Im trying to decide which route to go building a P4 3.06 and Im hung up on RDRAM vs DDR motherboards. I know the new granite bay dual channel boards are out but I dont what to run into any stability issues its still early. I know eventually DDR is the route to go but not sure if Im ready.
RDRAM vs DDR Which is faster?
- Thread starter 1ManArmY
- Start date
Go with DDR, more future proof, and not all that much slower, and it's cheaper.
Stability? Are you thinking RDRAM would be more stable than DDR?
There are plenty of stable DDR and RDRAM boards out there, I would go for DDR because you can get more, use it later and there's not that much of a performance hit.
Stability? Are you thinking RDRAM would be more stable than DDR?
There are plenty of stable DDR and RDRAM boards out there, I would go for DDR because you can get more, use it later and there's not that much of a performance hit.
You have to actually list what kind of applications you're going to be using. If it is an NLE based system, I would be aiming for I850E/RDRAM without a doubt. If it is a gaming based system, i would be up for i845PE or a SiS655/DDR based system. If it is a encoding/streaming based system, RDRAM is the currently the best way to go regurdless of the politics of the company name that the text lashing members like to regurgitate.
Wow. What a pressing argument.
Forget the "4x 2x" arguments here for a moment. Look at total bandwidth.
RDRAM is currently topped out at 1066MHz (or 8x133MHz FSB)
DDR is currently topped out at 400MHz (or 2x200MHz FSB)
However -
RDRAM transfers 16 bits per clock cycle.
DDR transfers 64 bits per clock cycle. In DualDDR configurations, it handles twice that - 128 bits.
1066MHz * 16 bits = 2.133 GB/s of bandwidth
400MHz * 64 bits = 3.2 GB/s of bandwidth
400MHz * 128 bits = 6.4 GB/s of bandwidth
Why do you suppose video cards use DDR as their RAM of choice? The massive bandwidth pipes.
And of course, RDRAM is pricier.
- M4H
ViRGE
Elite Member, Moderator Emeritus
- Oct 9, 1999
- 31,516
- 167
- 106
But dual channel DDR is also 4x per cycle, and not nearly as expensive. However, because of memory inefficiencies between DDR and RDRAM, dualDDR is still slower. But in return, even in the most bandwidth intensive situations, RDRAM tends to barely outperform DDR(and more so, dualDDR), and makes its price very noticable.
Not sure if Anandtech has a 3.06 shootout between dual channel DDR vs. RDRAM, but TomsHardware does.
TomsHardware on Granite Bay vs. RDRAM 1066 using a 3.06HT:
Quake: RDRAM wins.
3dmark/Comanche/Unreal Tournament: RDRAM wins
MP3 Audio Encoding: RDRAM wins
Flask encoding: Single channel DDR wins (?!?)
Mpeg2 encoding: RDRAM wins
Sandra: cpu bench, multimedia, dual DDR wins
Sandra: Memory bench: surprise! RDRAM wins.
PCMark: RDRAM wins. By a pretty good margin.
Office/Internet: RDRAM wins, or TIES *OVERCLOCKED* DDR, Dual channel DDR trails significantly.
WinACE archiving: RDRAM spanks DDR.
Lightwave: RDRAM
Cinema 4d: Dual DDR
3dsMax: RDRAM
SpecViewperf: RDRAM
THG Conclusion: Considering the fact that Intel wants to position the E7205 chipset (Dual DDR266 SDRAM) as successor to the i850E (dual PC1066 RDRAM), the limited performance is disappointing, especially with regard to memory.
DDR's had so much time to catch up.... and it still hasn't. Maybe video cards should start using this stuff instead of the DDR they've *been* using
Not that it matters. Rambus already has far faster technologies available that blow even their own RDRAM out of the water. Time will tell if Intel ever plans on using them. If not, SIS 658/659 is on the burner for this summer.
TomsHardware on Granite Bay vs. RDRAM 1066 using a 3.06HT:
Quake: RDRAM wins.
3dmark/Comanche/Unreal Tournament: RDRAM wins
MP3 Audio Encoding: RDRAM wins
Flask encoding: Single channel DDR wins (?!?)
Mpeg2 encoding: RDRAM wins
Sandra: cpu bench, multimedia, dual DDR wins
Sandra: Memory bench: surprise! RDRAM wins.
PCMark: RDRAM wins. By a pretty good margin.
Office/Internet: RDRAM wins, or TIES *OVERCLOCKED* DDR, Dual channel DDR trails significantly.
WinACE archiving: RDRAM spanks DDR.
Lightwave: RDRAM
Cinema 4d: Dual DDR
3dsMax: RDRAM
SpecViewperf: RDRAM
THG Conclusion: Considering the fact that Intel wants to position the E7205 chipset (Dual DDR266 SDRAM) as successor to the i850E (dual PC1066 RDRAM), the limited performance is disappointing, especially with regard to memory.
DDR's had so much time to catch up.... and it still hasn't. Maybe video cards should start using this stuff instead of the DDR they've *been* using
Not that it matters. Rambus already has far faster technologies available that blow even their own RDRAM out of the water. Time will tell if Intel ever plans on using them. If not, SIS 658/659 is on the burner for this summer.
ViRGE
Elite Member, Moderator Emeritus
- Oct 9, 1999
- 31,516
- 167
- 106
Originally posted by: Ice9
DDR's had so much time to catch up.... and it still hasn't. Maybe video cards should start using this stuff instead of the DDR they've *been* using
The latency is too high for video cards at the moment, as I understand it.
Hey fellas thanks for all of your input, the system will be used for all-purpose applications, digital media, and gaming.
ed21x
Diamond Member
- Oct 12, 2001
- 5,411
- 8
- 81
rdram is definitely not as expensive as the anti-rambus people here like claim it is. The argument is about as valid as saying Intel is much more expensive than AMD, when in truth, the differentiation is around $15 per 256 mb. Heck samsung is selling 256 mb pc800 rdram for around $45 (but of course you have to buy two of them).
- Mar 20, 2000
- 102,389
- 8,547
- 126
i dunno what it is but that samsung rdram is about half the price of any other rdram (i.e. kingston, corsair, whoever)
its probably not 40NS stuff, won't work with 533 FSB systems
now if you go up to 40NS stuff the price for 256 is $85, the same price as twice the amount of 2700 DDR.
DDR is faster because if you use RDRAM it will sue all the components in the computer stating that without RDRAM nothing will work and in the end it will start to leech processing power from just about everything.
That's absolutely never been true. No one has ever been able to prove that RDRAM's "latency" made any bit of difference in the real world.
Rambus' QRSL technology was *made* with Video Cards (and similar devices) in mind. The only reason companies aren't using it is because they're mostly on the fence until Rambus' lawsuits are over. It'll be interesting to see what happens over the next year or so now that Rambus is winning in court.
They already did.
They were first found guilty of "fraud" in their case with infineon in virginia's court. But when it hit the FEDERAL level, it was 100% overturned by the CAFC.
All that's left is a case with the Federal Trade Commission, and they don't have much of a case that they can win once THAT case hits the federal level.
never understood how that could be overturned
I will at least never buy rdram or any other Rambus product if I can find something of the same function because of their business ethics, and I will never advice anyone else to buy Rambus products for the same reasons
Well then, you better fight hard to find some EDO
According to the CAFC's ruling, Rambus' 1990 patent is valid, and SDRAM infringes. DDR has never been in dispute as Rambus technology, not even during the original Fraud ruling (Infineon initially won only against SDRAM and lost on DDR.)
So... If you buy DDR, you're paying 3% to Rambus since it's their technology. Now that the CAFC has given Rambus the right to sue whoever they want, you're going to see some settlements - assuming Team DDR can afford to pay the back royalties.
I'm certainly not saying it's a reason to buy RDRAM since everything is DDR now, but at least Rambus will see its cut. If you're buying a 3.06HT like I did, Rambus is *STILL* the best choice. However, you may be locked into an SIS chipset board for your next upgrade since they'll be the only ones making future RDRAM chipsets for PC's.
That is, unless Intel gets blown out of the water by a competitor's chipset.
hehe, edo sucks, have alot of those sticks in a bag somewhereOriginally posted by: Ice9
Well then, you better fight hard to find some EDO
According to the CAFC's ruling, Rambus' 1990 patent is valid, and SDRAM infringes. DDR has never been in dispute as Rambus technology, not even during the original Fraud ruling (Infineon initially won only against SDRAM and lost on DDR.)
So... If you buy DDR, you're paying 3% to Rambus since it's their technology. Now that the CAFC has given Rambus the right to sue whoever they want, you're going to see some settlements - assuming Team DDR can afford to pay the back royalties.
I'm certainly not saying it's a reason to buy RDRAM since everything is DDR now, but at least Rambus will see its cut. If you're buying a 3.06HT like I did, Rambus is *STILL* the best choice. However, you may be locked into an SIS chipset board for your next upgrade since they'll be the only ones making future RDRAM chipsets for PC's.
That is, unless Intel gets blown out of the water by a competitor's chipset.
But Rambus didnt actually develope SDRAM or DDRAM as far as I know, they came to the JDEC(sp?) meetings and saw developers working on a standard, and they just hey lets patent this thing before it becomes a standard and then we can sue everyone who uses it when it does become a standard.
like so many on this forum hate Rambus for that and I'm one of them
Originally posted by: Czar
hehe, edo sucks, have alot of those sticks in a bag somewhere
But Rambus didnt actually develope SDRAM or DDRAM as far as I know, they came to the JDEC(sp?) meetings and saw developers working on a standard, and they just hey lets patent this thing before it becomes a standard and then we can sue everyone who uses it when it does become a standard.
like so many on this forum hate Rambus for that and I'm one of them
Actually, Rambus *DID* develop SDRAM and DDR, back in 1990. They used it in their RDRAM technology. This is the common misconception about Rambus. Everyone thinks that they went to JEDEC and stole the technology. What was shown (and upheld) in court is that several JEDEC companies (infineon being one) KNEW that Rambus had this technology patented, and then INVITED them to JEDEC. This is documented during the trial's opening statements.
See, the way patent law works is, you develop an invention and you patent it. Then you have a *separate* portion for "patent claims". Here's how I described it in another thread:
Let's say you invented "Tums". You file a patent stating the combination of ingredients as the technical patent info, with a "patent claim" that states it's "For the temporary relief of acid indigestion". Now, during some kind of seminar, you find out that your invention is also being used as a calcium supplement to help prevent Osteoporosis. You can now "Amend" your "Patent Claims" to include "for use as a calcium supplement for the prevention of Osteoporosis". YOU CANNOT CHANGE THE ORIGINAL PATENT DESCRIPTION THAT GOVERN THE FORMULA USED. YOU CAN ONLY ADD NEW CLAIMS. In order to do this, you MUST file for a new independent patent.
It's these PATENT CLAIMS that are in question. Not who developed the technology. There's no question that Rambus had this patent in 1990.
I've been following Rambus as a company for years now, since Intel decided to use them in the late P3 days. There always seems to be 2 types of people when it comes to Rambus. People who believe everything they read in the media and consider them property stealing frauds (which isn't even possible under US patent law) or people who think they're being smeared and trying to be put out of existence so JEDEC companies don't have to pay royalties.
There's a lot of interesting "facts" about Rambus that never make the media, because the media LOVES a great story, especially when words like "Fraud", "Antitrust", "Monopoly" and "Shredding" are used. But it's rare that anyone ever attempts to read the OTHER side of the story, which in this case is where most of the actual truth is. Unfortunately, the truth is RARELY as exciting as the accusations being thrown around.
Eh, some people refuse to believe anything else than what's in the media. The media machine is a powerful one. And Rambus is too small of a company to really defend itself against it.
Either way, it no longer matters since everything is now solidly in Rambus' favor when it comes to the foundation of LAW, even with the FTC case. Argue about it all you want, it really changes nothing.
Ice9,
thanks for the info, first time I have read this info
as for fkloster *inserthandgivingfinger*
thanks for the info, first time I have read this info
as for fkloster *inserthandgivingfinger*
TRENDING THREADS
-
Discussion Zen 5 Speculation (EPYC Turin and Strix Point/Granite Ridge - Ryzen 9000)- Started by DisEnchantment
- Replies: 24K
-
Discussion Intel Meteor, Arrow, Lunar & Panther Lakes Discussion Threads
- Started by Tigerick
- Replies: 19K
-
-
-
Facebook
Twitter