Intel's Larrabee Architecture Disclosure: A Calculated First Move
by Anand Lal Shimpi & Derek Wilson on August 4, 2008 12:00 AM EST- Posted in
- GPUs
The Design Experiment: Could Intel Build a GPU?
Larrabee is fundamentally built out of existing Intel x86 core technology, which not only means that the chip design isn't foreign to Intel, but also has serious implications for the future of desktop microprocessors. Larrabee isn't however built on Intel's current bread and butter, the Core architecture, instead Intel turned to a much older architecture as the basis for Larrabee: the original Pentium.
The original Pentium was manufactured on a 0.80µm process, later shrinking to 0.60µm. The question Intel posed was this: could an updated version of the Pentium core, built on a modern day process and equipped with a very wide vector unit, make a solid foundation for a high-end GPU?
To first test the theory Intel took a standard Core 2 Duo, with a 4MB L2 cache at an undisclosed clock speed (somewhere in the 1.8 - 2.9GHz range I'd guess). Then, on the same manufacturing process, roughly the same die area and power consumption, Intel sought to find out how many of these modified Pentium cores it could fit. The number was 10.
So in the space of a dual-core Core 2 Duo, Intel could construct this hypothetical 10-core chip. Let's look at the stats:
| Intel Core 2 Duo | Hypothetical Larrabee | |
| # of CPU Cores | 2 out of order | 10 in-order |
| Instructions per Issue | 4 per clock | 2 per clock |
| VPU Lanes per Core | 4-wide SSE | 16-wide |
| L2 Cache Size | 4MB | 4MB |
| Single-Stream Throughput | 4 per clock | 2 per clock |
| Vector Throughput | 8 per clock | 160 per clock |
Note that what we're comparing here are operation throughputs, not how fast it can actually execute anything, just how many operations it can retire per clock.
Running a single instruction stream (e.g. single threaded application), the Core 2 can process as many as four operations per clock, since it can issue 4-instructions per clock and it isn't execution unit constrained. The 10-core design however can only issue two instructions per clock and thus the peak execution rate for a single instruction stream is two operations per clock, half the throughput of the Core 2. That's fine however since you'll actually want to be running vector operations on this core and leave your single threaded tasks to your Core 2 CPU anyways, and here's where the proposed architecture spreads its wings.
With two cores, each with their ability to execute 4 concurrent SSE operations per clock, you've got a throughput of 8 ops per clock on Core 2. On the 10-core design? 160 ops per clock, an increase of 20x in roughly the same die area and power budget.
On paper this could actually work. If you had enough of these cores, you could get the vector throughput necessary to actually build a reasonable GPU. Of course there are issues like adapting the x86 instruction set for use in a GPU, getting all of the cores to communicate with one another and actually keeping all of these execution resources busy - but this design experiment showed that it was possible.
Thus Larrabee was born.
Facebook
Twitter
RSS
101 Comments
View All Comments
Griswold - Monday, August 4, 2008 - link
You seem to be confused. Time for a nap.MDme - Monday, August 4, 2008 - link
but AMD will have Cinema 2.0. did you see that demo? by 2010, AMD will have the RV990 or whatever...and Nvidia will have GT400?phaxmohdem - Monday, August 4, 2008 - link
Considering how long it took nVidia to release a single GPU significantly faster than G80, I'd be shocked if we wee GT300 by 2009/2010. however a GTX 295GT X2 ULTRA OC is not out of the question ;)shuffle2 - Monday, August 4, 2008 - link
mm², how hard is that to write? >.>1prophet - Monday, August 4, 2008 - link
They need to hit one out of the park with the drivers (software)as well.jltate - Tuesday, August 5, 2008 - link
I've got a bunch of comments, so I'll just list them all here.SSE doesn't have fused multiply-add operations. Larrabee does -- thus that 10 core processor could perform a peak of 320 floating point operations per cycle (it's mentioned in the SIGGRAPH paper).
Larrabee's programming model is variable width -- the hardware can and likely will be augmented in the future to perform more than just 16 operations in parallel.
The ring bus between cores was stated to be for each group of 16. Intel stated that for more than 16 cores they'd use "multiple short-linked rings".
Also, the diagram only shows one memory controller on one side with fixed function logic on the other, not two memory controllers as you showed on page 5 of your article. However, Intel stated in the paper that the configuration and number of processors, fixed function blocks and I/O controllers would be implementation dependent. So in effect it could very well have a half-dozen 64-bit interfaces like G80.
My forecast? This thing will rock. I for one simply cannot wait.
Laura Wilson - Monday, August 4, 2008 - link
that's the truththey say they know this. it sounds like they know this ... we'll see what happens :-)
gigahertz20 - Monday, August 4, 2008 - link
I'm going to predict Larrabee will provide a huge boost of performance over Intel's current crappy integrated graphic solutions, but will not be able to compete with AMD/ATI's and Nvidia's high end GPU's when it (Larrabee) finally launches. If Intel can deliver a monster that can push 100+ FPS in Crysis and doesn't cost so much that it breaks the bank like the current Nvidia GTX 280's, then they will have a real winner! When it finally launches though, who knows what AMD/ATI and Nvidia will have out to compete against it, wonder if Intel is just trying to push out a mainstream chip or go high end as well...guess I need to read the rest of the article :)JEDIYoda - Tuesday, August 5, 2008 - link
dreaming again huh??? you people who want top notch performance without having to pay for it....rofl..hahahaFITCamaro - Monday, August 4, 2008 - link
This isn't mean to compete with their IGPs. At least not initially.