“Although we didn’t consider it as such here today, Yonah will be quite impressive on notebooks. The thought of having such a cool running dual core processor in a notebook is honestly amazing, and the performance difference (especially for multitaskers) over what we have today will be significant. The other thing to keep in mind is that when you go from a single core to a dual core Pentium M notebook, you won’t be giving up anything at all. On the desktop side, you normally give up clock speed for dual core support, but Yonah will be running at very similar frequencies to what Dothan is running at today. In other words, you won’t be giving up single threaded performance in favor of multi-threaded performance – you’ll get the whole package.”
Lookout AMD. Quite impressive numbers and power consumption. This upcoming year will be most interesting that is for sure. Intel seems to have recovered well from the Prescott fiasco.
Hmmm, AMD Turion MT processors can be as good as Pentium-M in power consumption. Don’t count out AMD yet.
Anyway, it’s good to see that portables are finally catching up with desktops in perfomance.
Edited 2005-12-01 21:02
It looks like Intel will be able to remain dominant in Notebooks, and have a strong showing in general desktops, though it AMD may still be your best best for high powered and gaming PCs where powerconsumption is a secondary thought, though AMD still doesn’t seem to be able to field a serious thread against Intel in notebooks.
Seems like a rampup to Intel Macs. Pentium M based desktops what a great idea been looking at a new AMD based system but looking at this makes me want to hold out for Yonah.
I have been slightly disappointed not to find the comparison of the power consumption agains the single core Pentium M (Dothan???).
While it is interesting to have more powerful CPU’s, it is also very interesting to have a lower power consumption. Especially when the bottleneck of current laptops lie more in the 4200 or 5400 rpm hard disks.
By the way, how will a G4 Powerbook compare to a Dual Pentium-M iBook?? I’m eager to find out if the dismal SQL performance of G5’s is due to Mac OS or to the G5’s (even though altivec is impressive)
“Especially when the bottleneck of current laptops lie more in the 4200 or 5400 rpm hard disks.”
When I need performance most (processing photos), the bottleneck is not disk performance. Raw CPU / memory performance is of interest to me.
From what I have seen it is OS X, you might find these interesting:
No more mysteries: Apple’s G5 versus x86, Mac OS X versus Linux Part I and II
http://www.anandtech.com/mac/showdoc.aspx?i=2436
http://www.anandtech.com/mac/showdoc.aspx?i=2520
Its OS X actually. Anandtech did a thorough analysis. OS X has dismal performance for server apps. Something to do with context switching.
“Its OS X actually. Anandtech did a thorough analysis. OS X has dismal performance for server apps. Something to do with context switching.
I read that too, but I’ve also read some interesting followups from a number of differnt sources (still looking for the links unfortunately) that claimed the design of MySQL itself (being rather Linux centric) was partly to blame for Mac OS X’s dismal performance.
Don’t get me wrong, I am not claiming that the context switing involved due to XNU’s implementation is stellar. Damned near everything I’ve read about it says that it really could stand some improvement. However, taking one set of tests on it’s own is kind of silly, especially when (as in this case) the software that was tested is rather Linux centric; as in, it was designed, tested and evolved with Linux’s strengths in mind, as Linux evolved along side it while removing any of the previous performance disadvantages it once had.
In that respect (and a few others as well) Linux is ahead of OS X. But it is foolish to take anything that changes as fast as software at face value, as you’re only looking at one particular snaphot in time. Mac OS X already does some things better than Linux. It’s just the way things go.
AFAIK, MySQL is a pretty “funny” application when it comes to multithreading, as it has some pretty steep requirements to the threading implementation. It’s been fiendishly slow on FreeBSD a few years ago (Don’t know about now), so if OSX is based on the threading implementation of that time, it’s no real surprise…
Mac OS X’s kernel doesn’t use FreeBSD for the kernel’s threading, it uses Mach threads. It uses FreeBSD kernel bits for the *userland* POSIX threading interface.
Newer releases of MacOS X (10.3+ I think) is based on FreeBSD 5.3. There have been a lot of benchmarks with FreeBSD 5.3, 5.4, and 6.0 that show the default threading library (M:N libkse/pthread) to be on par with the LinuxThreads implementation. And the 1:1 libthr library is generally better than libkse for MySQL on FreeBSD 5/6. See the past 3 months of archives for the FreeBSD -current, -hackers, and -stable mailing lists for more info.
The interesting part, for MySQL performance on FreeBSD (and probably MacOS X), is the way MySQL handles time. It appears that MySQL uses the gettimeofday() family of functions several hundred times per second. And the gettimeofday() family of functions on FreeBSD are much more expensive and accurate (higher precision) than the gettimeofday() functions on Linux. Which means MySQL runs faster on Linux than FreeBSD, but has more accurate time on FreeBSD than Linux.
There’s a nice long thread on the -stable mailing list about this.
With FreeBSD 6.0, the threading library is no longer a bottleneck for MySQL performance.
Since MacOS X is based on bits from FreeBSD 5.3, though, threading may still be an issue when it comes to MySQL performance. But, since it’s not a pure FreeBSD kernel, there will be other factors in there that may impact MySQL performance as well.
The performance of FreeBSD is largely irrelevent to the performance of Mac OS X. OS X uses only small parts of the FreeBSD code in the kernel. Filesystems and networking, I believe, come from FreeBSD. Several critical bits, the scheduling and threading infrastructure, IPC, and the VM, are from Mach, and most of the BSD layer is a heavily tweeked 4.4lite2.
Believe me, if OS X were based on FreeBSD, it wouldn’t suck nearly has hard for system-level stuff as it does. I’ve been using a fast PowerMac for day-to-day development tasks, and stuff like dealing with tarballs or compiling source trees is much faster on my Linux machine clocked 100 MHz slower. I’ve tried FreeBSD (6.0) on the same machine, and it feels just as fast as Linux. One of these days, I’ll get around to trying Linux on the Mac, but unfortunately I can’t get Ubuntu to boot on my machine.
Newer releases of MacOS X (10.3+ I think) is based on FreeBSD 5.3.
No, it’s based on Apple’s own kernel called XNU, which in turn was based mostly on Mach 3.0, with quite a lot of FreeBSD thrown in to provide POSIX compatibility. Preemptive multitasking and threads are implemented by the Mach microkernel, with pthreads or other threading libraries sitting on top of that.
http://www.kernelthread.com/mac/osx/arch_xnu.html
I get a MySQL Sysbench score of 200 on my Ibook running PowerPC Debian Linux. This is not bad.
The MySQL performance of the G4 and the G5 is actually quite good (on Linux) even though MySQL is not really optimized for PPC yet.
Cheers
Gunnar
Digital Media Boost technology will make your porn faster.
I wonder if anyone’s tried to put 4 4gb or 8gb flash cards in a laptop in place of the HD? That would have to save you a ton of power and make things blindingly fast as well. For business users that would probably be enough storage if you don’t need tons of software installed. Obvously this would be somewhat expensive
Two points to consider. Flash memory is only useful for a limited number of writes. Sure it may be a million times, but hook a flash memory card up to a swap partition for a web browser cache and it’s gone in a few months.
second Apple is working on integrating Flash into the mother boards.
My thought on the subject is like this. Put the core OS on a flash disk and use the HD for applications, swap(remember OSX is unix files mount points can be anywhere). The OS will boot up literally tens of seconds faster, Applications won’t load much faster but greater storage can be had for the user. (ie OSX takes 2-3 gigs for install) a 4 gig flash card(like the iPod Nano’s hint). Power consumption goes down as the hard drive isn’t accessed as often.
Increased battery life, faster startup times, increased HD space for the end user, all in one shot. Of course if you could figure a way to hook that flash card into an L3 cache spot on the chip itself that would be even cooler, well faster, and more efficient. though how I don’t know.
“Put the core OS on a flash disk and use the HD for applications”.
Applications can also be stored on a flash disk with no penalty. Applications (the executable part) are changed not more than a dozen or so times throughout the life of a machine – so they are hardly likely to exhaust the re-write capability of flash disks.
Only things like swap, cache, configurable settings, logs and application data (the actual documents and files that users of the machine produce or download during the life of a machine) need to be stored on a HD.
http://www.bitmicro.com/products_edisk_25_ide.php
These guys have been making SSD’s (Solid State Disks) for some time now. The link here is for their 2.5″ drive you could throw in a notebook. Not so sure on the price, though I would imagine it’s fairly hefty (I tried to call them to find out but they had already closed for the day). The primary bennefit of Solid State Disks in industrial applications tends to be durrability. You throw the thing agains a wall and it just keeps on going like nothing happened. I’m not so sure it would be all THAT much faster though. These guys report sustained reads and writes of ~ 28MB/s and bursts up to 66MB/s. Not all that much more impressive that high end HD’s nowadays
I really look forward to seeing these in iBooks/PowerBooks. Should be quite nice.
that the test chip barely runs cooler than a desktop 90nm A64 (single core)?
http://www.anandtech.com/cpuchipsets/showdoc.aspx?i=2410&p=2
Laptops with chips that hot generally run 1.5 inches thick or more.
AMD has been selling dual core 2.0ghz Opteron’s at 55W max for months now (according to AMD.com) and this chip seems to be running quite a bit hotter than that as it’s only a bit cooler than the 90w max A64 dual core chip.
I don’t see this beating the Turion at all but it’s still nice that Intel is catching up.
It would have been nice to see the AMD chip tested at 64-bit as well so we can see both chips at their best. AFAIK 64-bit gives the AMD about 20% more performance.
Anyone else notice…that the test chip barely runs cooler than a desktop 90nm A64 (single core)?
Two problems with your conclusion:
1. Those are system power draw measurements.
2. The Yonah tests were done on a desktop board.
As a result, it simply isn’t possible to compare Yonah and Turion based solely on the information in the article. IIRC, Anandtech staff said in their comments section that they’d be doing more power tests in the future.
t would have been nice to see the AMD chip tested at 64-bit as well so we can see both chips at their best. AFAIK 64-bit gives the AMD about 20% more performance.
Not from what I’ve seen:
http://www.extremetech.com/article2/0,1697,1857476,00.asp
Remember that we are talking about Windows software.
Two problems with your conclusion:
1. Those are system power draw measurements.
2. The Yonah tests were done on a desktop board.
Neither of those things matter because I wasn’t talking about total system draw but the difference between the two chips. Putting the Yonah in a laptop board may lower the total system draw but it isn’t going to do much to change the difference between it and that same A64 also in a laptop board.
If the difference between the two desktop systems is 20w then the difference between two laptop setups with the same chips shouldn’t be much different.
Not from what I’ve seen:
http://www.extremetech.com/article2/0,1697,1857476,00.asp
Remember that we are talking about Windows software.
http://www.pcstats.com/articleview.cfm?articleid=1665&page=6
Cinebench, Divx, Gzip, WinRar, and most other apps with 64-bit executables or that can use the extra registers benifit greatly. Even some 32-bit apps like After Effects get large speed gains as well. Even Photoshop gets about 7-10%.
http://www.planetamd64.com/lofiversion/index.php/t7774.html
Gaming, sadly, doesn’t really improve much at all so anyreview that does a lot of gaming benchmarks might come to the conclusion that there is little increase. But then again, most games don’t benefit from dual core either.
Really, thinking that AMD will be able to match Yonah with a dual-core Turion seems highly unlikely. This model Yonah is expected to max out at 31 watts, a single-core 2GHz Turion is already at 35 watts. Sure there may be progress from AMD, but Yonah is not an easy chip to beat.
On the topic of hard drive and flash memory performance it is interesting to point out that there is a middle road being taken as we speak; Large flash caches on hard drives. Samsung are expected to have products on the market quite soon, see for example http://www.electronicsweekly.com/Articles/2005/04/25/35080/Harddriv…
Considering the fairly low prices on high-performance flash we can probably look forward to having a significant part of the data one accesses most often in cache and thus very speedy.
You have to remember, also, that the Turion is 90nm while the Yonah is 65nm. Makes a difference with power consumption.
Also, dual-core doesn’t really add much in the way of power consumption, even at 100% CPU (this is even true on Athlon64 X2’s).
Really, thinking that AMD will be able to match Yonah with a dual-core Turion seems highly unlikely. This model Yonah is expected to max out at 31 watts, a single-core 2GHz Turion is already at 35 watts. Sure there may be progress from AMD, but Yonah is not an easy chip to beat.
1. That 31w number was Intel’s original estimate. That number was later changed to 49w.
2. Intel’s TDP rating is based on 75% load and is deemed the “typical” wattage not the max. AMD’s TDP rating is a precautionary maximum that mobo makers must design their mobo’s to handle. So, for instance, an AMD chip rated at 89w never even comes close to that wattage even under the worse kind of stress. A64’s rated at 89w will only hit about 43w under full load.
3. AMD has 2.2ghz single core Turions @ 25w max:
http://neoseeker.pricegrabber.com/search_getprod.php/masterid=12725…
2.4ghz versions are at 35w max right now.
4. AMD ratings also take in account for the memory controller (northbridge) and Intel ratings do not.
IIRC, they’re going to be integrating the memory controller into the chip in the next release, there is a bunch of work being done on improving watt/output, and decreasing leakage – some of the ideas wouldn’t have made it into Yonah, BUT if they can increase the performance, but maintain the same level OR decrease the level of power consumption – all will be good.
With that being said, however, the CPU isn’t the only component; what they need to do is think smarter on how to lower consumption – first, more effort is needed in regards to coming up with smaller batteries with longer life; as another person said, dump the CPU onto a flash chip, on the computer itself – call it the RISCOS, but instead of it being on a ROM, its on a storage medium that can be changed and updated – and heck, to make it really reliable, you can make the flash read only in the OpenBoot, meaning, if something goes wrong, it is just a matter formatting the hard disk, all the settings are removed, and everything goes back to defaults.
Though, the chip looks to be real promising, I can almost gua-run-tee that the minute you install Winders XP or Vista, the thing will run like a crippled dog.
…for macs. Some of the performace tests should have faired better I think. Speaking of which, I can’t believe Anandtech apologized for not having a lot of tests/comparisons. Sheesh. I kept clicking on page after page of tests/comparisons.
I’ll be picking up a new laptop after Christmas. Since I’m not going to be 2d/3d content creation, The single core Dothan will work fine for me. The model(s) I’m looking at will have a Nvidia 7800Go, so gaming isn’t a problem (most games can be maxxed out except something like F.E.A.R), and it looks like that Dothan’s lower L2 cache latency actually helps it improve in the business benchmarks, so things like big IDEs that I use will run just as fast. I’m sure things will be more interesting in a year or so when Apple’s transition to Intel has stabilized.
Wait for the dual core laptops to come out. When they hit the market you’ll see a huge price reduction in anything with a single core. Single core laptops will basically become low-end or bargin laptops.
Wait for the dual core laptops to come out.
That would be nice, but unfortunately the time frame I’m working with isn’t going to allow that. Prices are always going to come down. I’m hoping for some post-Christmas deals that I can work with.
Is Intel’s dual core Yonah 64-bit capable?
Edited 2005-12-02 00:50
Every new Intel CPU will be EMT64 capable. Even the new Celeron will be EMT64.
It’s slowly the end for 32bits CPU. Everybody is waiting for real drivers for XP-64 or Vista-64…
Yonah will not be 64 bit, dont listen to what they say. Memron and Conroe will be 64 bit, as will all other intel chips after Q2/3 2006. I’m too lazy to google a link now, but if your really interested, it shouldn’t be hard to find.
What makes Intel think consumers will want to wait till Q4 of 2006 to purchase a laptop with Intel’s Merom 64-bit dual core processors? After all we can get that now with AMD’s Athlon64 x2 which is sold by Alienware, GamePC and other distributors. I find it surprizing that Intel is one year behind AMD when it comes to offering dual core 64-bit mobile processors.
Maybe because noone really cares about 64-bit?
> Maybe because noone really cares about 64-bit?
And there is no need for more than 2 computers in the world… 🙂
I care (therefore your statements is false…) and soon everybody will care.
You care because marketing convinces you that 64 bits is better than 32? Are you planning to install more than 4GB of RAM in your laptop? I don’t think so. For high-end desktops and servers, there are still only a subset of applications which benefit. In fact, on many applications, performance may suffer due to the increased amount of memory bandwidth necessary to deal with 64-bit pointers.
You may not want a Yonah, but other people will. Merom will be even better
A lot of people care because amd64 fixes the register issue with x86, which makes code run 10-15% faster. This is something I notice all the time on my system. For example, blender runs about 20% faster when compiled for amd64 as compared to when it is compiled for x86. These Athlon64 vs Yonah benchmarks are a bit off for anybody who can uses 64-bit programs (me, for example, I don’t use Windows and Ubuntu has a native amd64 version), because they test only 32-bit code.
A lot of people care because amd64 fixes the register issue with x86, which makes code run 10-15% faster.
Much of that isn’t due to the extra registers, but the improved calling convention. Standard x86 code is bogged down by passing arguments on the stack and maintaining the mostly useless frame pointer, thankfully AMD have done away with that when writing the programming manual for AMD64.
Of course you could do the same with 32-bit code (gcc actually supports it), only no-one, especially MS with it’s woeful compiler, ever bothered.
Weird how so much effort goes into getting more performance out of the hardware when software needlessly throws it away like that.
(By the way: does anyone know what calling convention Apple is using on x86?)
And of course it all has little to do with the 64 bits. In actual fact in many apps they cause a slowdown countering the benefit from extra registers, because bigger pointers take up more cache and memory bandwidth.
Shame AMD didn’t make the extra registers available in 32-bit mode as well, that would have cost next to nothing in extra hardware.
How the hell can you NOT be a multitasker on any given OS other than, you know, DOS?
I’ve got a dual-core AMD 4200+ system. I frickin’ love it. The funny part is, that I don’t notice nearly the difference in performance on Linux as I do when it’s booted up in Windows.
Windows apps can sometimes go “nuts”, consuming 100% of the CPU. Not that Linux apps don’t do the same thing, but it happens more often in Windows. With one CPU, this bogs down the system. With dual cores, it’s no problem.
And I can alt-tab from Quake4 or UT2K4 to another application quickly and easily, where with one core it takes forever.
I really wish my IBM Thinkpad was dual-core. The hardware of the Thinkpad is nice, but all those dumb apps that IBM loads onto their Windows builds are crap. Everytime it comes out of sleep, some eGather process eats up 100%, along with other IBM apps, and makes the system unresponsive right out of sleep for a few minutes if I don’t kill it.
It will be amusing to watch OS X demolish Windows performance-wise on the same hardware, just as it has on different chip families. XP is a big step up stability-wise vs Win2000, but speedy it ain’t.
I love these Hebrew sounding names. Yonah (dove) Dotan (it’s a flower, don’t ask me witch. quite a comman hebrew name. for boys.)