Now that the unveil has come and gone I think I should summarize this for those having a hard time. bkillian can tell me how wrong I am
MS back in 2009-2010 when was setting their targets. Fabs were having issues and certain technologies continued to slide (DDR4, FINFET, silicon interposers, stacked memory, etc). The current generation was running long in tooth and Sony and Nintendo surely going to release before the other technologies materialized. So like Sony they had to go with what was on the board.
They clearly set out with a number of goals. First was the next Xbox was to be what the vision was for the first Xbox platform: a central entertainment hub for all things digital media. This clearly means the focus of the platform is squarely Xbox LIVE moreso than the hardware delivery. The next issue was costs (BOM as well as TDP), specifically cost reduction--if they want it to be a media hub and to displace/compliment cable boxes, be integrated into other devices, and be so cost effective to eventually squeeze out the Rokus of the world cost reduction was vital.
So they walk into a room with AMD and get (a) the same general technology options and (b) the same BOM / TDP metrics. They both chose APUs for many reasons so their choices became really obvious.
MS chose the following:
- MS decided to fully embrace the media box it needed 2 distinct OSes, hence the hypervisor and a Win8 variant and GameOS. This would require a lot of memory.
- DDR3 offers high densities and is very cheap (but slow). At the time it looked like MS could get away with 8GB of DDR3 versus 2GB of GDDR5. The trade off was speed for space. As the APU was going to be a big chip pad limits weren't a huge issue over the lifetime of the design. The bet was they could reserve 3GB of memory to the OS and be at 5GB vs. 2GB. Or even if Sony went with 4GB at least 500MB was going to the OS so it would still be a 1.5GB advantage.
- Embedded memory to resolve bandwidth shortages. A separate eDRAM module requires another bus and likely Xenos issues so that was off the board as it distracted from the SOC goal. eDRAM on die would be costly due to process issues. ESRAM was the next best option (as other variants were off the table for licensing?) The problem is ESRAM takes up a lot of real estate. The good news is that it scales well with process shrinks. The ESRAM doesn't appear over engineered (e.g. just enough bandwidth for what the system can use). Surprisingly, probably due to how complex these chips are these days and AMD's limited resources, ESRAM is not shared between the GPU and CPU.
- Storage. For the media and gaming goals they need a delivery system and local storage. Online is still too slow and lacks penetration so BDR was the easy choice--and have consumers pay the licensing fee for movie playback. SSDs are too expensive/small so a single platter HDD would do.
This is where things went wrong: GDDR5 densities grew allowing Sony to match memory footprint.
So back to MS's platform. They obviously have a HUGE multitasking/multimedia edge--the entire platform is built around apps.
On the reverse side the technology budgets clearly show gaming was not the first priority with "everything else having to fit in after that."
Anyways it is a cascade effect. MS has fewer CUs because they have ESRAM taking up die space. They have ESRAM to compensate for slow DDR3. And they have slow DDR3 because [strike]they have a TON
more memory[/strike]. And they have a [strike]ton more memory[/strike] so they can compete in gaming while also serving up all their media services.
It all blows up with Sony obtaining 8GB of GDDR5.
Reverse the flow: If MS had 8GB of GDDR5 they would not have needed ESRAM. And if they didn't need ESRAM they would have the same number of CUs.
And the crappy part for MS is I am betting their BOM out of the gate will be on par with Sony's.
Now not all is doom and gloom.
First is that the GDDR5 is going to be costly for a long time. On the flip side the ESRAM is going to shrink QUICKLY so MS's APU will become more affordable, quicker (in theory--if process reductions continue at a snails pace I think this was a horrible gamble). So MS has much cheaper and common DDR3 and will get to leverage more cost savings on node reductions.
MS also has Kinect (this is where their launch BOM may be higher) -- which the tech finally looks fantastic but they need games to prove its worth. The Kinect's killer app is probably the dashboard and Skype (which MS paid BILLIONS for).
Ahhh yeah, this thing was getting expensive so the embedded 360 SOC was tossed out. Hey, this saves money and MAKES money when you resell the XBLA games!
(Cha-ching MBA's with dollar signs in their eyes!)
Anyways, for all these reasons I highly doubt MS ever entertained changing the HW once they knew what Sony had -- EVERYTHING was tied into this strategy. I am sure they expected to be bested by 50% in CU performance.
Where MS was gonna hit back is they are close enough for rough parity--resolution reduction (something studies show most consumers do NOT notice) will level the playing field. The extra memory in theory could have assuaged some of the load times, and their killer app was (a) XBL / Media centric and (b) Kinect.
Unfortunately the wheels on the bus have fallen off for the core market: The platform has 33% less compute, a similar reduction in texture performance, 50% less fill-rate, a more complex dev environment due to the ESRAM, slower main memory, less main memory, probably less CPU resources (Sony has some helper logic for background tasks and MS has to power all that media crap somehow), all in an expensive box sporting multiple OS's running side by side that is clearly not aimed at gaming as the core/prime use but as an equal, important, but comorbid function with the media suite all tied into Kinect.
This will make many consumer super duper happy.
This will make many gamers-only very unhappy.
If Sony was at 2GB, or even 4GB, of GDDR5 I think the discussion would like a lot different. It obviously all comes down to games. But as a gaming platform it is pretty clear Xbox One is at a disadvantage across the performance board. Multiplatform games are going to suffer. I think Xbox One will performance wise be similar to the GCN next to the PS2 and Xbox1 -- it ran the same stuff and play the same, just not always as pretty or as smoothly.
So stop dreaming of hardware changes or upclocks. All these scenarios were thought out well long ago. MS knew Sony would have more gaming beef. That was by design.
MS is content with "very similar" gaming experience because the 33% less compute means better cost reduction down the road and allowed MS to make KINECT and MEDIA (XBL) equal full-time partners in regards to access and utility.
As a consumer you have to choose: Are all the extra Media things MS doing worth a couple less pretty pixels here and there or not?
Obviously price, exclusives, multiplatform performance, how used games and online pricing plays out, (and stuff like 3rd party indie publishing) will all be important but in regards to hardware and what it was designed to do do you want something that can overlay your Cable Box and interact with it and has a killer 3D interface in Kinect always on and Skype and Apps always on ready to be used and "Almost as good as Sony" gaming performance or do you want Tier1 Console gaming (but not PC!) and the apps be more secondary (something like the 360 on steroids)?
I know what I would choose... I think many will be surprised what general consumers choose. A much better Kinect and media coming out of your ears, especially the likes of the NFL, will appeal to a lot of non-core gamers who are none-the-less technophiles.
