Slashdot Mirror

Actually, the G5 is a scaled-down Power4 .

From CPU Planet
The target for the standard CPU is power consumption under one watt during normal use (less than half that of the mobile Pentium 4), with overall thermal design power or maximum power dissipation of 24.5 watts for the 1.5GHz and 1.6GHz parts; 22 watts for the 1.3GHz and 1.4GHz; 12 watts for the low-voltage 1.1GHz; and just 7 watts for the ultra-low-voltage 900MHz chip. By contrast, the thermal design power of the 2.4GHz mobile Pentium 4 is 30 watts, while the desktop Pentium 4s range roughly from 50 to 80 watts.


Now, given that a 2.4 GHz Pentium desktop consumes 50 to 80 watts, and power demand is proportional to clock speed, a 600 GHz Pentium would
require around 600/2.4 * 50 = 25 Kilowatts, or enough energy to run four or five houses.

That is going to be one hot PC.

Your probably one of the idiots that modded me a Troll. But thats ok. I have too much Karma shit anyways.

xp can't do multiprocessing.

Low end centrinos idle in the low 60s. (watts)

No, Centrinos (just the Pentium M chip) run full-out at 7 to 25 watts:

Intel Centrino, AMD Athlon XP-M Spark Lightweight Laptop Blitz

There isn't that big a difference in the way 3d and 2d effects are rendered on a modern GPU.
There is a huge difference. Modern GPUs are basically general purpose vector microprocessors *with* some specialized hardware. One of those specialized pieces of hardware is a 2D ASIC, which has specialized circuitry for things like drawing lines. It is a seperate functional block from the general 3D core.

Yes, that is what would happen if the video driver didn't support the DrvTextOut
DrvTextOut doesn't rasterize the glyphs. It takes pre-rendered glyphs from the software font-rasterizer and alpha-blits them to the screen. This is exactly how Xft/XRender works.

Sure you can; it's called a "standard". I'm sorry that some other rendering engines are sloppy.
No you can't. The whole point of resolution-independence is to kill the relationship between pixels an coordinates. You'll get a different set of rendered pixels when you draw to a 300dpi target than if you draw to a 100dpi target. Also, it is not feasible to standardize the pixel output in the face of anti-aliasing, because hardware anti-aliasing algorithms in hardware are still evolving quickly.

I don't think that there is a seperate 2d and 3d engine on modern GPUs.
Yes, there are. Take a look at the block-diagram of the ATI R350 core (Radeon 9700).

Or are you saying that video driver writers have to resort to a hack to support alpha blending?
The video driver writers have to resort to a hack to support alpha blending. Most current 2D cores were not designed for alpha blending, so cores that support alpha blits tend to hijack the 3D circuitry to do it. Often, these hacks don't fully support all the blit features (like scaling), because the only critical software that depends on alpha blits is the font-renderer, which needs unscaled alpha blits for rendering anti-aliased fonts.

Is there some reason that the GPU can't do alpha blits on demand?
Hacks are necessary because the 3D core is set up as a pipeline. Vertex data goes in one end, pixels come out the other end. It's not easy to just hijack pieces of the pipeline and use them from the seperate 2D core.

There aren't seperate 2d and 3d drivers.
Yes there are. There is a GDI DDI (the Drv* functions), and a seperate Direct3D DDI.

As for concurrency, unless you have some really crappy driver that doesn't support asyncronous rendering
We're talking about concurrency between the 2D core on the card and the 3D core on the card.

Here: Transparency Bezier curves Anti-aliasing Gradients
The anti-aliasing link points to something that has nothing to do with anti-aliasing. Aliased output is simply not acceptable today. Transparency, bezier curves, and gradients might be supported in the GDI, but it's not efficiently implementable given the limitations of the GDI driver model.

Contortion could easily be made a space transformation
No it cannot. Do you understand graphics? *Convolution* requires a color matrix, which GDI doesn't supply.

and pixel shaders is too vague; pixel shading is a means, not an end.
Pixel shaders is not vague. A pixel shader is a program that can perform arbitrary per-pixel operations on rasterized pixels. GDI has no mechanism for doing such things.

Are you trying to say that these support functions don't really work; that they are just for pretend?
What I'm saying is that most vendors don't bother to properly hardware-accelerate these functions (because it's hard within the confines of the GDI, and nobody uses it anyway), and supply fast software-emulation instead.

The API already supports device-independent and scale independent rendering through coordinate space transformations as I stated earlier.
It's not a resolution-independent API if you have to set-up the coordinate transform for resolution ind

Here is an interesting FAQ about the new Mobile Athlon 64 processor.

Personally, I am waiting for some vendors other than emachines to release a notebook with this chip.

I am hoping for a configuration like this:

Athlon 64 3000 or 3200
Mobility Radeon 9600 128 MB
60-80 GB 7200 RPM HD
512 MB ram single stick or a full 1024 MB
1280x???? display

Background: I've used single CPU systems, HT systems, and SMP systems. I've taken courses on OS design and even in the process of writing my own. I'm quite familiar with the 80x86 32-bit instruction set and aware of the new 64-bit design as planned by AMD.

My $0.02 (this GREATLY SIMPLIFIED)

In the beginning there were CPUs. And CPUs were good.
Soon we realized the limitations and said.. Hey! Why not add another CPU and SMP was born.

SMP was good as well, however the additional cost was something of a deterrent for all but the power-users (and commercial applications of course).

Then Intel tried to develop a middle-ground, HyperThreading. It was a decent idea, however did not work quite as well as originally expected. AMD does not use it for a reason

From my experience I see HT as a hack developed by Intel, trying to duplicate true SMP. Might work sometimes and in certain environments but it's been show to actually slow execution in some situations (cache thrashing). In addition, SMP systems have much better responsiveness than HT ones under a high CPU load.

Which is why AMD is working on multi-core CPUs. This is the *correct* way (at least in my opinion) to tackle the problem, asides from getting true multiple CPUs. More can be read about it here. This combined with the new 64-bit instruction set (read more about that at the above link) will truly create a new era of CPUs.

Apple article

By the way, even though it is a 64-bit CPU, it has a 42-bit physical address space:

CPUplanet article on G5 and its memory

The release date of April 22 was released a long time ago

(ie: January 21, 2003, just incase you didn't get the picture)