Slashdot Mirror
Intel Readying Dual-Core Desktop Chip
sunisha.shah4eva writes "CoolTechZone is reporting that Intel is planning on introducing a dual-core Dothan chip for desktop computers. According to the article, Intel has plans to turn the performance table around with AMD. From the article: 'Finally, it looks like Intel has learned from its mistake and secretly prepping a surprise for the rest of the industry. According to the information we received, Intel is currently working on a desktop, dual-core Dothan microprocessor with SSE3 instruction set that Intel plans to launch sometime in the future. Whether the launch will take place this year or in 2006 is currently unknown.'"
Communication efficiency and information sharing between the two cores.
On AMD Dual Cores, there is a specific bus for communication between cores and with the memory module, while in Intel types they have to use the main bus.
So intel choice for Netburst dual core lowers the total efficiency (since the cores have to share with the rest of the system, situation akin to regular dual processors) while AMD dual cores have a special bus which is even faster than the regular main bus, lowering latency and increasing communication capacities between the cores, on top of making them compatible with regular mobos.
But one has to remember that the choice Intel made for Netburst's dual core was more than likely done in a hurry, to release DC faster than AMD.
They'll probably design a much more specific processor for their Dothan dual cores.
"The way we can tell it's C# instead of Haskell is because it's nine lines instead of two." -- wadler
The funny thing is that AMD is planning on releasing Quad Core chips in Q1 2006. Intel's DC Dothan may be too little, too late.
Help find a cure for cancer. Join the [H]orde
I still don't understand why Apple users care so much about which processor is in their system.
We're no better or worse than the Intel vs AMD crowd
...and that's all there is to it.
Because intel's cores have to communicate with each other through the FSB, which is slower and also eats into your other bandwidth. AMD's dual cores have a special connector IN the chip that accomplishes this MUCH faster than the FSB. All intel did was take two chip and package them together, they still communicate in the same way that any normal dual chip system does. AMD's don't.
Smithfield is two cores on one die.
The difference is both cores access the system bus directly, there's no on chip core to core communications as there is with AMD's solution. That shouldn't surprise anyone though, SMP by deffinition is done in the same mannor, each chip sharing the system bus. Intel doesn't have the same abstraction between the core and the system AMD has.
Intel has shown plans for two seperate dies on a package (I forget the name, a version of Pressler maybe it was), but that should only help Intel, if one Smithfield core is bad, they throw both away (or more maybe sell them as single core prescotts, but we'll see), independant cores makes it easier to discard only the bad dies.
The power hungry shit processor is the regular Pentium using the Netburst architecture, high clock rates, low efficiency per cycle, fucking radiator, the dothan on the other hand is another story, much closer to AMD's approach: lower clock rates (and upper limit of the architecture) but better efficiency per cycle and MUCH LESS power hungry (while latest PIV crank out above 130W peak out of the box, dothan are rated under 27W)
"The way we can tell it's C# instead of Haskell is because it's nine lines instead of two." -- wadler
You can buy them now: http://www.anandtech.com/news/shownews.aspx?i=2437 3
HJ
Hmm...
"The Consortium is led by founding members Advanced Micro Devices, Alliance Semiconductor, Apple Computer , Broadcom Corporation, Cisco Systems, NVIDIA, PMC-Sierra, Sun Microsystems, and Transmeta." (my emphasis)
IBM wasn't a founding member. Sure, they're a member, but Apple is higher up in the hierarchy than IBM. If Apple wants HyperTransport on an Intel chip, they can get it, because they've got power to license it, AFAICT.
I'll say it again, I LOVE competition. Ever since AMD became a threat to Intel, we've seen outrageous processor wars and benchmarking tribunals. I can buy a P4 3 gig processor for about $150 now.
Most likely, Intel will take that performance throne with their "secret". They have a way of doing that (like HT); but, we'll see something better come from AMD. And so the cycle continues...and we all benefit!
Hey MS Windows users: He's talking to you! Would you rather sit at home pretending that XP's new icon theme is "innovation," or using a system that's trying to compete through real improvements? Do you think that MS would have been forced into handling spyware and security if competition existed? What spectacular Next Great Thing would come out of Redmond if you opposed their monopoly by using Linux/Mac?
Pinto's Rule of Responsible Consumption: No matter how good it is now, a monopolist's product will be better next year if there is competition. Therefore, never buy from a company with > 75% market share, ever.
~p
Yes, you can buy what are called dual core P4s for a third the price of an actual dual-core Opteron, but that's because you get what you are paying for.
Opterons have their own high-speed dedicated bus for core to core communication. Dual core P4s are really two separate P4s on a single chip and use the regular bus for communications (along with memory, i/o, etc.).
The dual core P4 you mentioned is operationally no better than dual P4s (single core).
The Pentium 3 does not equal the Pentium M. I don't know what it'll take to get that into your head, but I'm going to stop trying after this post.
The Pentium M is actually a whole, different chip, just as the Pentium Pro and the Pentium 2 are whole, different chips (Pentium 2 and Pentium 3, on the other hand, have so much similar that it's almost a bad example). The Pentium 2 introduced to P6 (Pentium Pro) MMX. The Pentium 3 introduced to the Pentium 2 SSE(1/2). The Pentium M introduced to the Pentium 3 Micro-op fusion, Netburst-style bussing, and a bunch of other blessings that it inheirited from the P4.
So your statement "Saying the P3 was green lighted before the P4 is a pretty stupid statement", is in itself a stupid statement. The Pentium 4 project (Williamette) actually started in the mid-90's as a replacement for P6, so I'm sure plenty of cross-pollenation happened as P6 grew older, just as it did with the Pentium M.
I didn't say the Pentium 3 was greenlighted before the Pentium 4. I said the Pentium M was probably in development and minimal production long before they decided to greenlight and produce the P4.
"Victory means exit strategy, and it's important for the President to explain to us what the exit strategy is." G.W.Bush
o this day, a Dual 2.5GHz G5 still pounds a 3+GHz P4/AMD whatev into the ground (unless you're gaming).
I think that a dual 2.5 GHZ G5 would have all kinds of problems going against 2x AMD64s at 3+ GHz. Also, you probably need to take a look at some server benchmarks before you make comments like that.
Dissolve... Resolve... Evolve...
HT is marketed as something that gives the P4 an advantage over competing chips, and as such it is indeed a joke. Sure, it saves the P4 from it's own performance woes (remember the Willamette? I'm stuck with one on my desk at work), but it's usefulness ends there.
I encode movies, run GIMPS and offer remote (FreeNX) access to friends nearly 24/7 on my Athlon 3200+ and I have no problems with responsiveness. I think perhapse that the impression that HT is useful comes from the fact the the P4 is so terrible at dealing with pipline stalls.
At work I operate a Windows 2003 terminal server on a P4-2.8 machine with 2GB of memory. At home I operate Debian on a K7-2.2 with 1GB of memory. Both provide remote desktop access, usually 3 or 4 concurrent connections. With HT enabled on the Intel machine, the performance is reasonable. With HT disabled, the Intel machine is strangely lagged, to the point that I get support calls about it from folks trying to work from home. The AMD machine, of course, isn't equipped with HT at all and it runs just fine.
If HT was simply a trick to squeeze 20% more work out of a processor, then the difference would not be so pronounced. Truth is the P4 is terrible with branchy code, and the problem is exacerbated when running many simultaneous threads of branchy code.
HT is an excellent way to minimize the damage done by pipeline stalls, though I think it's given much more attention than it deserves. The problem that it was inteded to solve can be avoided entirely (and more gracefuly) by building a shorter pipeline!
Never eat more than you can lift -- Miss Piggy
>To this day, a Dual 2.5GHz G5 still pounds a 3+GHz P4/AMD whatev into the ground
See here
Top 20 2-way SPEC systems
Top 20 SPECint_rate2000
2 2600 Opteron 40.5 36.1
6 3200 Pentium 4 Xeon 34.3 32.9
10 2200 PowerPC 970 21.5 20.2
Top 20 SPECfp_rate2000
2600 Opteron 45.8 42.3
3600 Pentium 4 Xeon 28.6 28.2
2200 PowerPC 970 20 19.2
Extrapolating linearly results for a 2.5GHz, x86 is still about 1.5x to 1.75x on ints, and 1.4x to 2x on floats. From this I must conclude that you are as the subject says, or that "pounds into the ground" has aquired the slang usage meaning "is pwnd".
For some reason, IBM PPC processors seem to have aquired Jobs' RDF, from the G5 to Cell.
The road to hell is paved with good intentions.