nVidia's GT300 is "smaller, faster than Larrabee"?

[Ancient76]

http://www.brightsideofnews.com/news/2009/5/12/nvidias-gt300-is-smaller2c-faster-than-larrabee.aspx

It looks like nVidia is dead set on calming the PR storm the competitors are releasing, and the mood in the company is more relaxed. nVidia is certain they have a winner on their hands, but only time will tell whether they managed to pull it through or not.

We already gave you details on the architectural side. GT300 or G300 or NV70 is consisted out of 512 cores [or Shader Processors, whatever you like], features a 512-bit memory controller connecting to GDDR5 memory and well, it is a beast. nVidia packed everything in around 2.4 billion transistors using TSMC's 40nm high-performance process [yes, the one with currently sucky yields]. Given that the original GT200 packed 240 shaders in 1.4 billion trannies, one could argue that nVidia did some heavy dieting on its own architecture and managed to get a body builder shader power in a size zero body. GT300 packs 512 MIMD-capable cores and yet it uses "just" one billion transistors extra. I'll be first to admit that I wondered how GT300 packs at least three billion transistors, but according to our highly confidential source, the 2.4 billion transistors are packed in just 495mm2.

Yes, you've read that correctly. 2.4 billion in less than 500mm2 will put sweat on both ATI and Intel's forehead, since this chip could be profitably manufactured and yet pack performance to potentially blew the competition out of the water. Each of original 65nm GT200 chips took 576mm2 of wafer space, while 55nm refresh GT206/GT200b eats up 490mm2.

In comparison, Intel's high-end Larrabee part is manufactured in 45nm and takes around 600mm2.

Now, if you are wondering why nVidia created a single-PCB GTX295, think again. This part is more than an engineering exercise - but let us put it in this perspective - if nVidia managed to put 2.4 billion transistors in less than 500mm2, with an interesting performance delta between GT200-GT300 in clock-per-clock comparison… we could say that Jen-Hsun definitely opened a can of whoop-ass on its competitors.

http://www.hardware-infos.com/news.php?news=2944&sprache=1

http://resources.vr-zone.com/image_deposit/up2/1242158238e28c529255.jpg

[CaptNKILL]

I just peed a little bit.

[andyr3]

Larrabee is going to be huge, so the "smaller... than Larrabee" is no surprise. Faster? lets see.

[Dreamweavernoob]

This...................is good news :D

[Ninja Prime]

Q4 2009, look like the delay rumors were true.

Also, the article is a little deceptive, as we don't really know how big larrabee is or what process it will debut on(at least that I've seen). Not to mention, by the time the GT300 is slated to come up, Intel is supposed to have moved to a 32nm process, so the whole "lol its smaller" isn't really valid simply because its on a smaller process, because Intel will overtake them in process size like a month or two later. I also like the "512 cores" BS, I guess they never got over it, since larrabee is doing the same work with 32 cores. Efficency much? ;)

[AthlonXP1800]

2.4 billion transistors... HOLY CRAP! :afraid:

[walterman]

... I also like the "512 cores" BS, I guess they never got over it, since larrabee is doing the same work with 32 cores. Efficency much? ;)

nVidia calls core to a single scalar FPU unit (basically). 128bit SSE is like 4 single scalar FPU units. Larrabee with 512bit SSE units is like 16 single scalar FPU units, thus, 32 * 16 = 512 FPUs. They are on par.

In fact, since the core2, intel can exec up to three 128bit SSE instructions per clock, so, a quad core could be seen as a 4 * 4 * 3 = 48 'cores' cpu.

http://www.tomshardware.com/reviews/Intel-i7-nehalem-cpu,2041-5.html

Never apply gpu terms to the cpu world :)

[josiahsuarez]

LOL, this is the polar opposite of the GT300 story The Inquirer just posted. I won't bother linking to it because it's just a lot of hyperbole. Anyway, I thought the consensus was that the initial implementation of Larrabee won't be competitive in the high end space anyway. And the wafer Gelsinger showed at IDF had folks estimating >600mm2 die size for Larrabee, so I can believe both the faster and smaller claims/speculation.

[Sazar]

Speculation is good (popcorn)

/me predicts the next rumour states that the next-gen cards will destroy the Borg.

(nana2)

[SH64]

GT400 has a mini-death star in it :cool:

[Ninja Prime]

nVidia calls core to a single scalar FPU unit (basically). 128bit SSE is like 4 single scalar FPU units. Larrabee with 512bit SSE units is like 16 single scalar FPU units, thus, 32 * 16 = 512 FPUs. They are on par.

In fact, since the core2, intel can exec up to three 128bit SSE instructions per clock, so, a quad core could be seen as a 4 * 4 * 3 = 48 'cores' cpu.

http://www.tomshardware.com/reviews/Intel-i7-nehalem-cpu,2041-5.html

Never apply gpu terms to the cpu world :)

Heh, exactly my point. ;)

[Ancient76]

NVIDIA Talks About GT300, 40nm, Ion and Tegra.

http://news.softpedia.com/news/Exclusive-NVIDIA-Talks-About-GT300-40nm-Ion-and-Tegra-114087.shtml

Putting that into perspective, can we think about a future without x86?

I.S: I don’t think anybody is saying x86 is out of the picture, but CPUs are losing their breath and GPUs becoming much more important. It’s very simple. Just check how much time you need for transcoding of video on ION and how powerful CPU you need to match ION speed. I think that future is in well balanced PC with CPU and GPU living in harmony. CPU will have more control tasks in operating system and GPU will be focused for most of visual computing staff including some heavy computational tasks. This is the architecture of the future. I believe that PC with 1000 USD CPU and Intel integrated graphics card is already history.

I hope so!

[Heinz68]

NVIDIA Talks About GT300, 40nm, Ion and Tegra.

http://news.softpedia.com/news/Exclusive-NVIDIA-Talks-About-GT300-40nm-Ion-and-Tegra-114087.shtml
I hope so!
Can you tell us something about the alleged GT300 that has been rumored in the media recently?

I.S: I am not aware of that code name.
:D

[sethk]

I think there is very, very, very little chance that Larrabee will be faster than GT300 in existing games out the gate for so many reasons.

One is the SIMD computational throughput of the x86 uarch based Larrabee on all the preliminary information is not in the same ballpark as the GT300 or the R800.

Two, Intel has shown us that they are not exactly graphics driver mavens, and will probably not have good drivers or the same aggressive driver update schedule that Nvidia has.

Three, the huge library of DX10 and DX9 games that Nvidia has provided optimizations for and will support on day 1 will give them an edge in all review comparisons (yes the much maligned TWIMTBP program has actually made a difference).

Four, SLI. Nvidia has it, and Intel has not announced a similar driver level technology, although they will probably get there.

Finally, I think Intel won't launch on time. I think by the time they release a real enthusiast grade part, they'll be competing with die shrunk and up-clocked GT300 derivatives.

[CaptNKILL]

I almost forgot about Larrabee. :lol:

Has there been any news about it in the last few months?

[josiahsuarez]

I almost forgot about Larrabee. :lol:

Has there been any news about it in the last few months?

you know who made a post about it on his new site. unfortunately though it's basically just a verbose way of saying Larrabee has some hardware similar to Avivo or PureVideo.

http://www.semiaccurate.com/2009/07/21/larrabee-has-two-hd-decoders/
Larrabee has two HD decoders
by Charlie Demerjian

July 21, 2009

WORD HAS REACHED our tender ears that Larrabee, the upcoming Intel GPU, will not be quite as generalist as they claim. The fixed function parts will be two HD decoders.

This addition seems like a last minute thing, so it is probably pulled from Intel's existing IP catalog. The most likely candidate seems to be the G45 decoders, they are small, low power, and have (finally) decent drivers. Then again, does area really matter in a 700+mm^2 chip?

Don't take this to mean Larrabee is underpowered, it most definitely is not. The GPU can decode HD streams until you get bored watching them, the problem is that it can't necessarily do so while staying in an acceptable power budget. Having all the cores pump out HD would likely produce enough heat to keep the fans spun way up, and that is death in the HTPC market.

Why anyone would want the first generation Larrabee for an HTPC rig is an open question, one that brings us to how Larrabee will evolve. The first one can be considered a hybrid between a software development platform and a science experiment. Intel aims to do nothing less than change how graphics are done, a laudable goal that will take years. During that time, avenues will be explored, concepts implemented, and others shelved.

The end goal is to have a CPU that can do what a CPU does now in addition all the GPU tasks. Both AMD and Intel are driving to the same end point along very different highways, but will almost assuredly end up at the same point.

Getting back to power, Larrabee 1 was the first of it's kind, so once silicon came back, there were obvious things that needed doing on the next version, and the one after that. Intel is going for a fast cadence to roll out new Larrabee chips, with Larrabee 2 coming out about 9 months after 1, and number 3 about 9 months after that. There is an obvious comparison here that we won't stoop to.

Intel has a 12 month CPU cadence that goes new core -> shrink/optimize -> new core -> shrink/optimize. This is mainly dictated by a cross between the 24 month silicon shrink cycle and common sense. With Larrabee on a 9 month cadence, things can get a little odd.

Luckily, a lot of Larrabee is not all that complex. 32 cores across 7xx mm^2, minus a bit for cache, bus and memory controller leaves you with far sub 20mm^2 cores. Those cores can be revved with a lot less effort than a Nehalem core, then cut and pasted. The uncore can simply be left alone for the most part, or more likely revved separately. This is the same thought process that AMD pioneered with the K8, variants were relatively easy to pull off.

So, what you will see with Larrabee is initially to get it out there, then optimize for Larrabee 2. A 32nm version will lose half the area right off the bat from the shrink. Then Intel will unleash the optimization elves from the caves under Hillsboro (After dark of course......), and you could be looking at a second generation that is closer to 250mm^2 than 500.

Power is a little more problematic, purpose specific logic will always be more efficient than generalist units, and that is where Larrabee gets caught up. You have a big, hot chip that isn't optimized for much out of the gate. Running video on that takes wattage, and wattage is the enemy of thin and quiet PCs.

We have heard that the initial Larrabees were so power hungry that the clocks needed to be dialed back a bit, GT200 style, to get into a rational PC power envelopes. At those power levels, even using a fraction of the GPU to decode movies breaks the 'quiet' budget, so two HD decoders were added.

Given the rapid pace of optimizations that are already coming for the Larrabee line, we don't expect the decoders to remain in silicon for long. If they are there in the second revision, that will be somewhat surprising, but they are almost guaranteed to be gone by the third.

[jolle]

if running video decoding would have the chip running at full blast, then what about something like Aero Glass on Vista or Win7? that stuff uses D3d and runs on GPU, so shouldnt that mean your running flat out pretty much 100% of the time?
With the power consumption, heat and noise associated with that.

[FastRedPonyCar]

I heard the new GT300's are nothing more than old TNT2's that Chuck Norris is rubbing with his beard at the factory...

It would make perfect sense.

[jolle]

I heard the new GT300's are nothing more than old TNT2's that Chuck Norris is rubbing with his beard at the factory...

It would make perfect sense.
Then both ATi and Intel are boned, if their parts havent been touched by Chuck, what chance do they have?

[CaptNKILL]

Reminds me of the fake CPU-Z image I made years ago...

http://www.nvnews.net/vbulletin/attachment.php?attachmentid=37629&stc=1&d=1248378108

[jolle]

4 petachucks AND chuck +?
clearly, that cannot be.