be rated for CL2.5 or CL2 at 333MHz.
As for why I use DDR: I hate Rambus (the company) and I prefer open standards when possible. Also RDRAM is still more expensive. "Not much more" is still more. There's a 14 dollar difference in 256MB modules for PC2100 DDR and RDRAM, and a 21 dollar difference for 512MB modules. PC2100 gives me everything I need. PC2700 modules are more expensive, but they're also brand new, RDRAM is years old now and is just reaching the low price of DDR. The performance difference with RDRAM in SOME applications is not justifiable to me in terms of cost and preference. Everyone is free to choose what they want, I prefer not to use RDRAM.
As for why I use DDR: I hate Rambus (the company) and I prefer open standards when possible. Also RDRAM is still more expensive. "Not much more" is still more. There's a 14 dollar difference in 256MB modules for PC2100 DDR and RDRAM, and a 21 dollar difference for 512MB modules. PC2100 gives me everything I need. PC2700 modules are more expensive, but they're also brand new, RDRAM is years old now and is just reaching the low price of DDR. The performance difference with RDRAM in SOME applications is not justifiable to me in terms of cost and preference. Everyone is free to choose what they want, I prefer not to use RDRAM.
<< I asked in another thread, maybe won't have read it yet, but please point to proof that the technology of SDRAM/DDR automatically gets higher latencies with higher frequencies. This is NOT the same as saying that DDR266 modules have to go to CL3 in order to operate at DDR333 speeds, which is simply due to them not being rated for the lower latency at higher speeds. THAT will be changing as DDR333 becomes the norm, and more modules will be rated for CL2.5 or CL2 at 333MHz. >>
I answered in another thread, but I'm probably incorrect. The current implications of overclocking DDR SDRAM versus RDRAM still shows me that RDRAM is more desirable.
<< As for why I use DDR: I hate Rambus (the company) and I prefer open standards when possible. >>
Understandable.
<< Also RDRAM is still more expensive. "Not much more" is still more. There's a 14 dollar difference in 256MB modules for PC2100 DDR and RDRAM, and a 21 dollar difference for 512MB modules. >>
I hadn't even noticed this. I recall RDRAM being at about $75 for a 256mb Kingston RIMM a few weeks ago. I suppose that the situation has changed.
<< PC2100 gives me everything I need. PC2700 modules are more expensive, but they're also brand new, RDRAM is years old now and is just reaching the low price of DDR. The performance difference with RDRAM in SOME applications is not justifiable to me in terms of cost and preference. Everyone is free to choose what they want, I prefer not to use RDRAM. >>
Well spoken. I think the main reason that RDRAM is only ahead in SOME applications (though it is a better choice for overclocking) is that the i850, like the i845 and i845D, is a lackluster chipset. If you compare an i850 to an i845 (and, to make it more fair, equip the i850 with PC600 RDRAM), the i850 doesn't have any difficulty pulling ahead. If a company such as VIA or SiS designed a chipset around RDRAM, we might see the performance we should.
Whether RDRAM becomes very high performing or not, there will always be open standard memory designs that are as fast or close enough that it can't be called (like with processors, some apps like one more than the other). RDRAM is essentially to me just an alternative memory, one I'd push people away from due to it being non-open, and because Rambus profits from it without even doing anything anymore (and again, I hate Rambus ). I'm not entirely against royalty-based design, I like FireWire and would use it if I had anything that needed it. But RDRAM isn't so much better that I consider it worth investing in given all the issues.
). I'm not entirely against royalty-based design, I like FireWire and would use it if I had anything that needed it. But RDRAM isn't so much better that I consider it worth investing in given all the issues.I really think the main thing is that there are many "Dragon" type DDR boards avialable, and many ppl want those goodies, while there's only 2 main 850 boards avialable (Asus P4T-E and Abit TH7II-RAID) which have AC'97 sound if any, no NIC, and only ATA/100 RAID if any. And the other thing is that ppl are under the assumption that not just RDRAM memory, but 850 boards do not oc as well as their 845-D counterparts, which is also false (Abit TH7II is the ultimate ocing platform for the P4).
I disagree with Lord Evermore about the price difference. Plus, the thing about RDRAM, is that when running at PC1066+533fsb, then the difference is definately big enough to take RDRAM.
I don't think that refusing to buy RDRAM because it's made by Rambust is accept (just MHO).
I disagree with Lord Evermore about the price difference. Plus, the thing about RDRAM, is that when running at PC1066+533fsb, then the difference is definately big enough to take RDRAM.
I don't think that refusing to buy RDRAM because it's made by Rambust is accept (just MHO).
My opinion is that Rambus is an amoral company that found that their product wasn't doing so well, so they took unethical paths to try and get money from other companies without actually having to do anything for it. That's why I don't want to support them. I do not subscribe to the "buy whatever works best", my computing doesn't exist in a vacuum (but damn would it be runnin hot if it did).
RDRAM at 1066 doesn't even exist yet for purchase, nor do any boards or processors that officially support 533MHz. When they do, you can bet there's going to be a price premium, just as there's a premium for DDR333 memory compared to DDR266. I'm not sure how you can disagree with me about the price, as those prices are ones I pulled from Pricewatch, i.e., fact, not opinion.
RDRAM at 1066 will be considerably higher bandwidth, and quite possibly better latency. That still won't make it worth it to me to buy it.
Oh yeah, RDRAM isn't made by Rambus. Rambus makes nothing, they just collect fees for other people using their technology.
RDRAM at 1066 doesn't even exist yet for purchase, nor do any boards or processors that officially support 533MHz. When they do, you can bet there's going to be a price premium, just as there's a premium for DDR333 memory compared to DDR266. I'm not sure how you can disagree with me about the price, as those prices are ones I pulled from Pricewatch, i.e., fact, not opinion.
RDRAM at 1066 will be considerably higher bandwidth, and quite possibly better latency. That still won't make it worth it to me to buy it.
Oh yeah, RDRAM isn't made by Rambus. Rambus makes nothing, they just collect fees for other people using their technology.
RDRAM is an excellent performing memory solution for the Pentium 4. Check out the Serious Sam benchmarks in this review at GamePc. RDRAM clocked at PC1066 had a performance gain of nearly 40 percent.
Check out the Anandtech Review In the Serious Sam benchmarks the 2.2ghz Northwood was badly beaten by the Athlon XP 1700. Now we know why = Memory Bandwidth. If you were to run the benchmarks again clocked at PC1066 specs, the Pentium 4 would have come out on top.
This is just one example, but it shows the importance of the memory system. RDRAM (the technology) is a good thing. People dislike Rambus (the company) and their licensing scheme.
Check out the Anandtech Review In the Serious Sam benchmarks the 2.2ghz Northwood was badly beaten by the Athlon XP 1700. Now we know why = Memory Bandwidth. If you were to run the benchmarks again clocked at PC1066 specs, the Pentium 4 would have come out on top.
This is just one example, but it shows the importance of the memory system. RDRAM (the technology) is a good thing. People dislike Rambus (the company) and their licensing scheme.
A little something that people tend to forget in regard to both RDRAM memory and the i850 chipset is that it is a dual channel chipset. This is actually what makes the 3.2GB bandwith (RDRAM PC800) possible.
The i845D is uses single channel memory, and the fact that it comes even that close in "real world" performance to a i850 RDRAM rig is quite impressive. The rig I could tentatively compare the i850 to, memorywise, would be an nForce 415/420 driven Athlon rig, since it also uses a dual channel approach.
Apart from that an i850 RDRAM rig is quite a screamer, and there really isn't any reason not to go for an RDRAM rig costwise (at least at the moment). IMHO, the i845D boards with 3 memory sockets give you a bit more flexiblity in your choice of memory, even considering the "1 double bank/2 single banks" situation. In that regard i850 boards compare more with i845D boards with only 2 memory sockets, since you always have to install the modules in identical pairs (I don't know if you can force single channel operation on i850 boards, and why would you want to).
Also, I believe that DDR RAM is more "futureproof" than RDRAM, not least because it is an open standard. And I tend to agree with Lord Evermore in questioning RamBus Inc.'s business ethics. If they had their way we *all* would pay more for our *DDRRAM* modules, simpy because RamBus Inc. tried to press money out of memory manufacturers due to some patents they had filed after spying on the JEDEC committee. How sick is that!
The i845D is uses single channel memory, and the fact that it comes even that close in "real world" performance to a i850 RDRAM rig is quite impressive. The rig I could tentatively compare the i850 to, memorywise, would be an nForce 415/420 driven Athlon rig, since it also uses a dual channel approach.
Apart from that an i850 RDRAM rig is quite a screamer, and there really isn't any reason not to go for an RDRAM rig costwise (at least at the moment). IMHO, the i845D boards with 3 memory sockets give you a bit more flexiblity in your choice of memory, even considering the "1 double bank/2 single banks" situation. In that regard i850 boards compare more with i845D boards with only 2 memory sockets, since you always have to install the modules in identical pairs (I don't know if you can force single channel operation on i850 boards, and why would you want to).
Also, I believe that DDR RAM is more "futureproof" than RDRAM, not least because it is an open standard. And I tend to agree with Lord Evermore in questioning RamBus Inc.'s business ethics. If they had their way we *all* would pay more for our *DDRRAM* modules, simpy because RamBus Inc. tried to press money out of memory manufacturers due to some patents they had filed after spying on the JEDEC committee. How sick is that!
<< A little something that people tend to forget in regard to both RDRAM memory and the i850 chipset is that it is a dual channel chipset. This is actually what makes the 3.2GB bandwith (RDRAM PC800) possible.
The i845D is uses single channel memory, and the fact that it comes even that close in "real world" performance to a i850 RDRAM rig is quite impressive. The rig I could tentatively compare the i850 to, memorywise, would be an nForce 415/420 driven Athlon rig, since it also uses a dual channel approach.[/d]
Have you seen any dual-channel DDR boards that actually showed an improvement over single-channel DDR boards? I though dual-channel had pretty much bombed out with DDR, due to timing issues.
Dual-chanel rdram has been around for 2 years, and is pretty mature. When dual-channel DDR shows a real performance gain, people will take notice; until then, we're arguing about unicorns.
<< Also, I believe that DDR RAM is more "futureproof" than RDRAM, not least because it is an open standard. >>
Would you share how you can say that? RDRAM's roadmap shows that it could be up as much as 9.6GB/ps of bandwidth (32-bit, dual Channel 850, PC1200), meanwhile, DDR, is likely stuck at PC2700 for the rest of it's life.
<< would be an nForce 415/420 driven Athlon rig, since it also uses a dual channel approach. >>
A nForce P4 would most likely rock.
<< A little something that people tend to forget in regard to both RDRAM memory and the i850 chipset is that it is a dual channel chipset. This is actually what makes the 3.2GB bandwith (RDRAM PC800) possible. >>
Well, I am going to note that it is possible to get 3.2GB/ps on PC800 without Dual Channels. Originally, the 850e (850 with 533fsb suport) was going to be Tulloch. What is Tulloch? It was going to be a single channel 850 but by using 32-bit RIMM's, it would give 3.2Gb/ps on PC800 (or 4.2GB for Pc1066). What is a 32-bit RIMM? It is a RIMM that would have DUal 16-Bit channels on the RIMM. So, Tulloch would be much cheaper than 850. But Intel canceled it
so we're stuck with 850.<< Have you seen any dual-channel DDR boards that actually showed an improvement over single-channel DDR boards? I though dual-channel had pretty much bombed out with DDR, due to timing issues. >>
The reason for this is because of the platform that the Only Dual-Channel DDR solution is on. The Athlon's fsb is as it is, already being maxed out by a single DDR channel, and that's why nForce despite Twinbank, is no faster than KT266A. There would definately be a fairly large performance increase with Dual-Channel DDR on the P4 platform because PC2100 on the P4's 400MHz fsb, is hardly saturating it.
The question really isn't would Dual Channel DDR reap a performance benefit, but rather will Dual-Channel DDR be cheaper than Dual-Channel RDRAM. I'd like to think no, but considering nForce boards with all iit's features, are the same price (if not cheaper) than 850 boards, so I dunno. Intel does have a Dual-Channel DDR chipset in the making though (Garnet Bay) slated for release late 2002-early 2003. Only time will tell I suppose.
<< A little something that people tend to forget in regard to both RDRAM memory and the i850 chipset is that it is a dual channel chipset. This is actually what makes the 3.2GB bandwith (RDRAM PC800) possible. >>
It's also 16-bits per channel, for a total of only 32-bits. 32-bit and 64-bit RIMM standards already exist.
<< The i845D is uses single channel memory, and the fact that it comes even that close in "real world" performance to a i850 RDRAM rig is quite impressive. The rig I could tentatively compare the i850 to, memorywise, would be an nForce 415/420 driven Athlon rig, since it also uses a dual channel approach. >>
The real world performance isn't as close is that close, and the nForce is a poor comparison due to the 2.1GB/sec limit on the Athlon front side bus.
<< Apart from that an i850 RDRAM rig is quite a screamer, and there really isn't any reason not to go for an RDRAM rig costwise (at least at the moment). IMHO, the i845D boards with 3 memory sockets give you a bit more flexiblity in your choice of memory, even considering the "1 double bank/2 single banks" situation. In that regard i850 boards compare more with i845D boards with only 2 memory sockets, since you always have to install the modules in identical pairs (I don't know if you can force single channel operation on i850 boards, and why would you want to). >>
The system will not post without at least two identical RIMMs installed, and you need to install continuity RIMMs in all unused slots, which are usually provided with your mainboard.
<< Also, I believe that DDR RAM is more "futureproof" than RDRAM, not least because it is an open standard. And I tend to agree with Lord Evermore in questioning RamBus Inc.'s business ethics. If they had their way we *all* would pay more for our *DDRRAM* modules, simpy because RamBus Inc. tried to press money out of memory manufacturers due to some patents they had filed after spying on the JEDEC committee. How sick is that! >>
RAMBUs is not exactly moral, but they are no more unethical than Microsoft. I do not care for their questionable ethics, but I, unlike Lord Evermore (for the most part), buy solely on the basis of performance/compatibility.
I guess I need to clarify my position a bit. Firstly, even though I do not condone Rambus Inc.'s business practices, it's not what is holding me back buying an RDRAM platform (actually it's that 256MB ECC DDR RAM, that's doing it ). I was merely agreeing with Lord Evermore that Rambus as a company sucks (IMHO).
Secondly, when I referred to DDR RAM as being more futureproof, I was thinking of the current DDR RAM modules out there. Unless I'm wrong all DDR RAM modules currently available are mechanically and electrically compatible. All DDR RAM based boards seem to run fine with DDR200 (though not as fast) or DDR333 (which as of yet isn't JEDEC approved AFAIK). Likewise you can of course put RDRAM600/700/800 on an RDRAM board (though it seems hard getting anything else than RDRAM800, and why would you).
But when considering next-gen RDRAM, I may be way off the mark here, but I don't believe that 16, 32 and 64 bit RDRAM modules are compatible. Which means that even if you'd want to reuse your expensive (I think they still are) RDRAM modules when upgrading, you may not able because they just won't "fit". I don't think the same holds true for DDR RAM, but again I can be wrong (and most probably are).
As for single channel vs. dual channel memory systems, I still believe that dual channel has the most potential and would love to a P4 based dual channel DDR chipset. The nForce chipset was merely mentioned because, like the i850, it is a dual channel chipset, albeit for Athlon CPUs, which of course (currently) are held back by FSB266.
That's it... Hope I haven't alienated all of you now...
). I was merely agreeing with Lord Evermore that Rambus as a company sucks (IMHO).Secondly, when I referred to DDR RAM as being more futureproof, I was thinking of the current DDR RAM modules out there. Unless I'm wrong all DDR RAM modules currently available are mechanically and electrically compatible. All DDR RAM based boards seem to run fine with DDR200 (though not as fast) or DDR333 (which as of yet isn't JEDEC approved AFAIK). Likewise you can of course put RDRAM600/700/800 on an RDRAM board (though it seems hard getting anything else than RDRAM800, and why would you).
But when considering next-gen RDRAM, I may be way off the mark here, but I don't believe that 16, 32 and 64 bit RDRAM modules are compatible. Which means that even if you'd want to reuse your expensive (I think they still are) RDRAM modules when upgrading, you may not able because they just won't "fit". I don't think the same holds true for DDR RAM, but again I can be wrong (and most probably are).
As for single channel vs. dual channel memory systems, I still believe that dual channel has the most potential and would love to a P4 based dual channel DDR chipset. The nForce chipset was merely mentioned because, like the i850, it is a dual channel chipset, albeit for Athlon CPUs, which of course (currently) are held back by FSB266.
That's it... Hope I haven't alienated all of you now...
Folks! You may find this editorial at Tom's Hardware interesting to read. It's called Rambus - Take 2 Is a Rambus Comeback in our Future? http://www.tomshardware.com/column/02q1/020215/index.html
<< Why are so many people purchasing DDR SDRAM based P4 boards when RDRAM costs the same as DDR SDRAM now? >>
PURE SPITE!
<< Skysailor: I haven't taken anything Tom's says seriously for about 3 years. >>
Every once in a while Tom's does very good work, mind you.
Zap
Elite Member
- Oct 13, 1999
- 22,377
- 2
- 81
I'll give my ten cent's worth (inflation, ya know)...
Currently I'm sticking with DDR.
1) even if RAM pricing similar, motherboard pricing is less with DDR
2) many more motherboard options for DDR (VIA, ALI, SiS, Intel)
3) if I decide to switch to AMD I'll need the DDR and thus won't have to buy more RAM
Currently I'm sticking with DDR.
1) even if RAM pricing similar, motherboard pricing is less with DDR
2) many more motherboard options for DDR (VIA, ALI, SiS, Intel)
3) if I decide to switch to AMD I'll need the DDR and thus won't have to buy more RAM
TRENDING THREADS
-
Discussion Zen 5 Speculation (EPYC Turin and Strix Point/Granite Ridge - Ryzen 9000)- Started by DisEnchantment
- Replies: 23K
-
Discussion Intel Meteor, Arrow, Lunar & Panther Lakes Discussion Threads
- Started by Tigerick
- Replies: 18K
-
-
-
Facebook
Twitter