Follow along with the video below to see how to install our site as a web app on your home screen.
Note: This feature may not be available in some browsers.
Rambus says that the XDR2 memory interface is open to licence now and expects products based on the technology to ship by 2007
MiXdNuTs said:I will never own a Rambus based product. PS2 or anything else that uses it. I despise that company.
Malogato said:While I agree that rambus, inc can suck the big one.. Exluding yourself from products that use it will ... seriously limit .. your purchases. You would deprive yourself of playstation 3 just because of the shady business practices of one of the companies that provide a small part of the system? I'm sure if you delve into xbox360, you'll find a reason not to buy it, either. IBM can be shady at times.
Makes sense to me. And 8GHz is 8GHz!! I don't care if it has 2x more latency than GDDR3, having ~5x more bandwidth would be like a gush of refreshing spring water to bandwidth hungry video cards.Scali said:Actually, memory latencies don't matter at all for videocards, since they're pipelined processors, with fixed positions where memory can be read or written. The pipeline can be designed to hide any latencies.
GDDR is also a high-latency memory.
In fact, Rambus would be the perfect memory, since it offers the one thing that does matter to videocards: bandwidth.
And despite the negative attitude towards Rambus, it was the platform to beat for a long time, on the P4. Not until we saw dual DDR at high speeds, did DDR finally catch up with the raw performance offered by the Rambus systems, which by then were running on outdated chipsets, and memory that was no longer being actively developed.
Actually, I think that that would just short them outjebo_4jc said:[..]having ~5x more bandwidth would be like a gush of refreshing spring water to bandwidth hungry video cards. [..]
Hannibal said:Adaptive Timing
One of the more controversial aspects of the original RDRAM when it first appeared was the fact that different RIMMs could be at different distances from the CPU. At boot-time, the memory subsystem would go through a special initialization sequence that matched the varied latencies of the channels to the latency of the slowest channel. That way, the CPU would always see the same latency for every bank of DRAM, regardless of its physical distance from the processor socket.
XDR2 takes this general idea and applies it to individual trace lengths. One of the hard parts of manufacturing a memory bus is in getting all of the wires (or "traces") to be the exact same length, so that each wire has the same delay time. XDR2's adaptive timing circuitry gives motherboard makers a bit more margin for error in this regard, because it uses a special initialization sequence to sort out the delays of all the traces and to artificially match those delays to the delay of the longest trace. So adaptive timing can lower manufacturing costs by moving some of the complexity of dealing with mismatched trace lengths onto the memory die, where Moore's Curves make it cheaper.
Transmit Equalization
Mismatched trace lengths aren't the only problem that can drive up the cost of a board. The noise and filter effects associated with high-frequency bus operation inevitably distorts the signal that the bus carries, with the degree of distortion being more or less dependent on how much the board maker is willing to invest in keeping the bus clean and quiet.
XDR2's transmit equalization circuitry "pre-distorts" the memory bus signal in such a way that by the time the signal arrives at its destination the bus's filter effects have brought the signal into its proper shape.
Given the price and Intel's and Rambus' bullish tactics, good riddance.Scali said:And despite the negative attitude towards Rambus, it was the platform to beat for a long time, on the P4. Not until we saw dual DDR at high speeds, did DDR finally catch up with the raw performance offered by the Rambus systems, which by then were running on outdated chipsets, and memory that was no longer being actively developed.
MiXdNuTs said:I don't like consoles in general. Anyways, graphics cards will never use rambus memory because despite high frequencies, RDRAM has insanely high latencies which makes it horrible for use in a graphics card. This was quoted from an ATI interview way back when PC-1066 RDRAM was still around. PS and similar optimized closed systems can better use it though. I make all of my hardware purchases political because I can. Ultimately all companies will have to bow down to my consumer dollar and they know this. Thats why you have seen motherboards like DFI LanParty that are just insane in appearance but meet a demand. There was no demand for RDRAM so it failed. I was a participant in RDRAM's failure by not extending my dollar![]()
jebo_4jc said:Projection: when we're using 8GHz RDRAM in our video cards, AGP definitely won't be able to hang any more!![]()
MiXdNuTs said:I will never own a Rambus based product. PS2 or anything else that uses it. I despise that company.
Short n sweet.Brent_Justice said:I'm all for whatever can improve video card performance.
yep. i actually have terra ignored becuase i find it annoying.slamgoku said:Slamgoku - Anyone annoyed by my Dynamic Signature yet?
Ram running faster than your cpu is just sexy.lithium726 said:yep. i actually have terra ignored becuase i find it annoying.
anyway - im all for it if itll improve graphcs card performance, 8ghz ram is 8ghz ram after all, and sounds pretty sweet to me.
Just turn off signatures somewhere in User CP. Makes for much cleaner reading.lithium726 said:yep. i actually have terra ignored becuase i find it annoying.
its actually just him faking a signature like thisAraanor said:Just turn off signatures somewhere in User CP. Makes for much cleaner reading.
lithium726 said:yep. i actually have terra ignored becuase i find it annoying.
seems like they keep making up for the latency issue with mhzModules Type Voltage VDD,VDDQ Frequency Range Read Latency
128M bit GDDR1 2.5V, 2.5V 183-500 (800)MHz* 3,4,5
256M bit GDDR1 2.5V, 2.5V 183-500 (800)MHz* 3,4,5
256M bit GDDR2 2.5V, 1.8V 400-500MHz 5,6,7
256M bit GDDR3 1.8V, 1.8V 500-800MHz 5,6,7,8,9
MiXdNuTs said:I will never own a Rambus based product. PS2 or anything else that uses it. I despise that company.
Parabellum said:to adopt Rambus even despite performance was not as good as PC2100 and was way too much expensive.
Malogato said:While I agree that rambus, inc can suck the big one.. Exluding yourself from products that use it will ... seriously limit .. your purchases. You would deprive yourself of playstation 3 just because of the shady business practices of one of the companies that provide a small part of the system? I'm sure if you delve into xbox360, you'll find a reason not to buy it, either. IBM can be shady at times.
CompMage said:Ummm can any one tell me why Intel told RAMBUS to take a hike....
Projection: when we're using 8GHz RDRAM in our video cards, AGP definitely won't be able to hang any more!
0mega said:if AGP is dead, so is SLI and crossfire <_<
hahah waht a very funny joke...
poor ATi cant get there product out... while nvidia made a great lauch... boy am i glad i picked nVidia on my current build