AMD vs. Intel, choose your weapon!

[ArticleRobot]

Duron vs. Celeron II, Pentium III vs. Athlon, etc. Confusing to say the least. Afraid that your newly bought CPU will be obsoleted in a few months? Here's a brief guide to the combatants, present and near-term future and what they have to offer, and more importantly, if they fit your bill.

http://www.hardwarecentral.com/hardwarecentral/reviews/1826/

 

[Lord DreadWrath]

A good article IMO. One thing it obviously points out is that it's really not about processor vs processor anymore. Both AMD and Intel processors are close in performance. The real issue is the memory bandwidth bottleneck.

Some people think that eventually we'll have to go to a Rambus platform in the future. But from what I have read, DDR will far better fill this gap for the next 2 years at least. Nothing at all is even using all the bandwidth that Rambus provides.

Another note that can't be overlooked is RDR's latency issues. For RDR to have better latency than SDRAM, you have to push SDRAM to it's bandwidth limit. Something that home users still don't even achieve with PC100 SDRAM. DDR SDRAM also will improve the bandwidth issue thus still keeping SDRAM technology with overall better latency.

On the Willamette side, the reason they are looking to DDR for the high end is because of these latency issues. High end servers also usually have more memory. When you have more than one Rambus chip, the latency gets worse each time you add another RIMM.

I see why Rambus should be considered for the future. But it was pushed by Intel WAY before it's time. There are far better and cheaper options right now for Rambus to even be a good platform for Willamette. 2 years I would say is the time it will take before Rambus might be needed. 2 years is a lot of time though. Time in which a better memory solution may get introduced. In 2 years, Rambus may be better also. Who knows...

 

[Mystichead]

The battle between the Celeron and the duron should be an interesting one.. The new Celerons perform poorly when overclocked compared to their predecessors.
Not only did Intel Cut the cache on the new Celerons but they changed the cache configuration from 8 way associative (found on the P3) to 4 way associative.
With the predicted benchmarks released around the net, the only thing capable of stopping AMD taking the lead is the possibility that maybe the Duron will fail in the overclocking stakes (which I doubt).
The new Celerons have been crippled in too many ways by Intel for them to be considered on par with AMD's upcoming offerings.. looks like another ooops from Intel.

------------------

 

[Toxygene]

This article fails to mention that both Asus and Abit have already stated that their current KX133 motherboards will support the Thunderbird. It seems likely that slotket adapters will be designed for them to support the Duron as well.

 

[Lyngemark]

June 5th AMD will announce 600, 650 and 700Mhz Durons as well as: 700, 750, 800, 850, 900, 950 and 1Ghz Thunderbirds.
Since the core for Duron and TBird are the same I would assume that all those low cost Duron have the capability to be overclocked to the speeds of TBird.
However with a socket we don't have the "magic golden finger" to overclock our CPUs. Hopefully Abit and ASUS will have great overclockable boards that can take these lowcost Durons to their full potential!

Anyone knows the die size of the Timna? My expectation is that it should be very large since it's basically a Cu-mine 128 with an embedded i810e. I think the mfg cost of the Timna should be double the Cu-mine 128, but the pricing from Intel is only US$102 for 600MHz and US$140 for 667MHz. I don't see the economy for Intel to "waste" wafers on this non-profit CPU that probably no one will use. For low-end a Cyrix III or AMD K6 coupled with an all-in-one chipset is still a lower cost than a Timna.
Intel also plan to phase out the Celeron in favor for Timna by Q1'01 according to the latest roadmap.
Fortunately for Intel the Armador will make the Willy affordable, previously this CPU was out of my radarscreen due to the expensive RDR.

 

[Mr. Bandwagon]

FSB, FSB, FSB! give us a 128 bit data bus! Forget Rdram and Sdram, this is the real bottleneck. But it still looks good to me though, fast chips, low prices and many choices. Hope these guys fight it out for a long time to come.

 

[Lithium2k]

Intel have lost a lot of shares and are financialy in trouble claims a source on the web, although thier PIIIe is superb for overclocking and thier prices are starting to match AMD's can they afford to start new projects to rival the AMD Thunderbird in the furure, so if you ask me AMD 1 - Intel 0.

 

[Mystichead]

Its hard to see Intel matching anyones prices when they cant even keep the market supplied with chips.
They claim its because of a worldwide increase in demand for their chips.. they fail to mention little things like quality control and recalled chips like the xeon 933's.
Intels one of those big companies that doesnt care about customers... even though for years they have been living off the backs of them.

 

[DFWPilot]

Agreed! DDR,DDR2 and QDR are going to out perform RAMBUS and guess what -- no licence fees!

Originally posted by Lord DreadWrath:
A good article IMO. One thing it obviously points out is that it's really not about processor vs processor anymore. Both AMD and Intel processors are close in performance. The real issue is the memory bandwidth bottleneck.

Some people think that eventually we'll have to go to a Rambus platform in the future. But from what I have read, DDR will far better fill this gap for the next 2 years at least. Nothing at all is even using all the bandwidth that Rambus provides.

Another note that can't be overlooked is RDR's latency issues. For RDR to have better latency than SDRAM, you have to push SDRAM to it's bandwidth limit. Something that home users still don't even achieve with PC100 SDRAM. DDR SDRAM also will improve the bandwidth issue thus still keeping SDRAM technology with overall better latency.

On the Willamette side, the reason they are looking to DDR for the high end is because of these latency issues. High end servers also usually have more memory. When you have more than one Rambus chip, the latency gets worse each time you add another RIMM.

I see why Rambus should be considered for the future. But it was pushed by Intel WAY before it's time. There are far better and cheaper options right now for Rambus to even be a good platform for Willamette. 2 years I would say is the time it will take before Rambus might be needed. 2 years is a lot of time though. Time in which a better memory solution may get introduced. In 2 years, Rambus may be better also. Who knows...

 

[Juan]

You people must remember that the PIII is made with de p6 architecture, the real competitor for the athlon will be the willamette, at it kick Athlon's ass.
the architecture from the athlon is completely different from the k6 family , but the PIII is derived from the Pentium Pro.
The Willamette has a FSB of 400mhz and it's base speed is 1ghz.