The future of Pentium III?

[GFORCE100]

Hello,

I want to see your views on one thing before I ask what the future of Pentium III in your eyes will be. I've been thinking about the Ahtlon and Pentium III, I don't intend to start a I love Intel war but the Ahtlon has 128 + 256 = 384KB of full speed memory. The Pentium III has 32 + 256KB = 288KB. As the Pentium III's class is inclusive we need to take away 32K which leaves us with 224KB. Now my logic is, if the Pentium III had (384 - 288) KB of L2 cache (so both chips have equal full speed memory) wouldn't that cause the P3 to outperform the Ahtlon quite well?
Or lets put it the other way round, to do a fair comparison take away 96KB off the Ahtlon's L1 cache so it has the same amount as the P3 and then run the same benchmarks and what not. Place your views on this logic below:

And now for the future of Pentium III. On the roadmap it says the yet new enhanced Pentium III will be out in Q3 2001. It also seems that 1.13Ghz will be the last of the 0.18micron current P3's. I think Intel will kick that one out (fixed this time) by December at the latest so what's going to happen for the 6-9 months then? Will the 1.13Ghz model be the king of the P3 hill for that long? (simalar to how the P2 450 was for 5 full months).

So as I say, post your thoughts on the above below:

Adam

 

[Dark4ng3l]

the more memory in the cache the longer it is to acces it.

 

[Dark4ng3l]

1.13 ghz wont come out until quarter 2 2001 intel said this and it's in their road map

 

[Henry Kuo]

and also I think the overall datapath, etc will contribute to the performance. I think it is just fair to say that for the same series of CPUs, the more cache the better.

 

[Czar]

Intel will have to fill it with something untill the P4 comes because the P4 will not be a mainstream processor untill 2002.

 

[lowtech1]

...No soup for Intel

 

[Windogg]

There will supposidly be a .13 Coppermine w/ current 256k that will push past 1.13Ghz. A slightly revised P3 core called Tualtin is suppose to have 512kb of on-die L2. Guess Chipzilla needs a backup when P4 + Rambust gets nuked by Mustang + DDR.

While the enhanced Coppermines will be able to use the i815, the Tualtin will require a new chipset.

Windogg

 

[GFORCE100]

Yeah the i830 chipset.

 

[Rigoletto]

"the Pentium III's class is inclusive"
sorry but no book or Ace's hardware taught me what that meant!
Anyway I think that the Athlon has more profound advantages than cache size. What of all those execution units and the DSP? Intel chips look a bit better than they are due to unequal support for SSE over 3DNOW! athlon-version.
AMD is calling the shots right now and the question is with what pricing strategy and change to the celeron will intel react with. I think that they will wait till after christmas to improve celerons to 100MHz FSB, but do little else apart from a little price cutting. I don't think any real changes to PIII will happen until Q3 2001 but I'm just guessing.

 

[VladTrishkin]

There are several other facts that effect x86 CPU performance. We cant say that exclusive cache is better than inclusive or the other way around, but remember that inclusive cache promises a better hit-rate and lower latency.

 

[GFORCE100]

I don't know, maybe it's just wishful thinking but if Intel released a 1.2-1.4Ghz Pentium III then I would be out to get it. Until AMD sorts out the heat their chips spit out and all this fus over VIA drivers that and this etc I will stick to Intel. It's like with the Detonatoe drivers, you can go crazy installing new one's every 3rd day or the week.

 

[VladTrishkin]

Hey bud,



<< and also I think the overall datapath, etc will contribute to the performance. I think it is just fair to say that for the same series of CPUs, the more cache the better. >>



You are on the right topic, more is better in our case, but datapath length is another story. Datapath length is associated with overall latency (cache/chipset/memory-DRAM), but is not a major factor in the performance field. Although, it does contribute to the core-yield.

 

[Dark4ng3l]

<< I don't know, maybe it's just wishful thinking but if Intel released a 1.2-1.4Ghz Pentium III then I would be out to get it. Until AMD sorts out the heat their chips spit out and all this fus over VIA drivers that and this etc I will stick to Intel. It's like with the Detonatoe drivers, you can go crazy installing new one's every 3rd day or the week. >>



The 1 ghz p3 is like 200$ more than a 1.2 ghz t-bird can you imagine what the price on 1.4 ghz p3's would be?. if intel could push 1.4ghz p3's amd would have the t-bird at 1.6+ on the same process anyways.

 

[GFORCE100]

The Pentium III's will drop in price a lot once the Pentium 4 hits the shelves. I think Intel is planning to play the who can go faster game with the P4. It will go up in 200Mhz steps at least and if Intel wants to lose AMD off it's back-end it will jump in 400Mhz steps. I'm really looking forward to seeing a Pentium 4 live upto all this new architecture talk. I have the datasheets and in lots of places Intel says it's 2.8X faster the the P6 core. I still honnestly believe that Intel has been disabling some critical BIOS features in the P4 benchmarks we've seen. I don't believe Intel plans to pull off a big marketing roll-out with the poor figures all of us have been viewing lately.

 

[Dark4ng3l]

2,8 times is impossible the x86 architecture makes this impossible. Go read about that 20 stage pipeline and branch prediction and you will understang why the cpu clocks fast bus is slower in reality. Im pretty sure Intel wil ajust the new .13u p3 so that it performs slower than a p4(similar to how they limit the celeron)

 

[Windogg]

I pretty neutral when it comes to Intel vs. AMD. Nothing against Intel (I have 5 Intel systems at home and no AMD) but Chipzilla has been a real fart in the wind lately. AMD is beating them silly on every move they make. I have yet to see a 1Ghz P3 processor on the street which tells me that there are REAL serious production issues at hand. The jump to .13 may just compound the problems. It will take time to ramp up production as well as get the learning curve down. Tualitin with 512k of L2 may not be easy to make. I'm guessing that many will have part of the L2 disabled and sold as Enhanced Coppermines.After that the big mystery is how well their new chipsets will behave.

 

[GFORCE100]

I'm not sure, if I was to vote I would say Intel has a surprise lined up. In the specs it says at 1.4Ghz the L2 cache speed is 45GB/sec. On average it's 1.8x faster the the L2 cache in the P3. It's going to be an interesting chip, plus overclockers will break the 2Ghz barrier soon.

 

[VladTrishkin]

<< 2,8 times is impossible the x86 architecture makes this impossible >>



x86 is a CPU binary code, not a CPU Architecture Desktop CPU's are simply based on the x86 code, so we call 'em x86 CPU's.



<< Go read about that 20 stage pipeline and branch prediction and you will understang why the cpu clocks fast bus is slower in reality >>



i am lost with you here Deep pipelining will help you reach a higher yield, but it &quot;might&quot; decrease performance. A 20-stage (P4) pipeline has some good advantages (ex: more exe. instructions per/in flight, higher hit-rate, etc)..

 

[VladTrishkin]

Oh, and 2.8GHz is no way a limit, not even close.

 

[GFORCE100]

Who said anything about 2.8GHz? I said 2.8x more throughput compared to the P6 core. Intel says that so I just act as a replicator. One thing though, I wish they would use Slot 1, the black packaging looks so much better than the cheap bare FC-PGA kind. Oh well, I'm not the boss so I have to live with the Socket 423 variety.

 

[VladTrishkin]

ok, so its no-more than 2.8x for 6'th generation core..

 

[SuperFreaky]

6th gen core = PIII = which max out it seems at about 1GHz which * 2.8 = 2.8GHz

hmmm math sounds good to me

 

[DDad]

I'll make a bold prediction- the PIII is at the end of it's line! (as we know it now) Tualatin is being readied for:
A: In case the P4 runs into problems (aka bugs, recalls, chipset issues, etc)
B: Political problems. Please correct me if I'm wrong, but isn't the only fab working on the P4 located about 10 miles from the Isreal-Palestine border? One terrorist incident, and Intel would be in deep kimchee.
C: Future usage- Can anyone say Celeron III /100 fsb?

 

[GFORCE100]

Perhaps if we're lucky we'll see a Pentium III at 1.26GHz. That would be nice,