Intel G965 to support SM4.0?

But have you used refrast lately? Even a GMA900 should beat refrast (on any current CPU) in nearly every conceivable situation by at least an order of magnitude or more, assuming that the GMA900 doesn't refuse to run the application outright due to its horrible driver.
 
Chances are high that a SGX based IGP could have much higher potential, but either way an IGP is an IGP; meaning lowest level of performance anyway.

Question mark remains what IMG exactly meant in it's PRs when they stated computing devices in connection with SGX and Intel.
 
GMA950 is already 4 pipelines (a full quad). But as expected, though, it performs worst than ATI or NVidia 2 pipeline (half a quad) IGPs. So G965 should be at least 4 pipelines (with the kind of bw avalaible I don't see what will be the use of additional pipelines).
 
IMO, all the new Intel IGP needs is to be able to support Aero Glass and also have some basic hardware acceleration for h.264/WMV9 decode.

This would pretty much match necessary requirements for a VIIV system which I think Intel wants to push so I wouldn't expect that they will be looking to add much else. The dual core chips already have about enough performance to decode h.264 at 1080p I believe so the video acceleration capabilities wouldn't even need to be all that comprehensive.
 
RoOoBo said:
GMA950 is already 4 pipelines (a full quad). But as expected, though, it performs worst than ATI or NVidia 2 pipeline (half a quad) IGPs. So G965 should be at least 4 pipelines (with the kind of bw avalaible I don't see what will be the use of additional pipelines).
Click to expand...

yep.

for the record, 950 easily achieves 1.3GPix/sec untextured out of its theoretical 1.6GPix on an SMP system with dual channel mem.

throw in a single texture, though, and you get a clock penaly regardless whether the whole tex is in the cache.

on the bright side it does a dot3 per clock (i have yet to check how it co-isses, though)
 
Demirug said:
Sorry but I can’t follow you. Aero use the D3D9 interface and the D3D9 part of a driver.
Click to expand...

Yes, but on a D3D10 class device, the D3D9 driver is probably having the standard D3D9 pipeline "emulated" via D3D10 features. (similar to how the old OpenGL pipeline is 'emulated' on programmable hardware by appropriate shaders)

The have. It’s called WinSAT and my current System has an overall rating of 3. My 6800 GT is rated with 5.9 for the performance and 4.7 for the 249 MB usable graphics RAM.
Click to expand...

Sounds like a good start, although I'd be curious how they "evaluate" the performance of D3D10 HW, as it seems they are many factors of performance to test, and weighting the number appropriately could be tricky. As a developer, you might want a minimal level of GS performance, for example.

I am not sure if it is logarithmic but at least there is a small set of numbers.[/QUOTE]
 
DemoCoder said:
Yes, but on a D3D10 class device, the D3D9 driver is probably having the standard D3D9 pipeline "emulated" via D3D10 features. (similar to how the old OpenGL pipeline is 'emulated' on programmable hardware by appropriate shaders)
Click to expand...

On the driver level this is not a good idea. It would be better to build a universal driver that contains anything from D3D9 and D3D10 and use small layers that let the runtime connect.

Mapping one version of Direct3D to another is not always that easy. D3D8 to D3D9 was easy to do. D3D7 to D3D8/D3D9 is much more complex. D3D9 to D3D10 is a very funny job because they have changed nearly anything.
 
Uttar said:
I'd be VERY surprised if GMA965 was SM4.0. - everything I know points towards it being SM3.0. without "optional" features. It's a kind of "minimum minimora" for Intel, in order to get something sufficient for Vista without having to pay licensing fees (=> royalties) for IMG/PVR IP. Obviously, having to pay royalties to other companies is not acceptable in some of Intel's productlines.

Uttar
Click to expand...

if they stick to a DX9 design for simplicity and Aero glass compatibility, wouldn't they stay with bare SM 2.0 ?
 
DemoCoder said:
No where was I suggesting a "wrapper" approach in my message.
Click to expand...

The problem with the D3D10 features is that in some cases D3D10 is more limited as D3D9. If a driver use only D3D10 features to execute D3D9 applications it will need some kind of wrapper. That’s the reason why I write about a more universal internal driver interface.
 
I was at E3 talking to Mark Rein, when my co-worked asked "I wonder how Intel is going to handle DX10." Just then Tim Sweeney was walking by, so Mark says to him "hey Tim, how is Intel going to handle DX10?"

Tim snorted, shook his head, and just said "badly."

Now maybe he had some experience with their new IGP, maybe he was just going off Intel's less-than-stellar reputation for integrated graphics. Who knows. It made me laugh, though.
 
I've seen elsewhere unified shader units being mentioned for the 965; now assume it is true and tell me how many "whatexactly" it could have. For one measuring such a chip with "quads" would be nonsense.
 
JasonCross said:
I was at E3 talking to Mark Rein, when my co-worked asked "I wonder how Intel is going to handle DX10." Just then Tim Sweeney was walking by, so Mark says to him "hey Tim, how is Intel going to handle DX10?"

Tim snorted, shook his head, and just said "badly."

Now maybe he had some experience with their new IGP, maybe he was just going off Intel's less-than-stellar reputation for integrated graphics. Who knows. It made me laugh, though.
Click to expand...

When was that? Since this year's E3 hasn't started yet i suppose you are referring to last years event?
 
good

These are good news. This chip will be very cheap and hope it will be available when Microsoft decides to open to the public the Windows Vista Beta 2 and use DX10. At the moment to work with the reference rasterizer to test is a pain, so motherboards with it will be very good, specially if they come before the GeForce8 and the ATI2000.
 
Uttar said:
I'd be VERY surprised if GMA965 was SM4.0. - everything I know points towards it being SM3.0. without "optional" features. It's a kind of "minimum minimora" for Intel, in order to get something sufficient for Vista without having to pay licensing fees (=> royalties) for IMG/PVR IP. Obviously, having to pay royalties to other companies is not acceptable in some of Intel's productlines.

Uttar
Click to expand...

I can assure you that it is a fully unified chip.
 
Does this have powervr sgx/series 5 written all over it? I do not believe Intel have made made these big improvements to the G965 in house but what say others?

According to http://www.answers.com/topic/powervr:

"Trivia: the Intel 915-/945G chipsets use PowerVR technology licensed from ST Microelectronics"

Okay this isn't hard evidence but Intel's "Zone Rendering" in the 915/945 always looked uncannily similar to powervr's technology to me and the improvements made to the G965 chip go beyond the Stmicro' series 4 license so does this mean Intel have are using their license from Imagination Technology for eurasia aka SGX/series 5?

Eldar
 
Thanks, yes I noticed it had been removed after I made my post.

Doesn't alter the fact that Intel's "extreme gaming' chipsets won't be entirely in house designs IMO and the leaked spec's for the Intel G965 look uncannily like they have utilised series 5/SGX to me.

Kristof, care to make a no comment? ;)

Eldar
 
You must log in or register to reply here.

Similar threads

R
The Unreleased CPU Thread
Replies
10
Views
8K
kalelovil
K
A
Holy Wall of Text: My impressions post Playstation 4 (PS4) unveil
2 3 4
Replies
70
Views
20K
(((interference)))
I
A
The Official IGP Rumours & Speculation Thread
2 3 4
Replies
64
Views
12K
Ailuros
Ailuros
S
Intel launched Clear-Video to against AVIVO and PureVideo
Replies
2
Views
2K
asicnewbie
A
M
Nvidia Ageia acquisition - console edition
Replies
17
Views
5K
sevanig
S
Back
Top