New Pentium Chipsets Launched

MojoDog writes "Today Intel has officially taken the wraps off their new mainstream Pentium D 820 Processor and i945 Express series chipsets. Additionally, they also cranked up the Pentium 4 6XX sequence line-up to include the new Pentium 4 670 at 3.8GHz. The Pentium D 820 is Intel's new dual core CPU clocked at 2.8GHz, which contains two Prescott cores per die but doesn't support HyperThreading like the Pentium Extreme Edition 840. The i945 is their new mainstream PCI Express based chipset, one version of which has Integrated Graphics and both supporting these new dual core CPUs. Additionally, Intel took their Pentium 4 6XX sequence processor, based on the Prescott 2M core, for a speed bump to 3.8Ghz."

Comments

No hzperthreading shame that

[FidelCatsro]

Well considering that security alert recently about the problenms with hyperthreading , and given the fact that the chip is duel core anyway which greatly reduces the need for hyperthreading i don't really see it as too much of a loss and quite possibly its an advantage.

Re:No hzperthreading shame that

[vlad_petric]

The security problem can be easily fixed in a new core revision. The main problem with Intel's hyperthreading is their somewhat simplistic implementation, which greatly limits performance (for instance, when you enable HT, most of your out-of-order structures are physically halved).

From a technical perspective, there's still a great deal of value from doing (proper) HT on top of a 2-core design - namely, you get 4 execution contexts. But that negatively impacts sales (very few would still need SMP, for instance)

naming convention

[y2dt]

a Pentium 4 670 at 3.8GHz, Pentium D 820 at 2.8Ghz, a Pentium Extreme Edition 840 w/o HT, and a Pentium 4 6XX based on the Prescott 2M core???

seriously, how is this naming convention better than the old one?

Dual core versus hyperthreading

[sunderland56]

The Pentium D 820 [...] contains two Prescott cores per die but doesn't support HyperThreading

Huh? Hyperthreading was a constrained, limited ability to run two concurrent streams of execution on one physical chip. Dual core CPUs allow unlimited execution of two streams. "Doesn't support hyperthreading" is listed here as if it was a limitation - but in fact dual core (in the benchmarks I'm running) conmpletely blows away any hyperthreaded chip. This is a far better, far more powerful, solution.

It is nice to see Intel finally catching up with AMD....

Re:Dual core versus hyperthreading

[theantipop]

The issue is that running HT on both cores in a dual core would allow four concurrent threads to be processed. There are many benchmarks of the Pentium XE (dual core, HT) which show the few applications that support multithreading give big increases in performance with Hyperthreaded dual cores.

Re:Dual core versus hyperthreading

[Erwos]

You're missing one thing: the Pentium D EE _does_ have hyper-threading on both cores (looks like a 4 CPU system to your OS). HT and dual core is not an either/or proposition - you can have both, and HT is not going to hinder performace on dual core CPUs.

-Erwos

Re:Dual core versus hyperthreading

[tomstdenis]

"It is nice to see Intel finally catching up with AMD...."

Now if they could only do this on a instructions per watt basis ...

The Athlons take less power per unit of wall clock of time as a P4 and they routine excute a higher instruction count per second. This means not only do you get a task [say compiling] done quicker, but you take less power while doing it.

So you may say "wow that dual core dual HT 3.8Ghz sure is fast" but when you realize it takes 300W of power to run [as opposed to the 40W the new AMD Venice core takes] ... the comparison looks really shameful.

Tom

Re:Dual core versus hyperthreading

[_Pablo]

Better IPC and IPW is exactly what the next generation based on Pentium M is for.

Even in it's current form the Pentium M can exceed the performance of even an Althon FX-55 at the same clock-speed with far lower power draw. When Intel transition this to the desktop as dual core with AMD64 (oops I meant EM64T) and a serious FSB it's going to give AMD a serious run for it's money.

Hopefully AMD is up to the task and we can all look forward to lots of new multithreaded apps and lower electricity bills...unless you have SLI and AGEIA to soak up the juice that is! Good riddance Netburst.

Way to suck...

[http101]

...power. Why is Intel consistently a prime waste of power? (http://www.tomshardware.com/cpu/20050509/cual_cor e_athlon-19.html)

When wattage is spiking that high, I'd rather use the AMD processor solely because of the ever-increasing demand and cost of electricity. So not only are they cost-efficient and energy-efficient, but they're also faster and more durable. In the past 4 years, I've burned up (plugged it in, turned it on) a handful of Intel chips just because they were defective (purchased at various stores) and lost 1 AMD to a direct lightning strike.

Re:Struggling

[PlazMan]

Uh, Intel's stock ticker is INTC, and the last six months have been pretty decent.

http://finance.yahoo.com/q/bc?s=INTC&t=6m

Not upgrading yet.

[JustNiz]

I'm waiting for the 9xx series, because they support VT (Vanderpool) machine virtualization in hardware.

Bye-bye reboots to switch between Windows and Linux.

Re:Not upgrading yet.

[kcb93x]

I'll be looking at AMD's version - memory management done in hardware rather than software. Much better performance.

But you'll be able to buy Athlon 64 X2's first

[Brian Stretch]

according to The Inquirer. They'll do the launch at Computex Taipei next week and be officially buyable on June 7th. Pentium D's (D'oh!) will take a bit longer to reach retail. Something about awaiting approval from the fire marshal, I think. Paper launches are blast furnace CPUs are a bad combination, methinks.

Re:lost again to AMD

[DrSkwid]

how does these INTEL and AMD chips compare to IBM Cell processor

there are all turing machines, therefore equivalent

review site pimpage

[!splut]

While the poster successfully pimps hothardware.com, let us even things out by linking to some other reivews.

Anandtech:
P4 670
PD 820
Tom's Hardware on the PD 840s and such

MIPS per Watt or MFLOPS per Watt

[Colin Smith]

It's the real performance limitation in data centres as we move to smaller, cheaper machines. Raw MHz horsepower is becoming irrelevant for most applications except games and certain forms of data processing.

Power supply and air conditioning are expensive. Transmeta are substantially better than AMD or Intel, which means you can install far more machines at a higher densities than you can with Intel or AMD.

Course, if you want better still then you need to move away from ix86 to ARM, MIPS, PowerPC etc.

Re:AMD best bang for buck, supports Free BIOS

[rpozz]

Your rant about freedom is fundamentally flawed considering that like Intel, AMD is also a member of the Trusted Computing Group.

https://www.trustedcomputinggroup.org/about/member s/

Re:It's the CPU not the chipset.

[i41Overlord]

Anand reviewed the Pentium M on the desktop and found that it couldn't compete with dedicated desktop chips. While it was energy efficient, it just didn't have the power to compete against less-energy efficient chips on the desktop.

In other words, it's great for laptops, but a bit slow for a desktop.

Re:Quad-thread on Home Edition?

[kcb93x]

What home users would *buy* a dual-core HT'd system anyway? If they did, it would come from Best Buy or Dell, in which case the System Builder/OEM would install an appropriate OS.

But, as detailed here:
http://www.microsoft.com/licensing/highlights/mult icore.mspx

Microsoft isn't charging per core, it's per processor, so this would count as "one processor."

Re:The editors are just imitating Intel

[chrish]

Clearly a threading error, they didn't protect the post with a mutex.

No ECC support.

[ChickenFan]

Am I the only person who wants ECC in mainstream desktop chipsets?

I kit out all my new machines with at least 1GB RAM and I want long uptimes on all my Windows, Linux and FreeBSD machines. I really want ECC RAM, but it seems that only Intel's server chipsets support it.

It's built-in to the Athlon64 memory controller, right?

You'd think Intel would be more on the ball.

Of course, finding even an Athlon64 motherboard that actually ENABLED ECC is a challenge.

Hm.

[stonecypher]

So, all "silicon inside" jokes notwithstanding, does that mean that a dual-cpu system would be a Pentium Double D?

Re:It's the CPU not the chipset.

[Noehre]

Unfortunately, the actual data in that article produces a different conclusion. Anand just draws the wrong conclusions. As usual.

The 2.0ghz Pentium M performs within 10% of a Pentium 4 3.6ghz in most tests. That isn't exactly not competing.

Anand then saws that it can't compete because of price and lack of chipsets. Well, no shit. They produce mostly Pentium 4 chips for the desktop, not Pentium Ms. Ever heard of economies of scale? And you can bet that if Intel decides to switch to a Pentium M-based design for its next desktop chip, they will have a much more modern chipset to go with it.