3

If I use a discreet GPU with Sandy Bridge, would Sandy Bridge's on-die GPU be sitting idle and be completely wasted?

Or would the on-die GPU somehow work with the discreet GPU and still be productive?

Hennes
  • 64,768
  • 7
  • 111
  • 168
Continuation
  • 155
  • 1
  • 4

2 Answers2

5

be sitting idle and be completely wasted?

It will not be in use. Just like most IGPs on current gen Core processors.

Sathyajith Bhat
  • 61,504
  • 38
  • 179
  • 264
  • Thanks. So all Sandy Bridge CPU's would be forcing me to waste a big chunk of money and silicon real estate on something I can't even use? This is a rip off! – Continuation Jan 03 '11 at 16:18
  • 2
    @Continuation: It's not going to make much difference to the price when you consider that the graphics is included in the mass-market chips and economies of scale. And it is the high-end LGA1356 amd LGA2011 and some of the Xeon LGA1155 chips that do not come with graphics. Some low end chips come with parts of the processor /disabled/, and low end mATX motherboards always had onboard graphics anyway. – paradroid Jan 03 '11 at 16:30
  • 3
    @Continuation - The i7 2600K is as fast as the i7 980-X and US$700 cheaper. I think you can have a little extra silicon you don't need... – Shinrai Jan 03 '11 at 17:43
  • @Shinrai - "i7 2600K is as fast as the i7 980-X." Not for highly threaded software. I run a lot of threaded apps for my workstation such as JVM, Lucene, MySQL, Erlang. I'd much rather have another CPU core than an on-die GPU which I don't want or need. – Continuation Jan 04 '11 at 14:57
  • 1
    @Continuation - Then you're in a boat that a lot of people aren't, but obviously if you can use the extra cores then go for it. Really, though, if your tasks are that highly threaded and CPU intensive you're better off with dual hex-core Xeons. :) I guess my point though is that it seems a bit silly to be to talking about the GPU being a wasted chunk of money when all the alternatives with similar performance levels are an order of magnitude more expensive. I might not want a moonroof in my Honda Civic but that doesn't mean the economical choice is a Maserati without one. – Shinrai Jan 04 '11 at 15:16
0

It depends on what you mean with a 'discrete GPU'.

If you add a dedicated graphics card with its own outputs and only connect you monitors to that then on-die graphics part of the Sandy Bridge will not be used.

If you add a dedicated graphics card with its own outputs and connect monitors to both the outputs of the dedicated graphics card and the motherboards (to CPU/GPU die) part then both will be used. This can be useful when you need more monitors.

If you have a laptop with switchable graphics (e.g. Nvidia optimus) then the onboard chip will always be used. The extra GPU chip can assist with faster calculations, but the actual output is always connected to the on CPU die part and thus it is always used. (Bringing that up since your post mentions a GPU and not a dedicated graphics card.

Lastly, with DX12 both GPU's (dedicated on on CPU die) can be used simultaniously.

Hennes
  • 64,768
  • 7
  • 111
  • 168