Dual Core Notebooks - Page 2

acer?

Yes, DC Turions are clearly on the roadmap now, in March 06, and the code name is Taylor. And months of speculation has ended, Socket S1 will be introduced on the new Taylor platform.

Yonah, on the other hand, will perform reasonably well, but really is a holdover, until Intel can produce enough Meroms for a Q4 launch.

Yonah will still be very weak on certain fronts, such as having only one FP unit to go against three FP's per core on Taylor. But Merom should be 4 issue wide (can sustainably execute four instructions per cycle), and will have 4 MB shared L2, and full 64 bit memory address, and should be a pretty substantial improvement over any current PM stock, including Yonah.

Yonah itself, actually will likely be pretty underwhelming in terms of performance, and may not live up to some of the top end Dothans in terms of single thread applications.

Ahhh - OK, then I apologize for misunderstanding your post.

But you might actually be wrong - there may be optimizations for multi-core processors, due their somewhat different nature in memory and cache access.

As for multi-threaded apps - the average user has more than 20 processes listed in Task Manager. Even if they were all single-threaded, spreading them out over the available cores will make the system more responsive simply due to available time-slices on the core/processor that runs the shell or application.

While that may mean individual applications don't benefit, it would instead mean that system speed and responsiveness improves. For most users, that is actually more noticeable than single application speed gains (well, OK, for the average user, not you uber-geeks ).

you mean this? http://www.sagernotebook.com/pages/go_np9750.html

funny cause the sager will ship a whole month earlier too, and will costs hundreds if not thousands less

Dual Core:
2.0 Ghz = $600
1.73ghz = $400
1.6ghz = about $350
1.5ghz = about $330

Best bang for the buck is the 1.73ghz.

I forgot where i read this, but the info is out there some where..

Dual Core will be sorta affordable in my opinion, but then again, not many programs will support it.. ohwell, LOL, i guess its sorta future proofed or something.

Thanks,

MysticGolem

btw, new intel cores are being pushed back into production because of instability issues, so dont expect to hear anything from them any time in the near future in that regard. dual core fx 58 > all atm.

mystic, that is totally off for yonah. there wont even be a 1.73ghz version. if i do decide to go with intel over amd, i will certainly get the dualcore 1.66ghz one. it is only about 30 dollars mroe expensive than the single core version(240 dollars is the price, not 350), and when i throw it away and upgrade to merom, it wont be as big a loss as the higherclocked versions.

Where in the world did you get that information?

Yonah is not a hold over, nor will it even come close to underwhelming in terms of performance. May not live up to some of the top end Dothan’s? You are horribly mistaken, but you can see for your self in January.

Yonah was planned before Merom was even put on paper, and has been in development for a few years. Yonah is going to be faster at every angle compared to Dothan and has architectural improvements in every area. One of the many improves is the SSE instruction set that has been completely overhauled, while bringing SSE3 into the mix along with VT. There is even lower latency cache, 667MHz FSB, Dual Core, communication through the L2 Cache, DDR2-667 supported, the list goes on.

Everyone will be pleasantly surprised come early January 2006.


That was for Presler, and just the lower end Preslers. The Pentium D 9x0 series is still on track for a January 1st launch. Pretty much the stability issues boiled down to the motherboard, because the 3.2GHz, and 3.4GHz Prelser did not have these problems.

in single threaded performance, the 2.26ghz dothans will win simply because they have a mhz advantage.

the only reasons im waiting for yonah is because i want the new video cards and i want to be able to upgrade to merom.

Clock for clock Yonah is a decent amount faster then Dothan. That is why it can be released at a lower GHz, and still be faster then the 2.26GHz Dothan. Sure it’s also to keep the power envelop in check, but everyone needs to realize Yonah is much more then two Dothans slapped together with direct L2 cache communication.

what about idle battery life? will a yonah running at 1ghz give the same battery as dothan running at 800mhz?

OK, lets take your points one by one:

What does that have anything to do with performance?? And more importantly, nearly all earlier architectures come to market in the order that they are planned, designed, taped out, tested, sampled by OEMs, and shipped to OEMs. So what does that have anything to do with anything???

Do you see recent Prescotts with SSE3 having vast improvements over the earlier Northwoods before it without? Do you see the E steppings of Opteron/A64 having significantly better performance than C/D stepping before it at the same clock/cache/mem bandwidth?? What does new set of SIMD instruction bring? -- a limited set of applications that use large repetitive data structures, and those that are specificly designed to take advantage of them; certain encoding, graphic based, and a few others; but do you really see SSE3 give leaps and bounds of improvments over the previous Dothan revisions for the vast majority of the applications on the market???

Also, do you really see a lot of notebook owners wanting to simultaneously run 2 or 3 operating systems at the same time under virtualization, as Xen linux has done for quite a while now? Do you really see many laptop owners will be rushing to upgrade their machines simply because of Vanderpool or Pacifica. I think to include those type of technologies in consumer grade machines is little more than posturing and a waste of resources and something that will cost most owners, which they will never use in the life time of their machines.

Cache latency has never been one of the bottlenecks of Dothan or Banias performance; in fact, it is the relatively large amount of L2 that has helped PMs to keep pace with some of the features of A64 and to a lesser extent, P4, who both generally have larger memory bandwidth and in the case A64, much lower latency.

PM's much bigger problem is memory latency, where they done very little to close the ground at all in the last two years with the K8. Until PM starts to use on-die memory controller, that problem isn't going to magically diappear. While improving cache latency is admirable, and increasing the amount of cache always helps (which will be 4MB shared in Conroe/merom), they are only temporary patchwork solutions. Sooner or later they will have to move on that front. i.e. to bring out a real replacement for the Whitefield.

Again, that is already way behind time, A64 have 6.4 GB/s memory bandwidth for dual channel, and 3.2 for single, while at the same time having separate high speed I/O in HTT (800-1000MHz, and will be revised upwards next year) that is already much faster than any PM's FSB, which is shared by both memory and I/O.

The only thing on the horizon that will ease this on Intel's part is DIB, which will double the FSB transfer rate, but PM will have a chance in hell to see that implemented in a notebook in the next 6 months. You are talking about a technology that the Intel is already 2+ years behind the competition, and has no signs of catching up yet. Unless they eventually integrate CSI into something other than Itanium platforms, I don't see how they will.

Plus, one of the chief reasons for intel to crank down Banias original FSB is for power saving features, now that intel is forced to gradually raise the FSB, how do you think that will really affect the power consumption in NAPA??

How is DDR2-667 dual channel going to matter, if the FSB is still lagging in the year 2003. Plus, DDR2 inherently has much higher latency than DDR1 running at the same clock, only when we reach DDR2-800 would we see significant improvements on the overall performance of DDR2 RAM over DDR400; keep in mind, we have to look at both bandwidth and latency, there are two parts to this equation here.

Communication through the L2 is something that Intel should have done beginning with the Smithfield, and the lack there of (or other technologies such as communication through on die SRQ like the X2 A64s/Opterons) has really played out really badly in terms of Smithfield and Paxville's memory and IO bandwidth. Communication through the on die L2 is just the first step in alleviating the strained FSB. And if you are a regular reader of the board, and actually has seen some of my previous posts, you would know that I have been discussing the nuiances of these issues for a long time. And of course I recognize that shared L2 is a step forward for PM, but it can only be the first step in a long road in improving the bandwidth issues and data paths within the centrino and successor platforms.

And Again, if you read my post carefully, I was explicitly talking about single thread performance; which is what the vast majority of applications that typical notebook users would need. We are not talking about a database server or a CAD workstation here.

Another question, why did you take away the "Intel Development Forum" Icon from your avatar.

Edit: Are you concerned that we will misjudge someone base on them working for Intel, AMD, IBM, etc. Well, don't be, we will take every one's opinion seriously.

But since we know where you work, we will take you opinion with a large grain of salt, as I'm sure you will understand.

Everything I have read about Yonah says high power consumption, limited FPU performance, and 32-bit. It "may" be faster than Dothan, but first generation new technologies from intel historically have not been major improvements over the pinacle of the previous generation tech. Merom is the one to watch with generous FPU improvements, faster and larger caches and better power envelopes.

If Yonah is a 50w processor as has been published, I'll steer clear of it for laptops in favor of Turions or Dothans. Yonah "may" be targeted more at the desktop/SFF market.

The problem is that 1 transistor = X power consumption. If a dual core processor doubles the transistor count it WILL increase power consumption and heat. Both bad for lappies unless fab processes can result in a reduction of X sufficient enough to make cost of the increase in transistors less than the new X.

The other problem is can Intel produce .65 micron chips that do NOT leak current like a sieve like the .90 micron Prescott P4s do.

On that note... Im curious what AMD is doing with the dual-core turions. Do we think it'll be a hott chip with 2 cores, or are the implementing something really cool that'll allow it to still be feasible in the thin&light category?

the ones coming in march are 35w TDP IIRC

amd CLAIMS that there is no added heat due to teh dual cores, but we really wont know till we actually see them in lappys.

if it was just an issue of yonah vs taylor, then i would get the amd. the problem is that with yonah i will be able to upgrade to merom, which will more than likely be superior to taylor, and i dont know what amd will offer as an upgrade to taylor to counter merom.

FWIW, here what I got from the package detailing a certain Manufacturer's upcoming releases:

Here are the Cinebench 2003 (32bit) numbers

Dothan 1.73 228
Dothan 2.13 268
Yonah 1.5 383
Yonah 2.0 521
Pentium D 3.2 528

It's highly unlikely that Yonah will be a 50W processor, not when full load tests of its server variant, Sossaman, shows an entire system power usage of 80W~90W, and that's with a dual, dualcore 2GHz system.

http://www.2cpu.com/review.php?id=109&page=5
http://www.hkepc.com/bbs/viewthread.php?tid=490871

Why not, Dothan is made on the same .90um process and it's unmatched by any other high performance CPU for power consumption and leakage.

Intel has already published Yonah TDPs

http://www.anandtech.com/news/shownews.aspx?i=25077
http://www.digitimes.com/mobos/a20051117A6021.html
http://www.dvhardware.net/article7584.html

The last two categories will be for LV and ULV, essentially today's 5lb and less and 3lb and less ultra-portable categories.

Most people will end up buying notebooks with Yonahs in the range of 25 - 49W, in the thin and light 5-7 lb category. The full sized gaming notebooks will take >= 50W, probably reserved for >=8 lb category of laptops.

And Merom seems to have a TDP ceiling of 45W:
http://translate.google.com/translat...l%3Den%26lr%3D

So, TDP is only a paper specification, empirical measurements are what matters. Dothan 533MHz FSB has a TDP of 27W for the entire family from 1.6GHz to 2.26GHz, even though at 2.26GHz, it still doesn't hit the TDP while running cpu intensive applications.