Briefly: Intel Mac mini in production; iBooks due at later date - Page 3

OK...educated guesses. Thanks.

right.
the benchmarks posted is more like 'gma and cpu opengl performance' vs. 'ati-gpu opengl performance'.

the point of accelerating the gui with use of the gpu is to free cpu resources, no?

afaik, the requirement to fully enjoy hardware acceleration is (quasi-officially) a shader model 2.0 capable gpu. (kind of fishy to use directx-definitions, but still...)

the gma 950 does that, but im not sure if the (lacking) t&l-engine is considered...?
anyhow, the way coreimage is built, it will use whatever hardware features at your disposal and i guess t&l isnt very significant (if at all) when it all comes around.

That's the old Dothan Pentium M. The Celeron M 4xx is based on the Core Solo. Since the Core is a little faster than the Dothan, I think it may reach 2x a G4 1.4.

yes. both the imac and mbp got a 945gm-chipset with a gma950 built in.

(there were a lot of 'wtf?!s' when the first images of a gutted imac surfaced, remember?)

On the whole I'd agree although in some tests the G5 2.1Ghz iMac was quicker than the Intel at native apps.

I doubt that they will release something that isn't quicker but they've yet to release a machine faster than the PowerMacs. If speed is what you're after then PPC is it. The reason they've not is that Intel doesn't have anything faster than the dual core G5.

If you're running anything from Adobe, a G5 iMac is still quicker than an Intel iMac.



Original price. At launch the Intel iMac was £929 as opposed to £899 for the G5 despite Jobs standing on stage saying 'Same Price'.



Ultimately it's the speed of the apps that determine the experience and without native apps, currently the Intels aren't as fast as PowerPCs. I work in web design, programming and graphic design not sound. I don't know anyone professionally using a laptop that doesn't also have a desktop for real work. That's maybe just my sphere of users. The only laptop users I know are indeed the ones that use them for emailing grandma and general net use or they're road based so are stuck with a laptop.

LOL, over the weekend I took my machine over to my girlfriend (hardcore PC user) to get some work done and she was so impressed that she is switching to Macs .

She couldn't believe that while she was dealing with pop ups - she didn't even know where they came from, no browsers were running - had to reboot twice, constantly closing messages that her virus scanner prevented another virus attack ... etc., in the mean time I was just WORKING ... LOL. I also showed her that I run Windows 2000 Pro (Virtual PC version) to test my websites for Windows compatibility and also can run UNIX (she also works with UNIX at work). All that on one single machine that works and free of the frustrating she deals with every day. She also coudn't believe that in the past ten years I never even had a virus protection program installed. I told her many people just ignore Macs without even looking into how powerful they really are.

She was the second person I convinced to switch .

Those are pretty rare, and almost always functions that aren't optimized for dual processors. Once those few get optimized, they'll be faster as well.


That's better. People tend to forget that there are pro users who don't depend on Adobe.


As you say, it comes down to your apps. You're better off waiting. But many people are either good to go already, or will have their updates shortly. It doesn't make sense to make generalizations for all "pro users" (whatever that means) since it's different for everyone.

I guess I'm a 'Pro user', I design websites for entrepreneurs and small businesses so I am more concerned with the speed of my apps, or how fast I can get my work done. That is why I am interested in performance on the higher end models and the performance of the pro apps on these.

Now the Motorola chip used only 7 pipelines (or steps) for graphic processing vs. Intel had to go through 20 (!). There was a major difference there and that is one of the reasons graphic processing with apps like Photoshop is (was?) faster. So now with the Intel switch I really need to look what cost savings I get and what performance improvements I get for my money.

Now as a home geek of course graphics for gaming, video editing and the performance of other apps are also important for me.

So sounds like all in all I better stick with the PowerPC machines for now.

If you rely heavily on Adobe software, buy a new PowerPC Mac, and get an Intel Mac next year.

Of course if you're upgrading from an older G4, get an Intel Mac, because even under Rosetta (which is quite impressive), your Adobe software will still run at least as fast or faster.

So what are Apple's plans for switching the desktops to Intel? Anyone has good info on that?

It'll depend on Intel releasing a decent desktop chip. At the moment, the roadmap says it's probably a chip codenamed 'Conroe' running up to 2.66Ghz and shipping 2nd half of 06, probably nearer the end than the the beginning.

The one fly in the ointment is that Intel traditionally limits desktops to only two cores so Conroe replaces the bottom and middle PowerMacs. They have a server chip codenamed 'Woodcrest' which may be the replacement for the Quad G5 chip. Or perhaps Intel will stop limiting their desktop chips.

Around the same time is 'Merom' which is the replacement for today's Intel 'Yonah' based Core chips.

Merom, Conroe and Woodcrest are Intel's 64bit Next Generation Architecture. Yonah is the last of the old Pentium M 32bit architecture

Around the end of the year we may also see OSX 10.5 Leopard. A smart person might wait out for both Intel's next chips, Apple's second attempt at an Intel box and the new OS. That's if they don't need new kit now. If you're running Adobe software as I said earlier, and you've a decent G5 now, there's little point in upgrading now.

Thanks aegisdesign, yes I think it's sounds like a good idea for me to stick with the PowerPCs for a while. It makes me feel confortable that I wont have to worry about investing in upgrades for a while yet.

Thanks again!

I'm even questioning the assumptions that Apple will use Conroe for the new Powermacs. This seems to be the mainstream guess. If cost is the issue, maybe. Those chips are definitely faster than the laptop versions, the Yonah, and the Merom.

But, nevertheless, they are midrange chips. The PM doesn't compete against midrange machines. Not in the pro area.

I just can't see a Conroe PM stacked up against a Woodcrest machine. Right now, and for the last several years, PM's have gone up against Xeons and Opterons in the video editing arena. Also, for scientific uses, Apple would need something better than Conroe.
I think that it's at least 50/50 that Apple will go with Woodcrest, possibly, at least, for the top machine, if not the whole line.

By the way, I don't think that it's whether a quad core exists for it that will matter this year. The Quad has two dual core G5's, not a single quad core chip.

With Intel's memory bus problems, I think that two dual core chips will prove better than one quad core chip, especially as the Woodcrest uses up to a 1.333GHz memory bus, considerably faster than Conroe's max of 1.066GHz.

There's not much of a tradition for dual core yet. It really isn't a desktop/server dichotomy, the previous Xeon DPs were workstation and server chips for systems that had up to 2 processors. The MPs were almost exclusively server chips, generally for systems that have up to four processors. The PowerMac line is, IMO, clearly a workstation type machine so I would expect it to use the workstation chips.

http://www.intel.com/products/chipsets/945gm/

Now taking into account the actual cutout motherboard:

http://www.kodawarisan.com.nyud.net:...c_intel02.html

It makes sense that with what the 945GM chipset does as a mobile chipset that it be disabled since the iMac is not a laptop, lower-power system. Not to mention it is cheaper to disable this subsystem and conserve on manufacturing retooling costs than have a separate MOBO specific for the iMac tooled by Intel just for Apple.

Now the Mac-mini is a custom form-factor designed motherboard. I don't expect Apple to enlarge this system. I expect Intel to work with the original form factor and design a motherboard that meets Apple's requirements.

Having a disabled 945GM on-board is a waste of space.

Either there is a model that has this enabled for the lowest end possible or there is a model without this on-board and a slot mini-atx form-factor compliant GPU.

I'm in agreement with an earlier poster who pointed out that the Mac-mini will start at $499.

The iMac already limits the Radeon X1600 to 256MB when ATI clearly states it supports up to 512MB DDR3.

Ideally, if Apple wants this to be their entry into the Home Theatre market then to have an X300 upgradeable to the X800 BTO makes the most sense.

If it comes with integrated graphics and gets marketed as their product entry into the Home Theatre it will be the kiss of death.

I admit that I'm not terribly familiat with Integrated graphis, so I only know what I've been reading. But, what I wonder is whether a shortly available IG chip will be able to decompress 720p and 1080i H.264.

That's all it had to do for home theater, assuming that whatever cpu chip Apple will put in, can't do it itself, as a Core Duo Yonah 1.83GHz chip can, with ease.

Home theater doesn't require great 3D capability. Any GPU that is programmable, and that can handle core services will suffice for the 3D work.

Home theater requires a good GPU that can handle 2D decoding. If the cpu won't be doing it.

Look at the diagrams again.

Not only has the "Drawing Reuslts" moved over onto the GPU, so has Quartz 2D.

The fact that the results are stored in VRAM as opposed to RAM is a side effectin AGP systems (remember, that chart pre-dates PCI-E), it was prohibitively slow to move data from the GPU back to regular RAM. Uploading textures to the card isn't a bottleneck, thus storing them in VRAM alone has no benefits. The only benefits are the drawing speedup.

Additionally, you forget one thing about Integrated chips. They're, well, integrated. This means that they access regular RAM almost as fast as dedicated GPU's access VRAM, so even if the VRAM memory store WERE to make a difference, integrated chips wouldn't suffer at all.

And just to be clear, Quartz 2D Extreme isn't enabled yet, anyway.

Surely the X1300 is the obvious canditate for the Mac Mini, or maybe the x600 from the last iMac G5? I bet Apple got those pretty cheaply now they are obsolete, it would still be a hell of an upgrade from the 9200 in the current mini.

Yes, I know that. You still don't get it. The GPU doesn't actually run that Quartz 2D bit that much faster than a CPU. It just means the the CPU doesn't have to do the work, so a few cycles are freed up for the CPU to spend on other tasks.


Correct.


Incorrect. The pathway that was 2.1 GB/s is now 30 GB/s. That is a bottleneck in Quartz, as the output from the "Quartz 2D" block is huge bitmaps - a lot of data. Yes, you are right that the diagram was drawn in the days of AGP, and that PCIe is faster, but system RAM is still DDR2 whilst VRAM is DDR3, so VRAM is still faster.


Again, system RAM and VRAM are one generation apart in terms of speed, so dedicated VRAM does have significantly higher bandwidth.



Yes, I know. Apple seem to be having a lot of trouble with getting it working I've started to wonder if they're actually going to have to wait until QuickDraw is fully deprecated before enabling it, and that might not happen until 10.6

From the two Mac Professionals that I am in constant contact with, Rossetta is not all that impressive. Test showed that when using 3rd party software the PowerPC Macs are running MUCH faster. No matter what you do, if the app is not optimized for the processor you are using, the performance is much slower... Your overall advice is good, though. If you are a Mac Professional and rely on 3rd party apps, buy a PowerPC. You will be just as happy as you would have been even if Apple wasn't chaning chip manufacturers...

*cough* Currency market *cough* The US price is the same, he was also standing in the US when he said it.

I just filed my income tax. Bigger return than I hoped for. I have no need for a new laptop or desktop as my iMac G5 and 1.5ghz 15" PB are do me just fine.

However, now I am really hoping for either a true iPod video, tablet, or PDA type device so I can blow money I should probably save.
Alas, I have little hope for any of these products come tomorrow.

Ok, here is what I think that you should do.

Buy a shuffle and send me the rest of your tax refund...

an itms gift certificate will be sufficient.

You just inspired me to start my 2005 return.

I love how turbotax imports last years data.

Haven't done any deductions or additional earnings yet, but I owe some money. Not a big surprise though and its a good thing I made estimate payments throughout the year.

Main system memory isn't nearly as fast as VRAM. Single channel 667 MHz DDR2 RAM has a bandwidth of 5.3 GB/s. I highly doubt the mini would even get 667 MHz RAM, much less dual channel, so the memory bandwidth would look even worse. This has to be shared with whatever else the system is doing at the time. DDR3 is up to around 16 GB/s, and this is dedicated purely to drawing.

Also, it's more than just textures that are kept in the VRAM. Fonts, bitmaps, window textures, interface controls, etc are all cached in VRAM. Now think about when resolution independence hits, and all those bitmap elements need to be resized in real time. Having bitmaps cached in speedy dedicated VRAM will make a difference. If it didn't, Apple wouldn't go through the trouble of moving the backint store to VRAM at all.

Yes and no. The GPU is dedicated to drawing and can handle dramatically more vector operations than can either Altivec or SSE3. That's why people have GPU's. They're fast.

That said, you're also right, it also gets some nice speed benefits by freeing up the CPU.


You just contradicted yourselfbut you're right the first time. Say you're drawing to a large window... say, 1024 x 1024 pixels, or one megapixel. That works out to exactly 4 MB of data. 2.1 GB/s works out to 35.84 MB per screen refresh. So you can upload 9 uncompressed megapixels per frame. In fact, Apple's 30 inch display can only support 4 megapixels. So even with two 30" displays, you could never possibly be sending so much data to the memory card that it would be a bottleneck.

Therefore, ALL speedgains in Quartz 2D Extreme come from performing the drawing operations on the GPU itself, not from the caching them there.


I'm not sure what you mean by this...

My point was that while with dedicated GPU's, RAM accesses are slow (not a bottleneck in terms of drawing, but certainly in terms of Core Image operations or gaming). With Integrated GPU's, RAM access isn't nearly as slow because it doesn't have to go through AGP or PCI-E the GPU gets RAM as fast as the processor can, much much faster than a PCI-E GPU can access RAM.

I misunderstood and thought that you assumed that Integrated Graphics were as slow for RAM accesses as were stand-alone GPU's, which is not the case.

See above about this. For basic drawing and compositing, this makes no difference because so few resources are used. Bandwidth only becomes an issue when it's a bottleneck, it's not in this case.


Sort of. Fonts are not cached, because Macs use sub-pixel kerning. Each time a character is drawn to screen, it's slightly different. Also, interface controls are only cached as part of the window, the pieces themselves aren't cached.

Bitmaps can be cached if you program it yourself using OpenGL contextsit's not done for you and it's not all that easy.

Windows themselves have been cached since Jaguar, I believe, but in fact they often cache in system memory (and compressed) for two reasons. One, GPU's only support lossy compression, which is totally unsuitable for anything but game textures. Second, they add up really fastonce you add more than a handful of windows, your VRAM is all filled up.

Typically, only the front-most handful of windows are cached in VRAM.

?? No, I haven't. I said that an operation that doesn't need to send much data (drawing commands) isn't bottlenecked by memory bandwidth, but operations that do send a lot of data (drawing results = bitmaps) are bottlenecked by memory bandwidth.


The thing that is limited to the number of pixels on the screen is the output of the quartz compositor. In the case of the 30" display, this will be 2560 pixels wide by 1600 high by 24 bit pixel depth. The output of "Quartz 2D" is edit modifications to /edit bitmaps edit which are cached in RAM (Quartz 2D) or VRAM (Quartz 2D extreme) /edit with 32 pixel depth (24 bit colour + 8 bit alpha channel). Each individual bitmap is limited to the size of the screen, but there is not just one bitmap, there are many, many bitmaps (at least one bitmap for each window + plenty of other stuff), which are then blended together by the quartz compositor to make the final scene that is sent to the display. edit In the case of Quartz 2D, the compositor has to read all of these bitmaps from the system RAM cache, and in the case of AGP, this is bottlenecked to 2.1 GB/s. Moving the cache to VRAM gives the quartz compositor much faster access to the bitmaps it must blend together.

You are right that the memory bandwidth for an integrated GPU is higher than a GPU running over AGP, but it still is not as high as a GPU with dedicated VRAM. This means that the performance differential between no dedicated VRAM vs. dedicated VRAM is smaller for an integrated GPU than for an AGP GPU, but there will be a differential (probably only observable when there are many open windows). /edit



No, some of the gain is from the GPU doing the calculations (and this helps more as non-quartz CPU load increases), but most of the gain is from the increased memory bandwidth.

I suggest that you read this

If that were so, the Intel model should be cheaper as the $ is weaker now that it was back when the G5 was launched at £899. However it's irrelevant as the price of the G5 model stayed at £899. Even with Apple's policy of never changing the prices unless they release something new, it made us Brits do a double-take.

IIRC Fonts ARE cached in Quartz2DE as hints, not bitmaps, and redrawn by the GPU. This explains why enabling Q2DE on my old iMacG5 with FX5200 is so slow at drawing text. Slower than using the CPU to do it. It's the one reason I don't enable it as most of the time I'm using text. UI speed is a bit slower too.

Pity really as almost all the other graphic operations are faster.

International prices are what they are. Just deal with it.

Isn't an Apple computer that gets shipped to someone in Europe made in the same place as one that gets shipped to the US?

...easy for you to say

Yes - China/Taiwan, although we do get the better power supply in European iMacs.

It is pretty much the same computer, but it crosses different borders. Those border differences mean different currencies and different import taxes. Taxes make a difference except in the very smallest transactions. When I buy something from overseas, I don't consider duties, national sales tax, VAT or import taxes. A buddy from a different country might have to deal with one or more of those.

This site is truely screwed up today. Took me most of the day to get on. Now it seems as though posts are missing, even though I got notices in my e-mail.

Where are all the posts between Aegisdesign's and JeffDM's?

This isn't about taxes. It isn't about currency rates.

Our taxes didn't increase.

The US price remained the same. The UK price went up.

Apple UK are just weird. For instance, the UK store has this graphic on it's page...



yet you click through to the iMac and you get this...



Been like that since the iMac Core Duo launch when the G5 was £899. The G5 iMac 20" is coincidentally £1049 but they've got that price right. We quite often get prices plucked out of thin air here.

That's the problem with currency rates fluctuating moment by moment.

If the pound goes up by .5%, your price will reflect that. .5% on a 1,000 pound device is 50 pounds. the price on last weeks now discontinued item would have been less.

NO IT'S NOT.

Please read again and look at the graphics.

But if you want to play the exchange rate game, the price should have gone DOWN if US dollars is the baseline. We get more pounds to the dollar now than we did last year.

Here's how many pounds to the dollar ...



However, I suspect this has more to do with the Taiwan Dollar exchange rate which took a tumble at the start of this year. Apple USA seem to have absorbed it. Apple UK haven't. Or for the conspiracy theorists, Apple USA is clawing back the difference from Europe.