CORRECTION: Correcting processor system bus speed for 2.93GHz processor code-named Woodcrest.
Intel Corp.’s forthcoming micro-architecture promises to be tailored for both high frequencies and increased performance per clock, which will be the reason why the flagship server microprocessor Woodcrest will come out with up to 2.93GHz clock-speed, not much lower compared to today’s dual-core server chips that are based on the NetBurst architecture.
Intel code-named Woodcrest microprocessor will feature 1066MHz processor system bus and will operate at up to 2.93GHz in the second half of 2006, according to slides presumably from Intel’s roadmap which are signed as Intel Confidential and were published at HKEPC web-site. The new chips will be marketed under the well-known Xeon name, the pictures suggest. There will be server processors with 1333MHz processor system bus from Intel Corp.
The Woodcrest processors will be made using 65nm process technology and will be compatible with LGA771 socket as well as code-named Blackford chipset. It is unknown whether the Woodcrest central processing units (CPUs) will be drop-in compatible with Bensley platform which is slated to arrive shortly.
Earlier Intel Corp. officially confirmed that the next-generation Woodcrest chips will have 1333MHz processor system bus (PSB) and core-logic to support the chip will be able to handle two of such busses. Such technology is called dual independent bus (DIB) and is also to be supported on Intel’s code-named Blackford chipset that will be able to work with two dual-core Dempsey processors using independent PSBs.
Intel’s code-named Woodcrest dual-core processors are slated to be shipped in the second half of 2006, according to Intel’s roadmap.
Given that Intel’s server processors typically utilize technologies that are also used on desktops or were previously used on desktops, it is highly likely that Intel’s forthcoming desktop chips that are to be available in 2006 will support 1333MHz Quad Pumped Bus. With higher speed busses Intel’s server processors are likely to offer much higher performance levels compared to what is available now, which will add pressure on rival Advanced Micro Devices.
Intel has confirmed that processors code-named Merom, Conroe and Woodcrest would not be powered by the NetBurst micro-architecture that powers contemporary Intel Pentium 4 and Intel Xeon chips. Besides new micro-architecture, which is expected to feature shorter pipeline and high performance per clock, the chips will also sport capabilities like virtualization capabilities, LaGrande technology, 64-bit capability in addition to EDB, EIST and iAMT2.
The code-named Merom processor will feature 14-stages pipeline, down from 31 or more stages found in current Intel Pentium (Prescott) designs, 4-issue out-of-order execution engine as well as improved performance of the floating-point unit (FPU). This greatly showcases the substantial difference from the current NetBurst chips that have very deep pipeline and cannot boast with really high-performance FPUs. Furthermore, 14-stages pipeline is deeper compared to AMD Athlon 64’s 12-stages pipeline, which, on the one hand, allows slightly higher clock-speeds compared to the AMD64 architecture, but, on the other hand, may mean a bit lower efficiency.
Comments currently: 13
Discussion started: 02/10/06 01:54:00 PM
Latest comment: 02/15/06 10:11:00 PM
Expand all threads
| Collapse all threads
One of your better written articles, with a lot of good background information to flesh out the current point.
However, additional stages do not mean lower IPC, necessarily, nor do they necessarily allow for higher clock speeds. The perfect example is the Athlon 64 which has additional stages for increased IPC, not for clock speed.
You only reach increased clock speeds when you spend more stages on the same function and remove the bottleneck. So, if I use three cycles instead of two for whatever function, that particular part can run at higher clock speeds, and presuming it was a limiting factor, the processor can.
If I add additional stages that do something that was not being done, or is now being done better, it is not going to add to clock speed. I bring this up because my suspicion is that Intel may have added a stage or more for scheduling or stages that are in some way related to feeding the extra integer pipeline. I could very easily be wrong, but when you figure that AMD doesn't know how to do this, and Intel didn't with the Pentium III and Pentium 4, it could be they need an extra stage or two to help with scheduling instructions so the extra pipeline is actually utilized sometimes. Again, just speculation, but what isn't speculation is that the extra stages in the A64 were for IPC and not extra clock speed. So, it's a bad oversimplification to make that assumption. It might be better to say that generally longer pipelines lower IPC and increase clock speeds, but not in all instances.
One last point, the Pentium III had a longer pipeline than the Athlon, but couldn't hit the same clock speeds. So, as a generality, OK, but anything more than that is disinformation.
02/10/06 01:54:00 PM]
- collapse thread
But it is true. Longer pipeline = higher clocks but lower IPC. Otherwise the P4 would be Intels glorious victory. Which it...uhm...isn't.
Just think about it:
If longer pipeline = higher IPC plus
longer pipeline = higher clock we get
higher clock * higher IPC = higher overall performance.
Basically a win-win situation. So where are the backdraws of it apart from power consumption? Basically your assumption doesnt add up in reality.
And where does the Athlon 64 have a LONGER pipeline? It has 12-stages int and a 17-stages fp-pipeline. In contrast to this Merom/Conroe/Woodcrest have a 14-stage pipeline (int I guess is meant, not sure about fp). This allows somewhat higher clocks at the cost of IPC (slightly though). However, it's a 4-issue design, whereas Athlon 64 is 3-issue (so is Pentium M / Core Solo/Duo and Pentium 4). Thus it has a theoretically higher ILP which leads to higher IPC. However, it depends on the software if the core is ever possible to execute 4 instructions simultaneously. So while in theory it could be a 33% boost in ILP, in practice the gain will be way lower.
Bottom line, its kinda hard to estimate the performance of a new microarchitecture. Lets just wait a few more months to see what Intel has in store for us.
Oh and btw, why did you never consider Athlon 64 / Opteron for your servers? They boast great performance and power efficiency.
02/12/06 03:52:33 PM]
I'll only believe this when you can buy it. Frankly Intel hasn't hit an ontime release on anything they've laid out on any roadmap a year ahead of time, in about 3 years. Merom is highly likely to not come till 2007. Conroe in July is a pipe dream, smokin something alright.
They are good in what they do, but on time they are not. However they thrive on letting folks believe they might actually have a "better" product soon.
Dear Intel. Show us the goods. Like a working model chip? Nope!
(sorry my cynical side came on strong)
02/11/06 03:21:56 PM]
- collapse thread
Were you in a bad mood when you wrote that or somehow frustrated with Intel? It's not I disagree with the overall sentiment, but in some cases it's not true. For example, the move to 65nm and Pressler wasn't behind schedule. It seems the Itanium is not only behind schedule but way behind schedule every time though. Years and years.
I was reading the target numbers were a little lower than 2.93 GHz, but either could be true. I'm so happy that I can get a Sossaman very soon, I'm not too concerned with this release. I'm sure it will be better, but I don't need top performance and Sossaman has a very low power ceiling and should perform great. So, Intel is showing that chip and it is a reality, so why not give them some credit there?
I guess now is a good time to sell my new Pentium III motherboards on eBay. Once Sossaman comes out, I don't think many people will want them. It looks to be the first CPU that is superior in every way to the Pentium III for servers. Makes you wonder what took them so long though.
02/12/06 11:50:28 AM]
Keep the P3.
-At 2.93Ghz i dont think it will be very low power (lower than current P4 offering that will be for sure!).
-Intel will fail the schedule, belive me!
-I have persons that already told me the chip is hot! (Hot because is good? or because it heats too much? I just dont know that).
02/13/06 02:30:45 AM]
The Sossaman 1.67 uses less power per core than the Pentium III-S, and is going to be faster as well.
30 watts for a dual core is terrific. A pair of 1.4 GHz P IIIs is around 70 watts.
You can't beat that performance/watt.
02/13/06 07:09:57 PM]
Yes but that is achieved because of:
-A processor originally designed at 5.0 micron have been shrink over time (optimized)
AMD dual core at 1.6Ghz at 65nm will consume the same and perform the same of the Intel offering and it will be cheaper.
The only difference now is that Intel will catch up with AMD.
Since there are no longer differences between designs (since HT will go away), all will have SSE,..., VT,...
So we will go to a level where ATI and NVIDIA are now:
One performs better on one specific thing other on another.
Lets hope that Intel is late (i think they are late) for AMD grow their market share up to 30%.
02/14/06 06:20:15 AM]
Who cares why the processor performs better than the Pentium III at lower power. It's not that I disagree with you, it's just that I don't see the relevance.
All that stuff about AMD is pure speculation, and as you can see from the current 1.8 GHz Athlon with a single core, probably not even close to being true. But, more to the point, AMD doesn't make such a chip. I'm not really talking about theory here, or what someone could or couldn't do, I'm talking about what is out there. The Athlon is a poorly designed chip compared to the Pentium M line when it comes to efficiency, it never could measure up the P6 in that metric and the Pentium M is a P6 with the emphasis on efficiency.
I've had Athlons, they are miserable chips that get really hot and haven't been too reliable for me either. I don't trust things that get that hot, as heat is what causes electronics to fail. I almost laugh when these idiots overclock when they don't need to so they can sound cool on the internet. It shortens the lifespan by quite a bit, and slowly but surely the processor starts getting flaky and you have to keep backing off the timing. I did this with a K6-III+ to test this out (which is still very useful because you can change the clock speed on the fly with software) and for about six months could run it at 600 MHz. Internet Explorer started get rare errors, that got less rare all the time. I backed off to 550 MHz, was OK for six months and then the same thing happened. I just chucked it after that, rather than run it at 500 MHz.
So, for stuff you leave on 24 hours a day, a lot of modern processors are poorly suited. These ultra lower power processors are not only going to save you a lot of money in power (from AC as well), but also are going to last a lot longer. I suspect these ULVs should last pretty much forever, and with processors advancing so slowly now, a server that lasts for a decade has some nice advantages. If this sounds crazy, keep in mind that the Pentium Pro is still the preferred processor for many Linux servers, provided the application doesn't require too much processing speed. They are extremely reliable, very low power and run forever.
02/14/06 08:14:14 PM]
Ok, you don’t talk about theorically AMD chips, but you like to talk about the Intel ones?
I have (we all have) been pumped with Intel paper launches (future products speculation) all the time every day:
...end of the year we will have...
...next year we will have....
...in 2 years we will see bus speed at 1600Mhz...
...xxxx core will go to 2.93Ghz at Q1 of xxxx....
Why not say that we will have 10 Ghz bus speeds in 10 years, who cares? I don’t!
I have a 2500+ sempron (754), I can’t get power consuming numbers, but if:
3200+ at 2.0Ghz 512Kb cache does 11.4W idle and 41.9W burn
I think my 2500+ at 1.4Ghz 256Kb cache must be around 8W idle 25W burn (this theorically)! And is not even a mobile chip!!!!!
And if that is achieved at .09nm what about .65?
I must assume some 25% to 33% decrease, that will be at Intel knees for sure! Like Intel new dual core, AMD dual core doesn’t consume much more because of the second added core.
Of course this is all suppositions…
02/15/06 03:30:34 AM]
Add your Comment
Enter your username and e-mail address. Password will be sent to you.