P4

[Argo]

I've got OC3, how much faster do you need? 😉

 

[KarlHungus]

Just to clear things up...

RAMBUS bandwidth (800): 400MHz x 2(double pumped) x 2 byte path = 1.6 GB/s
SDRAM bandwidth (133): 133MHz x 8 byte path = 1.066 GB/s
SDRAM bandwidth (100): 100MHz x 8 byte path = 800 MB/s
DDR SDRAM bandwidth (266): 133MHz x 2(double pumped) x 8 byte path = 2.133 GB/s
DDR SDRAM bandwidth (200): 100MHz x 2(double pumped) x 8 byte path = 1600 MB/s

With all of these RAM types additional channels can be added to double bandwidth.

I really doubt the P4's quad pumped bus will show any benefits in the near future.
Right most programs aren't really pushing the effective bandwidth between CPU and
memory, so this additional bandwidth shouldn't have much effect (if any).

 

[skull]

I wasn't positive what RDRAM was running at I thought it was 800 but they kept saying 3.2G/sec. Is it 3.2G/sec with it in pairs?

<<I really doubt the P4's quad pumped bus will show any benefits in the near future.
Right most programs aren't really pushing the effective bandwidth between CPU and
memory, so this additional bandwidth shouldn't have much effect (if any). I really doubt the P4's quad pumped bus will show any benefits in the near future.>>

With the P4's 20 stage pipe line. Won't faster memory help cache misses. Even though 1.6G/sec is really slow compared to cache. In this case RDRAM's latency is really going to hurt it.

 

[Eli]

The first generation Pentium4 will be a joke, IMHO. Not even slightly worth it, with Mustang and DDR boards right around the corner. You'll have to buy a new case, with special supports, so the 450gram heatsink doesen't break the socket off the motherboard. You'll have to buy a new power supply. Having to buy new 300W power supplies for the Athlon was bad enough. Now you're going to have to buy a power supply that has a brand new spec!

I wonder how long till we start seeing ATX 2.03(or is it 2.04?) compliant cases/supplies? Think they'll be readily available with the P4's realease?

Intel is behind in the MHz wars. Their answer is the Pentium 4. I don't think performance is their main goal; MHz sells.

It will still be very, very interesting to see the benchmarks though.

If it's true that the Pentium 4 has a lower IPC than the P3(And therefor, the Athlon), and If AMD can keep up as far as MHz goes, I think Intel is in deep doggy doo.

Not to mention the motherboards are probably going to be far worse expense wise than even Athlon boards(which actually aren't very expensive anymore, average about 115 - 120$), because the i850 chipset is going to go for around 50$. The current going price for chipsets is ~30$.

Add to that the i850 chipset will supposedly be obsolete in 6 months.

Then add to that RDRAM ... You can see where this is heading.. 🙂

The .13um Copper Pentium 4 is their only hope. They had really better pray that their transition to .13um Copper goes absolutely flawlessly.

Ontop of that, They're not going to be switching all 1320948250 of their fabs over to .13um Copper instantly. They still have fabs at .25, and even .35um. Add to that the Pentium 4's massive die size(It will still be big, even at .13um, especially when we compare it to a .15 or .13um Thunderbird), they had better start prayin now. 😉


Mustang is (very hopefully) going to have a 256bit path to the L2 cache, which should help a little. Maybe a lot. It will have a longer pipeline, to allow for higher clocks, and hopefully more tweaks to enhance performance. This will enable AMD to at least catch up to the P4, assuming the current Thunderbird can't reach 1.4GHz(Which I believe it can, maybe with some tweaking).

If AMD can play their game right, I really don't think the Pentium 4 has much of a chance, in it's first lifetime anyway.


Oh, btw, Intel is absolutely not supporting FireWire. They're behind USB 2.0. If we ever see any P4 boards with FW, It'll be a mericle(merical, mercale? shrug)

Let's not forget that AMD's 8th generation processor (Read: 80886, K8) is due in Q1 - Q2 2001.

Intel appears to be way behind AMD. I wonder if we're going to start hearing about AMD's rabid bunnyman killer, the K9, anytime soon? 😉


Edit: I was close. Just added up all the lowest prices for all the Socket-A boards on Pricewatch. The average is 109$. Not bad.

 

[hans007]

karsin the hut->

1ghz classic athlon 56 watts. , 1.1 ghz tbird 55 watts, i heard the p4 at 1.5 ghz is like 70 watts.

 

[Mr.IncrediblyBored]

hans007, from Anand's article.


<< The 432-pin Pentium 4 should dissipate around 52W of heat when operating at launch speeds; this puts it below that of the 1GHz Thunderbird that is currently available. >>


Aren't the launch speeds supposed to be 1.4Ghz?

 

[Nemesis3]

-Eli and others in the AMD know.

Obviously when mentioning the p4 being out in October the Mustang should be mentioned too depending on when &quot;right around the corner is.&quot; Are we talking about the corner of 1st and 2nd or the corner of 121st and 122nd? I.E. when is it expected? Given it wll be out closer to 2nd generation p4 mobos and should we be comparing it to them instead of first gen?

Is the Mustang going to have:ATA100, USB 2.0, 1394b, 400MHZ FSB.., super double quad-pumped turbo-charged DDR solutions running at what MHZ🙂?

When it comes to AMD the problem has always been software peeps refusing to do 3dnow like they automatically do SSE and SSE2 and so on. As everyone always points out that and a solid mobo/chipset combo is often lacking as well.

Also the comments of the Athlon's fragile nature along with comments like this over at SE &quot;While we have only seen a small sampling of AMD's socket based CPUs, we have seen and heard of far too many Athlon Thunderbird processors and KT133 motherboards dying. Socketed CPUs, both on the Intel and AMD side, are inherently less durable than slot packaged CPUs due to having less protective packaging, something AMD has been learning the hard way.&quot;

>Mustang is (very hopefully) going to have a 256bit path to the L2 cache,
Anyone know?

 

[Nemesis3]

This thread is also over at Sharky Extreme at
http://www.sharkyforums.com/ubb/Forum5/HTML/000047.html

Check out Arcadian's breakdown of Rambus there as it clears a lot of the confusion up, plus brings up quite a few good points.

 

[Sohcan]

From recent interview with AMD President Hector Ruiz: &quot;AMD will introduce five new processors between Nov. 15 and Feb. 15&quot; (EBNews story). &quot;New processors&quot; refers to new cores (Mustang, Palamino, Morgan, Sledgehammer, and ?) not new speeds. Since Mustang is supposed to be the first processor released using the new Mustang core, a release date of November 15th seems very likely.

 

[Rigoletto]

Whered'ya get that info Karsin? I think you're right about the Athlon though, it gets hot!
I successfully overclocked a PII 233 to 350 today. Admittedly I didn't actually try to get into Windows with it but it still runs so cool in comparison. I'm almost tempted to use it instead to &quot;save the earth&quot;
But not quite!
>>evil laugh<<

 

[thorin]

1. Native ATA100 Any performance boost?
? Slight. Look for SerialATA as the next great thing for Hard Drives.
SerialATA Article
Worlds First SerialATA Drive (Check out those phone/cat5 size cables)
2. 400Mhz FSB Surely we should see some benifit here?
? Lots. Check out JCs PC News and Links, and some of the tech articles he links to.
3. USB 2.0 Will it be present?
? Yes it will but we (the consumer) should b pushing for FireWire (and FW2) because it's not CPU based and has higher bandwidth (and potential bandwidth).
4. i850= dual channel RDRAM memory setup... might offer bandwidth and speed but at what price?
? RDRAM Bites, look forward to Q2 or Q3 of next year and Intel's DDR chipset.

Thorin

 

[Nemesis3]

-Thorin are you disputing:

&quot;As I said, the Pentium 4 will use ICH-2. USB 2.0 is a feature of ICH-3, which won't be out until 2001. Sorry.&quot;

I realize 1394b is far superior to USB 2, that is why I mentioned it in my &quot;not included&quot; list🙂

Right now all the original features I have mentioned have &quot;dubious benefits in the eyes of most posters&quot; to the non-server consumer, which would include me as I would be primarily a gamer or &quot;home&quot; user. In the case of USB if it is even present is under debate it seems.

The main thing here is the quick demise of eveything Intel is going to be putting out for sometime to come and the lack of future compatibility in the least. This has always been true to some degree but a lot more time in between and a lot more compatibility with future systems overall.
In this case the 1st generation mobo will not be compatible with the 2nd gen processors, Rambus is suposed to change flavors quickly and then completely expire. Not to mention no one knows the expected lifetime of
even the 2nd generation p4 mobos. So for a guy on a 233 notebook a new machine that would be a &quot;wise&quot; buy appears to be years away🙂

 

[Rigoletto]

Let's FUDge the night away!

 

[OneEng]

I personally have jumped ship and went with AMD. You can place a pretty safe bet that the new AMD processors will be availible on the market before the P4. AMD will likely not even bother introducing the Mustang/Palimino until the P4 can be produced in some kind of volume. Intel has already said that P4 will be availible &quot;only in limited quantities&quot; until next year. This is the same thing they said about the 1Ghz PIII and it took over 4 months for that to ship in volume.

RAMBUS memory does in fact have quite high bandwidth. This is especially true for the dual channel configuration. The thing to think about is the latency (time required to start a stream of memory up) and its impact on P4.

The idea behind most modern processors is to keep instructions and data frequently used in high speed local L1 or L2 cache. This minimises the need to go out to local memory which is orders of magnitude slower.

The stream benchmarks using RAMBUS memory are quite good because it is a benchmark that relies on few accesses to RAM with each access getting a huge stream of information.

Processor performance is more dependent on latency than bandwidth. When the processor has a miss in L1 and L2, it goes out to main memory to get a relatively small chunk of RAM to continue processing.

This last statement along with the very large &quot;Hyper Pipelined&quot; design of the P4 will be the crippling performance issue. The long pipe makes it even more likely that L1 and L2 will miss and need refreshed from main memory.

IMHO, the P4 was designed for marketing reasons, not performance reasons. If AMD can keep up in Mhz, Intel will have a very serious issue on their hands.

Has anyone heard of a copper sample from Intel yet? I have not. AMD had samples of K6 on copper about 1 YEAR prior to the production launch of K7 on this process. This does not bode well for Intel.