New Intel Chipset

Ace Pace · Post by **Ace Pace** » 2004-06-20 08:31am

Well, we got 3.6GHZ, and some nice new goodies
http://www.tomshardware.com/motherboard ... index.html

Intresting, Intel's kinda forcing everyone to go their way, DDR2 support only on the high side, only 1 IDE connection, and no AGP.

phongn · Post by **phongn** » 2004-06-20 10:14am

DDR has some signalling issues if you don't use registered RAM. DDR2 fixes those problems, apparently. Parallel ATA is on the way out and most people buying new computers will get SATA devices anyways. If you must, there are always adapters.

PCI-X x16 is much better than AGP 8X anyways, but some motherboard manufacturers are putting on an AGP slot running in PCI mode for backwards compatibility.

Ace Pace · Post by **Ace Pace** » 2004-06-20 10:26am

I wasn't saying it was bad, just that Intel is forcing the change.

Okey, since I didn't read the whole articles(Sue me), it seems the new Chipset 925 isn't that much better then the old one with 3.4 GHZ, however I suppose when DDR2 gets faster, stuff might improve.

The Kernel · Post by **The Kernel** » 2004-06-20 02:56pm

Ace Pace wrote:I wasn't saying it was bad, just that Intel is forcing the change.

As they should be. If it wasn't for the market leader forcing new technologies down our throats, we would never be able to settle on new standards and we'd all still be using legacy IBM AT components.

Okey, since I didn't read the whole articles(Sue me), it seems the new Chipset 925 isn't that much better then the old one with 3.4 GHZ, however I suppose when DDR2 gets faster, stuff might improve.

It's all about the new features. Better RAID, more CSA support, PCI-e, 4-port Serial ATA, Socket 775, HD Audio, DDR-II, etc. The fact that the performance hasn't improved is actually a good thing; it means that we are switching to new standards before any of them become serious bottlenecks. Remember, when AGP first came out, it didn't provide much of a benefit either, but it was a necessary transition for later, which is exactly what most of the upgrades built into the 9xx chipsets and ICH6 are.

And its about damn time we moved off the PCI bus. Now if we could just get a serialized memory interface (XDR perhaps?) we'd be rid of the parallel bus for good.

General Zod · Post by **General Zod** » 2004-06-20 04:53pm

okay. 3.6GhZ is a good thing. but how much can it be overclocked and still function?

The Kernel · Post by **The Kernel** » 2004-06-20 04:57pm

Darth_Zod wrote:okay. 3.6GhZ is a good thing. but how much can it be overclocked and still function?

As of right now, overclocking at 90nm is going to be a problem. There is simply too much electron leakage from the transistors and wiring being pushed so close together that any reasonable clock scaling on both the Intel and AMD side is out. Everyone in the industry is seeing this problem right now and it isn't until they get it ironed out (expect at least till the end of the year, if ever) that heat will come down and thus clockspeeds will increase.

But I really don't have my hopes up for any significant improvement. I'm looking ahead to dual core Dothan-based 64-bit chips for the real gains. Either that or maybe we'll finally see some movement on IA-64 for the desktop...not likely, but hey, I like to be optimistic about such things.

Uraniun235 · Post by **Uraniun235** » 2004-06-20 07:28pm

The Kernel wrote:And its about damn time we moved off the PCI bus. Now if we could just get a serialized memory interface (XDR perhaps?) we'd be rid of the parallel bus for good.

What makes serial better than parallel?

Admiral Valdemar · Post by **Admiral Valdemar** » 2004-06-20 07:50pm

You're looking at the limits of silicon chipsets now. I doubt we'll venture past 5GHz using this technology, the chances of short circuiting are too high now. A new material needs to be found and fast, but I'm sure the large chip companies are already on the ball there.

The Kernel · Post by **The Kernel** » 2004-06-20 08:13pm

Uraniun235 wrote: What makes serial better than parallel?

Serial scales much better which means that you can ramp up frequencies to much higher levels. For instance, RDRAM was able to keep pace with DDR with only a 16/32-bit interface versus a 64-bit or even dual channel 128-bit SDRAM interface because RDRAM runs at a much higher clock frequency (and keep in mind, RDRAM had far less R&D going into it). Even more amazing is that the Serial ATA spec can do with 4 data pins what takes Parallel ATA 40-pins.

Rambus' new XDR memory looks far superior to anything DDR has on the horizon. XDR is going to start out at 400Mhz (with 8 data transfers per clock) with plenty of room to grow while DDR-II is currently stunted at 133Mhz (with 4 data transfers per clock) and doesn't look like it will improve much in the near future.

The Kernel · Post by **The Kernel** » 2004-06-20 08:24pm

Admiral Valdemar wrote:You're looking at the limits of silicon chipsets now. I doubt we'll venture past 5GHz using this technology, the chances of short circuiting are too high now. A new material needs to be found and fast, but I'm sure the large chip companies are already on the ball there.

Yeah, they are working on a few possibilities with carbon nanotubes looking to be the most likely candidate. The trouble is, Intel and the other major chip foundries weren't expecting to hit the limits of silicon for a few years yet so carbon nanotubes just aren't ready for primetime yet.

Intel was hoping that strained silicon would extend the life of silicon for several years, but so far both strained silicon and Silicon-on-Insulator (IBM/AMD's solution) haven't had much of an impact on the 90nm leakage issue. There are a few more potential solutions, but it looks like Intel is hedging its bets by deemphasizing clock frequency and concentrating instead on IPC. Multi-cores will also help keep CPU performance strong for years to come.

Ace Pace · Post by **Ace Pace** » 2004-06-21 01:14am

On the Overclocking issue, the presscot is fucking 70+C allready, how can you overclock without damaging it with reguler cooling?

Admiral Valdemar · Post by **Admiral Valdemar** » 2004-06-21 08:08am

Ace Pace wrote:On the Overclocking issue, the presscot is fucking 70+C allready, how can you overclock without damaging it with reguler cooling?

Get some liquid nitrogen, that'll help.

I hear that Fe-RAM may be the next best thing. It has a far lower search time than FLASH or SDRAM and uses tiny amounts of power to boot. It's a work in progress though I believe.

For processors, we're just going to have to live with hyperthreading or multiple CPUs per PC if you want more power. I don't see carbon-nanotube technology based chips coming out within the next few years yet, so we're just going to have to deal with it unless someone has an ace up their sleeve.

It would be nice if we had photonic computers now with no need for such wiring and leakage problems.

The Kernel · Post by **The Kernel** » 2004-06-21 10:57am

Ace Pace wrote:On the Overclocking issue, the presscot is fucking 70+C allready, how can you overclock without damaging it with reguler cooling?

Temps seem to be going down thanks to the extra power pins of the 775 socket, although certainly not enough. Still, a good Koolance system will probably allow you to hit at least 3.7Ghz with a new stepping of the 3.2Ghz Prescott.