NVIDIA GeForce GTX 480 Rev.A2 5009 Engineering Sample Obtained!

Some very nice ES cards you have there. Congrats.

From what I have found on the matter the GF100-A1 cards are extremely rare and don't work well at all. They also have a different cooler than the A2 cards so most likely the green pcb card you will receive is an GF100-A2 model. If you want a GF100 GPU with 512 Shaders the best chance to find one is with a GF100-A2 or GF100-A3 with the later being clocked at higher frequencies.
See here some examples of what can be found:

GF100-A2 512Shaders 384biti
[Blockierte Grafik: http://s7.postimg.org/4u5mulw2f/GTX480_ES37.jpg]
GF100-A3 512Shaders 320biti
[Blockierte Grafik: http://s2.postimg.org/795g7kn6d/GTX480_QS0042.jpg]

Cheers,
G1nX

Zitat von G1nX

From what I have found on the matter the GF100-A1 cards are extremely rare and don't work well at all.

NV had massive problems with it's memory controller, the manufacturing process (TSMC 40nm) and also within the design itself. I heared that nearly all the first silicon (GF100 Rev A1) were dead or simply die after a short period of time.

First real working silicon was A2, but it still had many major bugs in the design which also resulted in much heat dissipation / power consumption.

I'm sure there Nv had some fully enabled GF100-A3, but haven seen or heared of one. The only fully enabled GF100 are Rev A2. If you ask me, NV very early knew that they couldn't bring a full GF100-A3 to the market and that's why most 512core samples are Rev A2. Since NV had pressure to finally release something they had no time for another respin. That respin was called GF100B or GF100-B1 at first but became GF110.

Well, G1nX' screenshot does show an A3 with 512 shaders and a 320-bit memory interface right there.. I'm assuming that's not fake.

But if cards with 512 core GPUs are mostly A2, then it'd be a good thing for Obi to get another A2, hehe.

Zitat von GrandAdmiralThrawn

Well, G1nX' screenshot does show an A3 with 512 shaders and a 320-bit memory interface right there.. I'm assuming that's not fake.

Fully enabled I also mean memory interface

I'm nearly 100% sure about that, NV surely did performance tests with a 3 GiB GF100. Some GF110 custom cards do have 3GiB.

The Quadro 6000 even features 6GiB of memory

Zitat von Gold Leader

GiB is German for GB

No. Afaik, some crazy US-americans didn't know that 1 KB isn't 1000 byte but 1024 byte but the manufacturer calculate the capacity of a HDD with 1000. For example a capacity of 4 GB means 4.000.000.000 (4 billion) bytes (~3,725 GB) brutto and not 4.294.967.296 byte (4 GB). So manufacturers were sued by these stupid people and the they won. As a consequence the units KiB (KibiByte), MiB (MebiByte) and so on were created. These units are international and not a German invention

Zitat von CryptonNite

No. Afaik, some crazy US-americans didn't know that 1 KB isn't 1000 byte but 1024 byte but the manufacturer calculate the capacity of a HDD with 1000. For example a capacity of 4 GB means 4.000.000.000 (4 billion) bytes (~3,725 GB) brutto and not 4.294.967.296 byte (4 GB). So manufacturers were sued by these stupid people and the they won. As a consequence the units KiB (KibiByte), MiB (MebiByte) and so on were created. These units are international and not a German invention

Muhaahaha, Godlike

@GoldLeader:
Nice Card

True. This is now officially specified in the International Electrotechnical Commission's standard IEC 60027 to do away with the cheating that sometimes happens with the calculation of decimal and binary style prefixes...

They never did it "right", because - just like you - they learned, that SI prefixes (mega, giga, tera) mean something specific in regular sciences like mathematics or physics, but something different in computer science.

Basically, we in computer science were mostly wrong all the time to call 1024 bytes a "kilobyte". But it still sticks, so usually people just know that 1024 DRAM bytes are 1kB, but for harddrives, 1000 storage bytes are 1kB.

It is kind of wrong though, hence kiB, MiB, GiB etc.

They mix it up even worse in the above example, as they say:

1 gigabyte(gb)=1024 megabytes or ~1,000,000,000 bytes

That's a dangerous statement, they could at least say the "almost correct" thing:

1 gigabyte(gb)=1024 megabytes or 1.073.741.824 bytes

That's what we assumed was correct before the introduction of binary prefixes in IEC 60027.

Well:

1 KB = 2^10 bytes
1 MB = 2^20 bytes
1 GB = 2^30 bytes
1 TB = 2^40 bytes
1 PB = 2^50 bytes

So, 1 KB = 1024 bytes makes sense.

No, because you're counting binary, but SI is expecting decimal. So you can't do calculations to the power of 2 to determine the meaning of an SI prefix, that's the whole point.

1KB = 10^100 bytes (count decimal = SI)
1KiB = 2^1000 bytes (count binary = IEC 60027)

It's pretty new and not yet widely accepted, so people keep doing what they were always doing, wrong or not. For those of us who know where to count in which way it truly doesn't matter that much.

Plus (and that's my main point of criticizm) stuff like "Mebibyte", "Gibibyte" etc. just sounds fucking gay..

But there is one reason I do understand: When I was younger and not yet knowledgeable, I always wondered why that 540MB harddrive didn't give me freaking 540MB when formatted. It's because they counted decimally, whereas Windows counts binary.. even today!

Its a bit crappy, that a 4TB hardrive winds up being "3.6TB" in Windows. Just because of how they count and not tell you the differences. Back then when I was a kid, it took me a long time to figure that out and understand why HDDs are always smaller than what the manufacturer said they would be...

But yeah, just go ahead, amongst those who know, it doesn't matter.

When do you actually expect that second prototype to arrive?