Digital Foundry Interview With IMG

Lazy8s

Veteran
Caught this link from Rob Evans.

http://gamesindustry.biz is hosting an interview Digital Foundry just did with IMG. The interviewer, Richard Leadbetter, had done some good coverage of Dreamcast and its technical merits back in the day, so he puts some informed questions here for Kristof Beets and David Harold as an overview of PowerVR's market approach and offerings.

Some insight into balancing customization versus simply scaling when making a new variant core:
Q: You have other multi-core projects in the pipeline for the series five hardware. What advantages do they have over the SGX543?

IMG: In addition to SGX543 we have also announced SGX544 which offers the same performance characteristics but enables fully compliant DX9 Feature Level 9_3 capabilities so basically an extra bump in feature set to meet Microsoft requirements. Also available is the SGX554 which is our first 8 pipeline part (SGX543/544 have 4 processing pipelines) which offers improved compute density for customers focused on GP-GPU and shader processing since a single SGX554 would offer the same compute capability as an SGX543 MP2 but not the same geometry or pixel throughput.

This means that SGX554 offers more GFLOPS per mm2 since the design avoid overscaling the geometry and pixel capabilities of the design versus customer requirements - basically we do not believe in "one size fits all" solutions and we thus offer our customers various options.
 
Last edited by a moderator:
Caught this link from Rob Evans.

http://gamesindustry.biz is hosting an interview Digital Foundry just did with IMG. The interviewer, Richard Leadbetter, had done some good coverage of Dreamcast and its technical merits back in the day, so he puts some informed questions here for Kristof Beets and David Harold as an overview of PowerVR's market approach and offerings.

Some insight into balancing customization versus simply scaling when making a new variant core:

so the SGX543M4+ in the NGP don't really have much of a disadvantage compared to the SGX544 because it's not using DX9 & it won't have much of a disadvantage compared to the SGX554 besides when something is using the GPU for GP computing which shouldn't be too much of a big deal since the Cortex-A9 is also good at most of the things that would take advantage of the better GP-GPU computing?
 
Last edited by a moderator:
Some interesting information here. Looks like an SGX554 really does double the ALU:TMU ratio vs SGX544, which is what I expected.

so the SGX543M4+ in the NGP don't really have much of a disadvantage compared to the SGX544 because it's not using DX9 & it won't have much of a disadvantage compared to the SGX554 besides when something is using the GPU for GP computing which shouldn't be too much of a big deal since the Cortex-A9 is also good at most of the things that would take advantage of the better GP-GPU computing?

SGX554 cores will have the advantage over SGX543 cores for anything that benefits from improved ALU utilization. GPGPU applications are a good example since they don't benefit from an increase in setup rate, depth comparisons, tile binning, and various other fixed function capabilities of the chip.. and they may care less (but still care) about TMUs. But that doesn't mean that only compute benefits from more ALUs.. with more ALU capabilities you can do more complex fragment shaders, and this is something you can't keep on the CPU.

SGX554 does make more sense for compute but it's hard to imagine a company picking an array of 16 of these solely for compute purposes, surely there are better avenues for HPC.
 
Last edited by a moderator:
so the SGX543M4+ in the NGP don't really have much of a disadvantage compared to the SGX544 because it's not using DX9 & it won't have much of a disadvantage compared to the SGX554 besides when something is using the GPU for GP computing which shouldn't be too much of a big deal since the Cortex-A9 is also good at most of the things that would take advantage of the better GP-GPU computing?
Well since the PS3's GPU is DX9.0c and the Microsoft liked to classify the 360's as DX9.0c+, just looking at raw technical compliance between the SGX543 and SGX544 with DX9.0 support could be a point in NGP vs PS3 "power" debates. Of course, the smaller resolution, smaller screen size, etc. could mean full DX9.0 level 3 compliance is not necessary for the NGP for things like larger textures size support.

IMG said:
The total Series5 and Series5XT portfolio enables the industry's broadest range of performance/area options, from the smallest single pipe SGX520 core up to the 64-pipe SGX543 MP16. All popular APIs and OS are supported by all SGX cores, including OpenGL ES 2.0/1.1, OpenVG 1.1, OpenGL 2.0/3.0 and DirectX 9/10.1 on Symbian, Linux, Android, WinCE/Windows Mobile and Windows 7/Vista/XP.
This response seems kind of contradictory though. Are they really claiming that all SGX cores support DX10.1 since I thought that was the SGX545's unique feature. Or are they just generalizing that the Series5 architecture is capable of DX10.1, with it implied that only the SGX545 actually has it implemented.

Out of interest, Sony says that it has a PowerVR SGX543 MP4+ inside Sony NGP... what does the plus stand for?
IMG: That's to indicate the work Sony has done to implement the graphics. What they licensed is a SGX543 MP4.
I guess that mysterious "+" will still remain a mystery. I wonder if IMG actually knows what Sony has done or if they sold the license and that's that?
 
Some interesting information here. Looks like an SGX554 really does double the ALU:TMU ratio vs SGX544, which is what I expected.

Equally interesting:

Just as a reference, over time we have seen implementations of the same SGX core going initially from 110MHz to 200MHz and today designs are beyond 400MHz in silicon.
By the way congratulations Richard; a job well done :)
 
This response seems kind of contradictory though. Are they really claiming that all SGX cores support DX10.1...
I read it as "We support DirectX on all SGX, either 9 or 10 depending." Otherwise there'd be no need to mention DX9 as DX is fully BC, and DX10 support implies DX9 support.
 
Followed off a link from IPPaws:

another good piece by Richard Leadbetter for Digital Foundry -- hosted at http://gamesindustry.biz/ -- about the significance of the Apple A5's graphics performance. While Apple pushes cutting edge iOS hardware in order to serve the demands of increasingly functional apps and not necessarily to position themselves for a game console as Richard entertains, the article is well reasoned and provides great technical analysis and videos of A5's performance.

http://www.gamesindustry.biz/articles/digitalfoundry-in-theory-apple-home-console

None of the content of iOS's most demanding game software, even after the engines were updated to take advantage of the additional performance, stop the A5 from consistently maxing out the engines' refresh limits.
 
Last edited by a moderator:
I'm not quite sure why the framerate would be capped between 34-36 frames per second? Isn't there just a more consistent bottleneck for this, rather than what the original iPad ran into? And as far as performance goes, I'm not that impressed yet to be honest - is there even any real shadows or lights going on in there? Everything looks pretty much baked.

But yeah, it is an improvement over the first iPad, that much is certain. And a good thing that the resolution wasn't increased to Retina level. I have my doubts on how this performs at 1080 over HDMI. Will that even be native? (not that it matters much)
 
And as far as performance goes, I'm not that impressed yet to be honest - is there even any real shadows or lights going on in there? Everything looks pretty much baked.

But yeah, it is an improvement over the first iPad, that much is certain. And a good thing that the resolution wasn't increased to Retina level. I have my doubts on how this performs at 1080 over HDMI. Will that even be native? (not that it matters much)

Real racing2 HD: 1080p comes to IOS
http://www.eurogamer.net/articles/digitalfoundry-real-racing-2-hd-1080p-comes-to-ios

native 1920x1080 frame buffer for HDMI output , whilst still running a 1024x768 screen output on the ipads own screen.
 
EDIT: ah, this is newer and is in fact the actual analysis. So they did manage it. Pretty impressive - even if it generally looks dull, they even output to two screens at once.
 
Some extra analysis was added in a new Eurogamer article from Digital Foundry for the same theme of the possibility of an Apple games console. The visual improvement from the MSAA and also the performance impact of OS multitasking are particularly interesting to see.

http://www.eurogamer.net/articles/digitalfoundry-vs-ipad-2

In contrast to the theory of a gaming initiative by Apple proposed by the article, the choice of the SGX543MP2 and Cortex-A9 dual was always obvious based upon which mobile cores fit the die/power budget for that timeframe for a SoC trying to stay current with the rapid advance of mobile computing. Pretty much every semi came up with dual A9s, and PowerVR Series5 XT licensees who were refreshing at that time (essentially Apple and Renesas) also went 543MP2.

Apple doesn't need to make a gaming push; they'll just organically expand their scope until they envelope much of the gaming market.

And they don't need physical controls either; games that succeed in its market will just be centered around touch friendly gameplay and interfaces.

And the market will have to change, because the arbitrarily high prices supported by the current console model will be made obsolete.

And games will still be made at a competitive level of quality, but they'll just be developed/funded in a more logical way. Fewer costly art assets, populating gigantic -- mostly barren from a gameplay perspective -- levels will be developed. Instead, developers can design one part of a game world at first and just pack it tighter with more interactivity. They can take better advantage of the assets/objects within and exploit their gameplay potential by expanded physics models and gameplay dimension. Then, that one part of the game can be sold at a low price and with a low investment by the developer, and the game can be built upon with further downloads if the first release catches on with the consumer.
 
You are saying all that, but so far research in my country has just shown that so far it is only expanding the market. Gaming time went up 50% thanks to an increase in mobile gaming that did not see a decrease in console gaming.
 
You are saying all that, but so far research in my country has just shown that so far it is only expanding the market. Gaming time went up 50% thanks to an increase in mobile gaming that did not see a decrease in console gaming.

smart phone games is the free crack that gets them hooked on gaming then they have to come & buy the good stuff next thing you know your grandma is at the game stop shaking asking for the new Call of Duty.
 
Some extra analysis was added in a new Eurogamer article from Digital Foundry for the same theme of the possibility of an Apple games console. The visual improvement from the MSAA and also the performance impact of OS multitasking are particularly interesting to see.

http://www.eurogamer.net/articles/digitalfoundry-vs-ipad-2

In contrast to the theory of a gaming initiative by Apple proposed by the article, the choice of the SGX543MP2 and Cortex-A9 dual was always obvious based upon which mobile cores fit the die/power budget for that timeframe for a SoC trying to stay current with the rapid advance of mobile computing. Pretty much every semi came up with dual A9s, and PowerVR Series5 XT licensees who were refreshing at that time (essentially Apple and Renesas) also went 543MP2.

Apple doesn't need to make a gaming push; they'll just organically expand their scope until they envelope much of the gaming market.

And they don't need physical controls either; games that succeed in its market will just be centered around touch friendly gameplay and interfaces.

And the market will have to change, because the arbitrarily high prices supported by the current console model will be made obsolete.

And games will still be made at a competitive level of quality, but they'll just be developed/funded in a more logical way. Fewer costly art assets, populating gigantic -- mostly barren from a gameplay perspective -- levels will be developed. Instead, developers can design one part of a game world at first and just pack it tighter with more interactivity. They can take better advantage of the assets/objects within and exploit their gameplay potential by expanded physics models and gameplay dimension. Then, that one part of the game can be sold at a low price and with a low investment by the developer, and the game can be built upon with further downloads if the first release catches on with the consumer.

I agree that Apple won't specifically target games but get more capable silicon, probably at better prices than anyone else. For the quarter they just reported, they have $11 billion in purchase commitments for the supply chain. That kind of scale gives them the clout to use the best components without fear of being underpriced.

They don't need physical controls but there are several attempts by third parties putting out controller with SDKs. If Apple put out a controller with APIs, iPad gaming would take a step forward. A universally-supported physical control scheme for iPad games would skim them a few more percent of the handheld and console gaming market. The other part is continuing the same pace of improvement in SOCs every year.

A lot of apps., not necessarily games, are offered for free or very cheap and then add a lot of content via in-app. purchases. Developers and publishers are already pushing DLCs on XBL and PSN with retail games and they may come to rely more on the model of selling the basic game for a low price and then try to capture back more and more revenues through additional content. This additional content might have been features or levels they would have included in a shrink wrap game but for the App. Store, they withhold for in-app purchases so that they can distribute the basic game at a low-entry price to attract the highest volumes possible.
 
The rest of the article was so good that it was glaring to see them call it a dual-core 535 in A5!
 
Back
Top