NVIDIA GeForce 6200 TC 64MB

The Technical Blahblah
Okay, so now we forget turbo and focus on the definiton of cache: optimize data transfers between system elements. We are slowly getting close to what TC is all about. In short TC is an efficient way for the videocard to use the system memory with the results that less memory in that graphics card is needed.

Now with "older" AGP solution that little trick was out of the question. That AGP solution might be able to handle 2 GB/sec data from the framebuffer but transferring it back and forward over the AGP bus, that speed would be reduced to 266 MByte/sec.

64 MB of 64-bit DDR memory from Elixer

Here's one of the new advantaged of a PCI-Express bus, as you know it is capable of managing 4 GB/sec up and downwards (bi-directional). That's a huge improvement and has a serious advantage over that AGP bus. So with a relatively small amount of onboard memory, the graphics card can claim 4x the amount of system memory for its usage. A 16 MB version can claim 64/128 MB system memory. The 32 MB, 128 MB and our 64 MB model can make use of 256 MB system memory. That "external" system memory can be used for a number of things, among them texture information but it can even render 3D images in system memory. Obviously you need at least 512 MB system memory in your PC.

Look at the specs and notice that the NV43 based "normal" GeForce 6200 had 143 million transistors, again cost control... this new 77 million transistor NV44 based product is a much cheaper to manufacture chip. The GeForce 6200 series product line has only 4 Pixel pipelines and 3 Vertex processors that we confirmed to be working 100%, let's open up the videocard's BIOS and see what it reports back:

$ffffffffff ---------------------------------------------------
$ffffffffff Display adapter information
$ffffffffff ---------------------------------------------------
$0000000000 Description : NVIDIA GeForce 6200 TurboCache(...
$0000000001 Vendor ID : 10de (NVIDIA)
$0000000002 Device ID : 0161
$0000000003 Location : bus 5, device 0, function 0
$0000000004 Bus type : PCI
$0000000009 Base address 0 : e0000000 (memory range)
$000000000a Base address 1 : d0000000 (memory range)
$000000000b Base address 2 : none
$000000000c Base address 3 : e1000000 (memory range)
$000000000d Base address 4 : none
$000000000e Base address 5 : none
$ffffffffff ---------------------------------------------------
$ffffffffff NVIDIA specific display adapter information
$ffffffffff ---------------------------------------------------
$0100000000 Graphics core : NV44 revision A2 (4x1,3vp)
$0100000001 Hardwired ID : 0160 (ROM strapped to 0161)
$0100000002 Memory bus : 64-bit
$0100000003 Memory type : DDR (RAM configuration 01)
$0100000004 Memory amount : 65536KB
$0100000005 Core clock : 384.750MHz
$0100000006 Memory clock : 210.375MHz (420.750MHz effective)
$0100000007 Reference clock : 27.000MHz

Hey now, Palit has altered reference speed of the graphics core, we noticed a faster 385 MHz core frequency and a lower 420 MHz memory frequency. Hmmm. We will just assume the sample we received is a retail sample. If you buy this product and clocks are lower at reference level, be sure to drop me an email.

What's rather nasty here is that the local framebuffer has been limited to 64-bit memory. Bandwidth is effectively cut in half. Then again, 64-bit memory is dirt cheap and that was the whole idea about this product.