AMD: Southern Islands (7*** series) Speculation/ Rumour Thread
- Thread starter UniversalTruth
- Start date
boxleitnerb
Regular
Apart from that:
Is it really clear if this hidden water is rendered in normal game mode, too? Or just in wireframe mode, since in wireframe mode there is no culling. At least that is what the devs over at Crytek said in their forums. So maybe this whole outcry over the water was for nothing...
Is it really clear if this hidden water is rendered in normal game mode, too? Or just in wireframe mode, since in wireframe mode there is no culling. At least that is what the devs over at Crytek said in their forums. So maybe this whole outcry over the water was for nothing...
Download the free Perfstudio (http://developer.amd.com/tools/graphics-development/gpu-perfstudio-2/) and see for yourself.
Incorrect. No one ever talked about wireframe mode...they used perfstudio to see what the application actually did.
Maybe your whole posts screams revisionism
Silent_Buddha
Legend
Hidden water is generally present in all titles sporting water. There's usually one global (for the level) layer of water. The level is then built upon and in that. This usually has no effect on rendering performance as the graphics card won't render the hidden water.
So, nothing weird was done with regards to the water. Over tessellating flat surfaces on the other hand (cement barricades, for example), has to make one wonder why they would do that.
Regards,
SB
OpenGL guy
Veteran
If the water is rendered first, how can the GPU discard it? Also note that even if there are some pixels that are rejected due to occlusion, all of the vertices have to be computed first.
boxleitnerb
Regular
Download the free Perfstudio (http://developer.amd.com/tools/graphics-development/gpu-perfstudio-2/) and see for yourself.
Incorrect. No one ever talked about wireframe mode...they used perfstudio to see what the application actually did.
Maybe your whole posts screams revisionism
Correct me if I'm wrong, but that water looks a whole lot like wire frame mode to me:
http://techreport.com/review/21404/crysis-2-tessellation-too-much-of-a-good-thing/3
From here:
http://www.crydev.net/viewtopic.php?p=889047
OpenGL guy
Veteran
Just because you can't see it doesn't mean it wasn't rendered. That comment is very vague.
Perfstudio doesn't enable wireframe mode in general. It shows you all of the draw calls in the frame and when a draw is selected with the frame debugger loaded it draws all of the objects prior to the selected draw call and draws the wireframe of the selected draw call.
Everything in the pic other than the water was drawn first. The fact that there's so much missing from the pic proves the water was drawn early. Perfstudio would say for sure, but it seems like the water was written to the frame buffer and overwritten later.
I'm more interested on per-pixel displacement mapping tesselation..
They are tessellating surfaces to slow down their per-pixel displacement mapping?
I'm more interested on per-pixel displacement mapping tesselation..
They are tessellating surfaces to slow down their per-pixel displacement mapping?
my limited knowledge stop there lol.
Last edited by a moderator:
I have a belated query.
How do you guys think a 28nm, 384-bit VLIW4 chip could compare to Tahiti in games? I remember here on B3D that VLIW scaling was considered a problem with Cayman and Cypress but I do wonder if it was really that much of an impasse.
How do you guys think a 28nm, 384-bit VLIW4 chip could compare to Tahiti in games? I remember here on B3D that VLIW scaling was considered a problem with Cayman and Cypress but I do wonder if it was really that much of an impasse.
There is more to the change between SI and previous than the compute cores, so things are not exactly like for like, but the easiest thing to do is take a Juniper and compare it against Verde at the same speeds. I believe a few people did that initially, and we did it as well, and I seem to remember a 10%-20% uplift in performance; that may be rose tinted glasses, but then I'd wager it may even be more now if someone conducted the same test on more recent drivers.
Awsome. Can't get much clearer than that
http://techreport.com/discussion/24368/fate-of-amd-sea-islands-obscured-in-the-fog
This clears things up completely.
This clears things up completely.
Considering how it went up to this point, it was clear that such clarity was going to be brought about - good thing it got cleared by management! Sometimes keeping quiet has its merits, I would reckon...that being said, there's only so much wiggle room when you are AMD and want to reach the OpEx levels that Rory Read is aiming for. Putting cash into bring-ups and the whole enchilada attached to a new product launch (that will probably screw you in the proverbial back, since you have quite a bit of accrued inventory) is not the way to go in that context, I guess. Of course it's not sexy to admit that
Frankly I don't know why Scott is taking such a stance - simple fact is that the roadmap continues unabated, unimpacted. As discussed the focus of Sea Islands is for notebook products first and was stated as such when we started talked about them at CES (though not everyone would have attended the breifings). We are still taping out and bringing up the same number of products irrespective of where they are focused or who (OEM notebook, OEM desktop, channel) is seeing them first. Obviously we're not going to give the farm away on the roadmap and product plans right now though.
Still, others seem to pick things up clearly:
http://www.anandtech.com/show/6751/amd-reiterates-2013-gpu-plans-sea-islands-beyond
http://www.pcper.com/news/Graphics-Cards/AMD-wants-set-record-straight-its-future-GPU-strategy
http://www.hardwarecanucks.com/news...y-new-cards-same-architecture-some-surprises/
Still, others seem to pick things up clearly:
http://www.anandtech.com/show/6751/amd-reiterates-2013-gpu-plans-sea-islands-beyond
http://www.pcper.com/news/Graphics-Cards/AMD-wants-set-record-straight-its-future-GPU-strategy
http://www.hardwarecanucks.com/news...y-new-cards-same-architecture-some-surprises/
I have to say that the anandtech article on the roadmap issues was a lot clearer. It seem like theTechReport is trying to be as transparent as possible with their readers, giving them a direct look at their conversations with AMD. I appreciate that, but in this case it has simply caused further confusion.
The way I see things: AMD may release some 7000 series refresh parts for the desktop, possibly +10 like the old 4890 was to the 4870. Perhaps not across the board though. We could see only one added for the high end of each segment or even only a couple of segments. This sort of refreshing makes a hell of a lot of sense to me. The VLIW4 6900 series of 2010 was a bit of an anomaly caused by the delay of 32nm and shift to 28nm. I don't think we should expect these sorts of refreshes in the future.
8000 series are focused on OEM and laptops. Perhaps AMD's image was damaged here with the 7000 series on top of OEM's desiring new numbers every year.
I think that the products we see towards the end of the year will be 9000 series. I kind of doubt they will be 20nm unless things are going very well for that process at either TSMC or GF. It would, however, certainly be nice to see what I consider a proper new generation in that time frame.
The way I see things: AMD may release some 7000 series refresh parts for the desktop, possibly +10 like the old 4890 was to the 4870. Perhaps not across the board though. We could see only one added for the high end of each segment or even only a couple of segments. This sort of refreshing makes a hell of a lot of sense to me. The VLIW4 6900 series of 2010 was a bit of an anomaly caused by the delay of 32nm and shift to 28nm. I don't think we should expect these sorts of refreshes in the future.
8000 series are focused on OEM and laptops. Perhaps AMD's image was damaged here with the 7000 series on top of OEM's desiring new numbers every year.
I think that the products we see towards the end of the year will be 9000 series. I kind of doubt they will be 20nm unless things are going very well for that process at either TSMC or GF. It would, however, certainly be nice to see what I consider a proper new generation in that time frame.
Similar threads
