Slashdot Mirror
Nvidia Claims Intel's Larrabee Is "a GPU From 2006"
Barence sends this excerpt from PC Pro:
"Nvidia has delivered a scathing criticism of Intel's Larrabee, dismissing the multi-core CPU/GPU as wishful thinking — while admitting it needs to catch up with AMD's current Radeon graphics cards. 'Intel is not a stupid company,' conceded John Mottram, chief architect for the company's GT200 core. 'They've put a lot of people behind this, so clearly they believe it's viable. But the products on our roadmap are competitive to this thing as they've painted it. And the reality is going to fall short of the optimistic way they've painted it. As [blogger and CPU architect] Peter Glaskowsky said, the "large" Larrabee in 2010 will have roughly the same performance as a 2006 GPU from Nvidia or ATI.' Speaking ahead of the opening of the annual NVISION expo on Monday, he also admitted Nvidia 'underestimated ATI with respect to their product.'"
Good, learn from that and don't make that same mistake again!
Larrabee [...] will have roughly the same performance as a 2006 GPU from Nvidia or ATI.'
DOH!
Belief is the currency of delusion.
No wonder it's so slow. He keeps making reference to how it paints things. Can't move on to another frame until the previous one has dried.
So why is NVIDIA on the defensive?
Intel is aiming at number crunchers (note that their chip uses doubles, not floats). They don't want NVIDIA to steal that market with CUDA.
When Intel says "graphics", they mean movie studios, etc.
If Larrabee eventually turns into a competitor for NVIDIA, all well and good, but that's not their goal at the moment.
No sig today...
At least Intel documents their hardware. Fuck NVIDIA and their stupid proprietary hardware!
Glass
"OH MY GOD! CPU AND GPU ON ONE DIE IS STOOOOOOOOPIIIIIDDDDDEDEDDDD!!!1111oneoneone"
How stupid is it really? So what if the average consumer actually knows very little about their PC. That doesn't necessarily mean it won't be put into a person's PC.
If they were really forward thinking, they could see it as an effort to bridge the gap between low-end PC's and high-end PC's. Now maybe, at some point in the future, people can do gaming a little better on those PC's.
Instead of games being nigh unplayable, are now running slightly more smoothly. With advance in this design, it could really work out better.
Sure, for the time being, I don't doubt that the obvious choice would be to have a discrete component solution for gaming. However, there might be a point where that isn't in the gamers best interests anymore. I'm not a soothsayer, I don't know.
Still, I can't only help but imagine how Intel's and AMD's ideas can only help everyone as a whole.
A recent journal article on ArsTechnica points to an Intel blog on Larrabee: http://arstechnica.com/journals/hardware.ars/2008/05/01/larrabee-engineer-on-personal-blog-larrabee-is-all-about-rasterization Curious.
Putting syrup in coffee is some form of blasphemy.
Ten years ago you would see Nvidia GPUs in everything from low- to high-end. Today, not so much - Intel dominates the low-end spectrum, with ATI hanging onto a somewhat insignificant market share. The Larrabee is Intel moving upmarket. Sure, it might not perform as well the latest Nvidia or ATI high-end GPU but it might be enough in terms of performance or have other benefits (better OSS support) to win some of Nvidia's current market share over. Considering it's supposedly the Pentium architecture recycled, it's also reasonable to assume the design will be relatively cost-effective and allow Intel to sell at very competitive prices while still maintaining healthy profit margins.
It's a classic case of disruption. Intel enters and Nvidia is happy to leave because there's a segment above that's much more attractive to pursue. Continue along the same lines until there's nowhere for Nvidia to run, at which point the game ends - circle of disruption complete. See also Silicon Graphics, Nvidia's predecessor in many ways.
If, in the future, the trend evolves that all gpu's are integrated.
Intel, nvidia, AMD and ATI...
Who is the odd one out there?
I record my sleeptalking
Lots of people here and analysts have written off AMD. I think AMD is in a great position if they can survive their short term debt problems which is looking increasingly likely.
Consider the following:
AMD is in a great position like no other company to capitalize on the coming CPU / GPU convergence. Everyone jeered when AMD bought ATI but it is looking to be a great strategic move if they can execute on their strategy.
AMD has the best mix of technology, they just have to put it to good use.
G. Washington on Government "it is force. Like fire, it is a dangerous servant and a fearful master."
Company says competitor's product sucks! News at 11.
From the SIGGRAPH paper they need something like 25 cores to run GoW at 60Hz. That's 1Ghz cores for comparison though. LRB will probably run at something like 3Ghz, meaning you only need like 8-9 cores to run GoW at 60, and with benchmarks stretching up to 48 cores you can see that this has the potential of being very fast indeed.
More importantly, the LRB has much better utilization since there aren't any fixed function divisions in the hardware. E.g. most of the time you're not using the blend units. So why have all that hardware for doing floating point maths in the blending units when 99% of the time you're not actually using it? On LRB everything is utilized all the time. Blending, interpolation, stencil/alpha testing etc. is all done using the same functionality, meaning that when you turn something off (like blending) you get better performance rather than just leaving parts of your chip idle.
I'd also like to point out that having a software pipeline means faster iteration, meaning that they have a huge opportunity to simply out-optimize nvidida and amd, even for the D3D/OGL pipelines.
Furthermore, imagine intel suppyling half a dozen "profiles" for their pipeline where they optimize for various scenarios (e.g. deferred rendering, shadow volume heavy rendering, etc. etc.). The user can then try each with their games and run each game with a slightly different profile. More importantly, however, is that new games could just spend 30 minutes figuring out which profile suits them best, set a flag in the registry somewhere, and automatically get a big boost on LRB cards. That's a tiny amount of work to get LRB-specific performance wins.
The next step in LRB-specific optimizations is to allow developers to essentially set up a LRB-config file for their title with lots of variables and tuning (remember that LRB uses a JIT compiled inner-loop that combines the setup, tests, pixel shader etc.). This would again be a very simple thing to do (and intel would probably do it for you if your title is high profile enough), and could potentially give you a massive win.
And then of course the next step after that is LRB-specific code. I.e. you write stuff outside D3D/OGL to leverage the LRB specifically. This probably won't happen for many games, but you only need to convince Tim Sweeney and Carmack to do it, and then most of the high profile games will benefit automatically (through licensing). My guess is that you don't need to do much convincing. I'm a graphcis programmer myself and I'm gagging to get my hands on one of these chips! If/when we do I'll be at work on weekends and holidays coding up cool tech for it. I'd be surprised if Sweeney/Carmack aren't the same.
I think LRB can be plenty competitive with nvidia and amd using the standard pipelines, and there's a very appealing low-fricion path for developers to take to leverage the LRB specifically with varying degrees of effort.
but the extra programability larrabee have as its just a bunch of cpus with some gpu instructions
Agreed -- why stick to GPU applications, when you have a general purpose multicore machine? How about getting those new instructions into general usage -- remember how MMX was originally introduced for stuff we now run on GPUs.
As for traditional GPU applications, there's already an OpenGL driver for the Cell SPUs in development. A similar driver for a generic multicore machine would be nice, particularly if it's not limited to Larrabee and x86. Of course we already have software implementations of OpenGL, but I wonder how well those scale with dozens of CPUs.
Escher was the first MC and Giger invented the HR department.
With all OpenGL extensions supported working properly, to latest and greatest from NVIDIA where I can never be sure which extension work on which driver with which card.
So intel will only be about 4 years behind current in their graphics system when it comes out.
In that case, it's probably the biggest leap they'll have ever made.
The nVidia drivers are binary only, so they are not available in the standard source repositories and are not compiled and included by default in most opensource distribution.
Ubuntu has made the necessary arrangement and provides, out-of-the-box a tool that can automatically download and install binary drivers from within the usual setup tool.
It think that's why the parent poster may refer to.
That means that, instead of having to manually download a package and execute it (from the command line) - which isn't complicated but require some interaction with the computer - installing a binary driver under Ubuntu simply means clicking the button "yes" on a dialog asking "the following hardware requires non-free proprietary driver, would you like to install them".
It's made trivial enough so computer non-litterate users can still do it easily - well, almost. The users still need to think that maybe they should get some software to make the graphics work better.
Behind the scene, cliquing "yes" automatically add the non-free drivers repository to apt-get and selects the necessary package for installation.
The results are similar (although differently implemented) to opensuse's one-click install (where you click on a link in a web page to a file with name ending in ".ymp") and the corresponding repositories are added to YaST and packages selected for installation.
"Sufficiently advanced satire is indistinguishable from reality." - [Tips: 1DrYakQDKCQ6y52z6QbnkxHXAocMZJE61o ]
AMD is missing someone to develop and manufacture good motherboard chipsets.
Haven't been following the news recently ?!?
ATI/AMD latest serie of chipsets (the 790) is quite good. That's the reason why VIA announced dropping that market in the first place.
The only problem is that currently, nVidia's SLI is a proprietary technology requiring licensing. So that's why a lot of player still buy nvidia's chipsets and avoid ATI's - not that these are bad, on the contrary, but they only lack the license required for SLI.
This SLI problem is also explaining why nVidia may have to consider stopping producing intel chipset : They never licensed their SLI technology to Intel to have SLI-compatible Intel-made chipsets. (Either forcing gamers to use nVidia chipsets or requiring convoluted hacks with SLI chipsets acting as bridges between the main northbridge and the GPUs as in Skulltrail).
And Intel is now retaliating by refusing nVidia access to QuickPath Interconnect.
So either nVidia will have to drop the Intel chipset market (and only produce SLI-bridge like in the Skulltrail hack).
Or nVidia will have to give possibility to license SLI, and thus lose an interesting market that they had managed to lock.
Hence the rumors you mention.
Nonetheless they aren't going to stop producing chipsets for AMD (still popular among gamer) nor for VIA (they have even announced new chipset able to play DX10 games and Vista in all its aero glory on VIA Isaiah ITX platforms)
"Sufficiently advanced satire is indistinguishable from reality." - [Tips: 1DrYakQDKCQ6y52z6QbnkxHXAocMZJE61o ]
Didn't the 8800 series come out at the end of 2006? The first gen 8800GTS 640MB and the 8800GTX 768MB those are still powerful video cards by today's standards.... so if Larrabee is "a GPU from 2006" then isn't that a compliment to Intel?
As Moore's law makes silicon cheaper, what are you going to do with it? more cache, more cores... why not a GPU? Concurrent software to utilize multi-cores is not yet mainstream (maybe never), so that leaves cache and GPU.
In a way, the existence of separate GPUs is just a sign that the CPU wasn't powerful enough to deliver the graphics the market wanted (and would pay for). When CPU's are powerful enough (clock speed or multi-core), they'll subsume the GPU, as they did maths co-processors and cache. ie. The silicon would be partitioned into CPU, GPU and cache - but it would all be on the one chip (called the "CPU" no doubt).
Intel already owns a fair bit of the integrated graphics market. They have great access to channels. Even if this is only half as good as a separate GPU, they will increase market share. I can't see what could stop them... except maybe a patented technology that can't be worked around. Some manufacturers of separate GPU's will survive in specialized niches. Some.
What chips? Intel hasn't demonstrated any Larrabee hardware yet. They've published some specs, but we don't even know how many cores it will have or their clock speed.
Give a man fire, and you warm him for the night. Set a man on fire, and you warm him for the rest of his life.
Although I appreciate the attention from NVIDIA and Slashdot, I can't support that alleged quote from my blog (http://speedsnfeeds.com).
First, what's being described as a quote is actually just John Montrym's summary from my original post, which is here:
http://news.cnet.com/8301-13512_3-10006184-23.html
What I actually described as equating to "the performance of a 2006-vintage... graphics chip" was a performance standard defined by Intel itself-- running the game F.E.A.R. at 60 fps in 1,600 x 1,200-pixel resolution with four-sample antialiasing.
Intel used this figure for some comparisons of rendering performance. If Larrabee ran at 1 GHz, for example, Intel's figures show that it would take somewhere from 7 to 25 Larrabee cores to reach that 60 Hz frame rate.
Larrabee will probably run much faster than that, at least on desktop variants.
Well... rather than writing the whole response here, I think I'd rather write it up for my blog and publish it there. Please surf on over and check it out:
http://news.cnet.com/8301-13512_3-10024280-23.html
Comments are welcome here or there.
. png
For something that was back in the early K8 days, it seems like 99% of the boards on the market today for AMD CPUs have nvidia/via chipsets.
And since Phenom and AM2+ socket appeared, 99% of the boards on the market for these use nvidia/ati chipset.
The few VIA based motherboards you can see usually are based on derivative of the KT800 chipset that was already available back in the early K8 days (as the memory controller in on the CPU and the chipset only communicates using HyperTransport - one can pretty much mix'n'mach most chipset almost regardless of the processor generation).
And these mainboards are targeted to the budget segment (usually feature only a couple of slots, and sometimes integrated graphics).
All the high-end boards are nvidia or ati based.
The ATI are specially popular in research because they provide 4 long PCIe slots (16x physical, usually 8x bandwith when all 4 in use), often in altening succession (one PCIe 16 each to slot) enabling scientist to put 4 dual-slot cards for GPGPU (CUDA or Brook)
I'm not really seeing a lot of boards on the market (especially carried in stores) that use the AMD chipset.
I don't know, maybe the few stores you checked either carry only old (pre-Phenom) motherboard or sell more nvidia-based because they are popular because of the SLI support.
But most on-line shop I use have both nvidia and ati based motherboards.
"Sufficiently advanced satire is indistinguishable from reality." - [Tips: 1DrYakQDKCQ6y52z6QbnkxHXAocMZJE61o ]
I'm fixated on what engineers use their computers for.
I design things all day, and all I've got, all I need, is an ancient Intel 865 video chipset built into the motherboard of my Dell Optiplex.
I don't want or need a GPU, neither does anyone else in our department.
It's been a long time.