Howdy, Stranger!
It looks like you're new here. If you want to get involved, click one of these buttons!
This forum is in permanent archive mode. Our new active community can be found here.
Some PC questions
I'm an electrical engineer and I know an awful lot about how the innards of a CPU work, but I find there are some pretty awful holes in my knowledge of how thing on the market actually work in practice and what new technologies are coming out. So there are some things I've sort of been wondering.
DDR vs DDR2 Am I wrong in thinking that there is very little practical difference between having a single stick of DDR2 RAM and two sticks of DDR RAM? The single DDR2 stick takes up a bit less space and uses somewhat less power, but should I the consumer consider a motherboard with two DDR2 -800 slots to be better than a motherboard with four DDR-400 slots if I'm not worried about the power consumption of my RAM?
AGP vs PCI-E. There are lots of cards for each that seem to have similar specs, but I'm given to understand that PCI-E allows a faster connection to the CPU or some such. How important is that from a practical perspective, or maybe I should ask how good a graphics card it would take for that connection to become a significant bottleneck? Or is that a matter of the game in question?
SATA vs PATA. Why the heck is a serial connection faster than a parallel one? I mean, if you can do things in parallel, why not do them in parallel? It seems odd to me.
DDR vs DDR2 Am I wrong in thinking that there is very little practical difference between having a single stick of DDR2 RAM and two sticks of DDR RAM? The single DDR2 stick takes up a bit less space and uses somewhat less power, but should I the consumer consider a motherboard with two DDR2 -800 slots to be better than a motherboard with four DDR-400 slots if I'm not worried about the power consumption of my RAM?
AGP vs PCI-E. There are lots of cards for each that seem to have similar specs, but I'm given to understand that PCI-E allows a faster connection to the CPU or some such. How important is that from a practical perspective, or maybe I should ask how good a graphics card it would take for that connection to become a significant bottleneck? Or is that a matter of the game in question?
SATA vs PATA. Why the heck is a serial connection faster than a parallel one? I mean, if you can do things in parallel, why not do them in parallel? It seems odd to me.
Comments
Wikipedia says this about DDR v DDR2
"DDR is being replaced by DDR2 SDRAM, which has some modifications to allow higher clock frequency, but operates on the same principle as DDR. Competing with DDR2 will be Rambus XDR-DRAM. It is expected that DDR2 will become the standard, since QDR (Quad Data Rate) is too complex to implement, while XDR is lacking support. DDR Prefetch buffer width is 2 bits, DDR2 uses 4 bits. Memory manufacturers have stated that it is impractical to mass-produce DDR1 memory with effective clock rates in excess of 400 MHz. DDR2 picks up where DDR-1 leaves off, and is available at clock rates of 400 MHz and higher."
There are also articles on PCI-E and AGP, but in my view, you'll want the PCI-Express 16 because you'll want the fastest processor you can get and most MoBo companies aren't going to support AGP going forward.
Re: SATA v. PATA, again from Wikipedia:
"SATA drops the shared bus of PATA, giving each device a dedicated cable and dedicated bandwidth. While this requires twice the number of host controllers to support the same number of SATA devices, at the time of SATA's introduction this was no longer a significant drawback. Another controller could be added into a controller ASIC at little cost beyond the addition of the extra seven signal lines and printed circuit board (PCB) space for the cable header. Features allowed for by SATA but not by PATA include hot-swapping and native command queueing. To ease their transition to SATA, many manufacturers have produced drives which use controllers largely identical to those on their PATA drives and include a bridge chip on the logic board. Bridged drives have a SATA connector, may include either or both kinds of power connectors, and generally perform identically to native drives. They may, however, lack support for some SATA-specific features. As of 2004, all major hard drive manufacturers produce either bridged or native SATA drives. SATA drives may be plugged into Serial Attached SCSI (SAS) controllers and communicate on the same physical cable as native SAS disks. SAS disks, however, may not be plugged into a SATA controller."
In my view, what's really interesting are the quad core processors coming out by year end. This, to me is a serious innovation. R & S could not disagree more and you have to decide who has more credibility: two computer professionals or a lawyer-hobbyist.
Certainly there are good reasons why industry is switching to DDR2, but is there anything I as a consumer should care about? My impression is that the various RAM sticks aren't all on the same bus, so having more RAM will increase RAM bandwidth. I think that's true, but I'm not entirely sure so I'd like someone to confirm that. And in terms of latency DDR1 chips tend to have much smaller latencies measured in cycles, so I think it evens out.
I'm sure you're quite right that most new motherboards aren't going to support AGP, but should that effect my current motherboard purchasing?
I'm quite aware of all the various advantages of the newest IDE standard over the old one, but my question is why they felt that moving to serial instead of parallel data transmission would help them achieve it.
AGP is essentially dead. The only reason anyone makes AGP cards is because people still have old motherboards with AGP. If you are buying a new computer you pretty much have to have a PCI Express x16 slot and compatible video card. If you choose anything else you are simply denying yourself the possibility of video card upgrades in the future. Also PCI Express x16 is much faster than AGP, even if you don't always use all that speed.
SATA is leaps and bounds ahead of old school ATA/IDE. Not only is 3gbps transfer rates help to widen one of the worst bottlenecks in any desktop PC, especially those running MS Windows. Not only that, but SATA cables are nice easy to handle wires instead of ribbon cables. As far as I'm concerned SATA is the only way, IDE is more dead than AGP.
In conclusion, if you are buying a motherboard today, get all three of those features. For your use at this very moment in time, it might not be a super huge performance enhancement. However, the price of those features is not a significant cost factor. The main money sinks in a PC are the CPU, the amount of RAM, the number of hard drives and the video card itself. Pick a motherboard that can do what you need it to do, then worry more about what you're going to fill it with. Those are the hard choices.
I totally agree about SATA being far better than PATA, but I'm still scratching my head over why they went from parralel to serial. Was it capacitive coupling between the wires they were worried about maybe? I really don't have much of an intuition about RF electronics, give me a wafer of silicon or a good old compensation problem any day.
Parallel communications has a few problems. It requires much more complex cables and circuitry. You can't send a bunch of bits without a bunch of wires. Hence, the ribbon cable vs. the USB cable.
Also, parallel communications buses have synchronization issues. Let's say you have a parallel bus that is 8 bits wide. You have to make sure all 8 of those bits are there before you ask for the next set of 8 bits. The result is that you have to set the clock speed of the parallel communications bus to a much lower rate. If you don't, then the clocks on the different data transfer lines will not be able to stay in sync, and reliability of data transfer will be lost as clock speed increases. With a serial bus you can crank the speed way up. Sending 8 simultaneous bits every second doesn't mean anything if you can send a thousand individual bits in the same amount of time.
Crosstalk, i.e., interference across the different wires can be an issue with parallel buses. However, with modern shielding technology that hasn't been a serious issue for anyone other than the people developing those technologies.
And lastly, serial communications are wicked cheap. Chips that process serial data have fewer pins, simpler circuitry, and are generally much smaller and more efficient. Go look at some old computer with lots of parallel buses. Look at how big some of the chips are, look how wide the ribbon cables are, etc. It's not good.
If you want to see an example of high clock speed serial buses being superior to parallel buses, take a look at your PCI slots on your motherboard. The old PCs had ISA slots. ISA slots are gigantic black slots that take gigantic cards and are very slow. PCI slots have about half as many pins, are half as long and stupid fast compared to ISA. PCI Express is even smaller than PCI and even faster.
Modern computing architecture is all about having lots of really high clock speed serial data streams. If you can find a way to clock a parallel bus as high as a serial one without any other troubles, you'll be quite a wealthy person.
Thank you, thaneofcawdor.