64 Bit Pentium 4... any future for gaming?

[Mini D]

Skuzzy, you are thinking in "all else being equal" mode. That is seldomely the case. All else being equal, we could make a 60MHz P1 that wouldn't even require a heat sink at all with the current process. That is saying that 600k transistors aren't adding anything that needs to be worried about.

Now, you try ad jam 40million transistors into 1/4 of that area and tell me one more time how line size is what's making those things so hot.

Density and thermal transfer are the main issues. As the cross section are of a line goes down, the lenght does too. The overall resistance doesn't change much. The thing that changes is that where there were previously 10 lines there are now 15.

Also consider that the initial release of a chip is usually on the order of 15x12mm. Subsequent releases get reduced to about 12x10mm with at least 30% more transistors.

Very little about the Low-K process will reduce the resistance of the lines, it only allows you to put them closer together without reducing the width of the line. Anything you gain will be lost by increasing density once again. All else being equal, it would help. But all else is never equal and these things seldomely are the savior they are sold as.

The only solution to the current thermal "problems" is a major design change. Personally, I don't see them ever being solved. It is actually easier to figure out new and improved ways to cool these things than it is to figure out how to keep them from getting hot.

[Skuzzy]

I agree with most everything you said (pretty much fact so no argument from me).  Bottomline is Prescott is a bad design, compared to what it could be.  Performance is poor, in comparison to Northwood, for example.

The thermal control of the part is an issue as it applies to the surrounding parts.  Sure, the CPU itself can run insanely hot (comparatively speaking), but when you put a motherboard on a bench, with an HSF mounted on the CPU, and the neoprene rubber block on the backside of the motherboard gets welded to the motherboard from the heat of the CPU, that is a problem.

Epoxy can withstand a great deal of heat, but the surrounding capacitors and resistors will suffer premature failure from heat exposure as it stands right now.

It's just not a good part.  Even if the thermal issues were solved, the performance is just not there.  Of course, some of the handicap has to be put off on the use of DDR2 memory.  Makes for good marketing as the numbers are higher, but the latencies kill any performance increase you could have.
Then there is the heat problem with DDR2 memory.

All this heat translates from power usage.  The amount of power required for a Prescott based system is nuts, quite frankly.

I am just disappointed MD.  Intel had a good base design in Northwood and it was faster than the AMD counterpart as well, near the top end.  Prescott and the use of DDR2 ram is a giant step backwards in many areas.
I have heard rumblings about Intel working on a new part (code name starts with a 'C' if I recall) which should alleviate the thermal issues.  I anxiously await it.

[SkyWolf]

Originally posted by Wolf14
:

Did Northwood pretty much quit making cpu's and if so why?

also

If Prescotts are having problems with heat, why are they like the mainstream chip now? is it cost?

 

I buy about 60 -70  800MHz FSB P4 Processors a month. I'm always looking for Northwood Cores. Hard to Find now. We've tested the Prescott Core Processors and they do indeed run too hot for our application (enclosed in a housing and frame). I'm not sure what the deal with the Prescotts is....but we are actually re-engineering to use them because the Northwoods are goners.  

Woof

[Mini D]

Originally posted by Skuzzy
I am just disappointed MD.  Intel had a good base design in Northwood and it was faster than the AMD counterpart as well, near the top end.  Prescott and the use of DDR2 ram is a giant step backwards in many areas.
I have heard rumblings about Intel working on a new part (code name starts with a 'C' if I recall) which should alleviate the thermal issues.  I anxiously await it.

Hehehe... I thought about this thread yesterday when we were discussing a new ILD that we developed that actually looks like it might work. If things go as planned, you'll read about it in 6 months or so.

I actually mentioned that someone was thinking this would cause the chips to be cooler and everyone started laughing .

I totally agree on the design, but I doubt it's for the same reasons. Right now Intel is able to manufacture a processor that is running 50% faster than it's counterpart with similar heat emission. The truly sad part about it is the processor running at 66% the speed of the pentium IV is performing just as well. That can only point at design. No ammount of process changes in the world are going to solve that problem.

I haven't heard of a new chip except maybe the "nahalem". It's important to remember, though, that chips names are design based, not process based.

I will say, though, that some things are start to look promissing that will revolutionize the concept of "dual core". Now, if they'd just do something with the "core" itself, i'd be happy.

[Mini D]

c = "cedar mill" (I believe). I have no idea what it is.

[Skuzzy]

No offense taken.  Being outside the design process only allows outsiders to guess and speculate.

Only two reasons they would laugh about it.  1)  If they had alreadydoen the research and found it would not help, or 2) if they choose to be arrogant and agreesively reject ideas which did not start within thier clique.

No offense to you in that.  I have worked many years in, around, and on chip design and found many, if not most, chip designers to have a pretty arrogant attitude towards outsiders.

I do agree whole-heartedly the design is the issue with the current generation.  I just wonder how they went from a decent design to the train wreck which is Prescott.  Different teams?  No internal communication?  Prescott team had no access to the Northwood design (one of those brilliant, clean-sheet designs).
It's none of my business, but curiosity is a driving thing.

[Mini D]

It's not a team thing skuzzy. We know why there is a need for Low-K films quite clearly. Afterall, we are the low-k group. Reducing power consumption was never a target of the group. Being able to place lines closer together was. Capacitance is an issue because of the power necessary to run the chip. Reducing the capacitance would not allow you to run the chip at a reduced power, it simply allows you to bump up the power with less impact to the line next door.

Now... reducing the power needed at a transistor is an effective control for thermal issues. The problem is, this doesn't occur inversely to the proximity of the transistors. We end up dropping transistor power requirements by 15% while increasing the density 35%.

Also, I'm pretty sure the prescott and nothwood people had a little more exposure to each other than you've been led to believe. Even for a large company, we communicate just a bit better than that. Especially when it comes to things that are going to be manufactured in our virtual factory (all sites simultaneously). I do work with the design groups occassionaly on our test vehicles. It really is a small circle once you hit the management in these teams. Everyone knows everyone and they all regularly talk/meet/plan.

[StarOfAfrica2]

So what is the new Pentium D I keep hearing people talk about?  Is it an improvement?  I've always pretty much stayed with AMD, but I try to stay open to new things.

[Skuzzy]

Oh, I was not led to believe anything MD.  I was just speculating.  It's the outside-looking-in thing.  Trying to make the best guess as to how Prescott got a bit fubared (from outward appearances).

I understand the issues involved with thermal control as density increases.  It is an issue.  But I will maintain there are those who have successfully done 90nm with higher transistor counts (ATI GPU's for example) than Prescott which run significantly cooler.  I am not sure of the exact densities ATI is using on thier current parts though.  The die sizes do not appear to be that large.

I attribute some of this to possibly better layout of the part.  What would you attribute it to (I am asking out of curiousity and you probaly have a better idea than I do)?

[Kev367th]

Originally posted by Mini D
Hehehe... I thought about this thread yesterday when we were discussing a new ILD that we developed that actually looks like it might work. If things go as planned, you'll read about it in 6 months or so.

I actually mentioned that someone was thinking this would cause the chips to be cooler and everyone started laughing .

I totally agree on the design, but I doubt it's for the same reasons. Right now Intel is able to manufacture a processor that is running 50% faster than it's counterpart with similar heat emission. The truly sad part about it is the processor running at 66% the speed of the pentium IV is performing just as well. That can only point at design. No ammount of process changes in the world are going to solve that problem.

I haven't heard of a new chip except maybe the "nahalem". It's important to remember, though, that chips names are design based, not process based.

I will say, though, that some things are start to look promissing that will revolutionize the concept of "dual core". Now, if they'd just do something with the "core" itself, i'd be happy.

Please don't confuse Intels psudeo 'dual core' with AMDs true 'dual core' design.
Plenty of stuff on the web explaining why Intels offering isn't true dual core, far to techy for me to understand. Something about the way the two 'cores' are linked or communicate.

[StarOfAfrica2]

What I understand of it is this Kev.  The Pentium "dual core" is basically two physical CPU's glued together that still share the same voltage and run in the same power state, and have to communicate across an external FSB.  Any core to core communication is slower because of this.  The AMD on-chip Northbridge set speeds things up alot.  I'm re-reading an excellent writeup on it from last month that answers my own question from up above as well.  

http://anandtech.com/printarticle.aspx?i=2397

Talks about the current 64s, the upcoming stuff, and compares it to the Intel offering.

[Schutt]

I thought Prescotts problem were the leaking currents, not resistence.

Anyway the design had a lot of stuff in it which is new technology and intel wanted to performance test, i think they hoped that it would be easier with the heat and they could bump up the clock to 4+ GHz. Now that they found out they cant they will wait for the next design to do that.

Intels dual core is just as much dual core as AMD. Both are just a cheap start in the multi core future, slapping two single core chips together. That is good for programms that run multithreaded in a way which is thought of today.

Real future technology is something like the cell processor, where you have 10 instead of one processing cores on one chip. But to use it efficiently the programming has to be diffrent and the software producer already mourn that its hard to have 2 processors work at the same time.

[Mini D]

Originally posted by Kev367th
Please don't confuse Intels psudeo 'dual core' with AMDs true 'dual core' design.
Plenty of stuff on the web explaining why Intels offering isn't true dual core, far to techy for me to understand. Something about the way the two 'cores' are linked or communicate.

LOL!

Dude... neither have anyhing to do with what I'm talking about. Like I said, it will be very interesting and I doubt very much that anyone in the microprocessor industry will be doing it. We still have alot of bugs to work out before the announcement, but we've gotten the chips to yield. In about another month the next batch wich fixes some lithography aligment issues is due out. It might yield well over 650 ISO. If it does, I think there will be some serious internal discussions about an announcement. The problem is, it's only going to double the density of transistors yet again.

Schutt, more leakage occurs in the transistors than anywhere. The irony is that this is the job of the "High-K" group to fix.

[OOZ662]

Did Northwood pretty much quit making cpu's and if so why?

Wolf, Northwood is the code name for the previous generation of Pentium 4 CPU. Probably the best CPU Intel ever made. Definately better than anything they currently make for the consumer.

This thread is waay to long to read and see if someone else said this already.

I just bought one of the 3.0 CPUs and had to get Northwood because my motherboard was Prescott-unfriendly. I'm not exactly sure how new it is, but it's not an old one.

Northwood P4 3.0 GHz

[Mini D]

I missed this exchange:

Originally posted by Roscoroo
talk about a screw up .. shoulda bought intel stock when it was @ .42 the other day .. today its at 23.08 coulda made 4500. off of a 200. investment ...

Huh?

Originally posted by Skuzzy
There ya go Rosco. Was Intel that low?

Not sure what that has to do with the discussion though.

Are you attempting to ellude to they are doing well in spite of themselves? Just proves how many suckers are available in the marketplace. Most people have no clue and Intel is probably grateful for that.

Huh?

Intel Stock has not been anywhere near $0.42 ever. It's been bouncing around between $22-$24 for a while now.

The one thing that is abundantly clear is that the value of Intel stock has very little to do with the product that is available. The two are totally unrelated. Market trends have a much more significant impact.