Quick question: How low can latencies really get, I understand that you're not going to be pushing even 2-2-2-5 XMS dimms down to 1-1-1-2 in the bios obviously but is it possible that manufacturers will eventually be able to mass produce dimms like that? Or is there some theoritical restriction. I'm doubting this as if there was we'd almost have hit the holy grail of memory latency, but it seems too soon for that. Hope some of you can help me out, thanks.
or you can have latency of 0-0-0-0 by having the ram on the CPU die Ram dimm's will get closer to the CPU socket, as will distance between CPU and chipset shrink because it takes time to send signals even 1mm.
Thanks bud, now -1 would create an interesting slew of fancy space time fabric type words I'll bet with the whole data arriving before it was sent kind of thing, I'll bet Addis could give it a whirl
Yeah, I'm pretty sure you PC would quickly freeze or reboot, there's no way you could push anything on the market that low. You'd almost certainly fry your ram if not more. Especially since it's generic like mine dimm so it's probably got something like 3-3-3-6 timings or something, not happening, sorry
how low do you think i should go? also what difference having extreamly costly 1gb ram is it going to make compared ot my generic ram? surely it wont effect me that much :s
Can you give me a link of the ram with the latencies listed so I can guage it? There are some advantages of have higher quality ram, if you want to over clock it for instance generice ram isn't designed for this and won't yeild good results, high quality ram will have lower voltage, and the latencies will be lower out of the box so you may even be able to drop them a tad more if you lucky. But with generic ram you'll be lucky to drop them to the level of the high quality stuff without causing instability and risking frying your ram. And as a bonus you really don't have you pay that much to get yourself really nice ram, a dual-channel kit of low latency 1gig ram will cost you about maybe 70 or 80 quid if you know where to look, I'd show you a newegg link but I wouldn't want to discourage you as I realize you can't order from them, I'll just say than you can get some 2-2-2-5 dual-channel 1gig kit with 2.6 voltage for about $160 or so dollars, very nice.
A link to the ram you're using or want to buy whatever it was, so I could see the specs? If not what are they at right now? You're right you can lower those thing in the bios but it's not real healthy to do that, sort of like a soft mod, you can get away with it but there's always a risk. The thing is with high quality ram you can push it farther, sort of like if you wanted to overclock a 1Ghz processor and a 2Ghz processor of a newer make and model, the 2Ghz proc would overclock more and still be stable so the lead against the 1Ghz proc would be huge. You can still oc the 1Ghz proc but you simply can't push it as far as the 2Ghz proc as it already has a major lead, do you follow this analogy? I'm just trying to say that high quality ram is already at the level that you want you generic ram to be, the difference is the high quality will run cooler, more stable and you can still tweak it to your liking. It's your choice mate, there's nothing wrong with spenidng a little less money
The CAS (Collumn access strobe) latency, the latency we all try to lower was brought with SDRAM because SDRAM unlike DRAM uses a clock pulse. Like a CPU does a clock pulse is the timing in which the RAM operates. The advantage of chips using a lcock pulse is that the intructions and operations can be pipelined; meaning that while stage 2 of a read and write operation is being done, stage 1 is being executed. this means that the RAM doesn't ahve to wait to finish the current oepration before starting another. RAM is organised into rows and collumns (RAS and CAS), to read or write to a memory cell, the collumn and row must be activated. The time it takes to read from a cell after it has been activated is the CAS latency. This is relative to the clock speed. The CAS latency is the column access time divided by the lenght of the clock cycle. Lets say 40nanosecs/20nanosecs. 40ns/20ns = 2. The higher you overclock your RAM, the lower your clock time so the higher your CAS latency would be. But when we do that, the BIOS doesn't change to suit, and maintains a CAS latency of lets say 2. This stresses the RAM. Unless you know the exact specs of your RAM (i mean exact as in you know the collumn access time and details rarely supplied by the manufacturer) then you won't really know for sure how much you're able to tighten latencies. Of course, if this doesn't stop RAM manufacturers, the limitsof physics on current transistor switching etc will do someday.
errm, if i mounted some fans over the ram and had them blowing cold air towards the ram, would that help me get better performance outa the ram, and enable me to overclock it more?
If you want to overclock your RAM you would be better off buying RAMsinks, better cooling and may get a few extra MHz out.
command rate at 1T bank cycle time set down to 5 The Ras to cas delay is down to 2 annd my bank interleave is at 4B
Wow, I don't have anything that even resembles that in my bios, I don't think I can even look up my cas or ras in there
