I've already seen other articles interpreting that to mean that there is simply no console launch in 2012. He also says he's "not convinced that things are happening this year." That's some commitment.

And really.. how old was the X1800 by the time the 360 launched? Is the Xenos 1:1 an X1800? No. Is it still closely related to that architecture? Sure.

How is it shit? I'm more into Wii U right now than anything else because it's real and Nintendo is doing something different. If I want typical experiences, I already get loads and loads of them on the PC :D

The thing is, those quirky experiences sorta appeal to me since I can get mostly every other kind of experience on the PC.

If you also consider the AMD 7970 and the AMD 6970 closely related architectures, then sure, why not? :P

Well that's sorta my point though.. all of these people complaining that they're using tech that's already old seem to think the Xenos is based on tech that was cutting edge when it shipped. It was built up from the X1800 with unified shaders and a slew of other features from the next generation of AMD products. Is it really that different?

Hardly.. some dude from one of their European branches said he didn't think it was going to happen. I don't call that an official announcement.

I guess I wasn't very clear. You had said that:To which I responded:And what I meant by that is, tongue-in-cheek, is Xenos is architecturally very different from the X1800 (R420) and previous ATI generations (R3xx; R400 was an unreleased branch of development). I would contend more so than even GCN (7900) is from the 6900 series. The Xenos architecture, features, and implimentation are just far and away (a) built on ATI's portfolio and design methods but (b) a totally different GPU lineage.

The Xenos architecture. Xenos' unified shader architecture was a big change and the industry as a whole has followed. It is a major design win if you can get it to work and ATI did with Xenos (not so much the so-called R400). And there were other major architectural changes such as the TMUs being independent of the shaders (unlike say RSX where shaders could tie up texture units). The ROPs are also directly tied to the framebuffer memory with massive framebuffer bandwidth. It also had some tradeoff, e.g. it appears to have skimped on TMU caches which with the meager main system memory is a cause for poor AF performance (quality AF with solid performance was addressed with the R300 so this is a design set back). There is also MEMEXPORT for directly dumping data ("stream") from the GPU to system memory. The memory controller was for the CPU and GPU was on Xenos (the PC GPUs would have a memory controller to VRAM but use the PCIe bus for system traffic coordinated by the CPU/Northbridge memory controller).

The Xenos features. Single cycle 4x MSAA, FP10 buffers, true vertex texturing, hardware tesselation (500M vertexes/sec. setup rate, 250M tesselated triangles a second), FP16 buffers with MSAA, etc. Xenos had a lot of features not present in GPUs in AMD's lineup to date.

The Xenos implementation. The aforementioned ROPs, tied to the framebuffer, were off-die; Xenos essentially had a 128bit bus to the 512MB UMA system as well as another bus to the ROPs. The ROPs then had a much wider path themselves to the framebuffer. Xenos was a 25W GPU at 500MHz (memory, CPU, eDRAM, optical drive, HDD, cooling, I/O, etc make up a large amount of the systems power draw; even a PC GPU a lot of power draw is for the memory as well as some for mechanical cooling).

This is from memory so I am missing stuff and may have a couple facts misplaced. BUT the point is this: Xenos, architectural, was not related to the X1800. Xenos is less related to the X1800 than the 7900 "GCN" architecture is to the previous 6900 series in my opinion.

The R520 as a X1800 XL was a 256MB card with a 500MHz clock on 90nm with 32GB of memory on a 256bit bus. The R520 as a X1800 XT was a 256MB or 512MB card with a 625MHz clock on 90nm with 48GB of memory on a 256bit bus. All in 320M transistor in 264mm^2 with at least 123W TDP (I am not going to bother digging through all the reviews to find out what the idle base load is just for the GPU).

Anyways, just huge architectural, feature set, and implementation differences. And as a design it performed quite well for its purpose--an X1800 would not have fared as well as Xenos. Maybe an X1900 would have the additional shaders were cheap but then that is a bit later in development and more so expensive. The fact the PS3 constantly has to use CELL to play catch up to Xenos is pretty telling of how Xenos fairs against the R500/G7x gen of hardware.

The biggest concern I have is from what someone on B3D posted:The TDP issue is getting pretty bad for a small space. e.g. moving to 28nm at best is giving like a ~40% increase in transistors at the same frequency or, for the same tranies and frequency, like a ~30% or something like that reduction in TDP versus something like thr 40G process. As you can see to 2x your performance is going to increase you TDP.

Since the 360/PS3 launch what were formerly high end GPUs were breaking the 100W barrier are now topping off at 300W for NV's large GPUs. AMD has scaled back performance some (I recently posted about how cutting frequency on a bigger chip seems to win out in some cases over a smaller higher clocked chip).

Personally I think the solution is to address the memory draw (TEVS for stacked memory; an SI for a much wider bus at lower frequencies) first. Such approaches also addresses the bandwidth bottleneck issues. A revised R7950 on a mature 28nm process will be smaller and more power efficient... but ideally 22nm/20nm 3D transistors would solve a boatload of troubles.

The problem is it looks like at least one of the HD console makers is not going to wait. If they don't I do hope the competition plays their hand right and blows them out of the water. It is amazing what an extra $50-$100 put into a box can do -- unlike the hundreds put into crap into the PS3 that did little to actually change the user experience (unless you really care about BDR which, well, I can get a dedicted player for $100 if I wanted).

As for the concern for asset creation, developers are already making multimillion poly models for all those normal maps. Using those same source assets, bumped up and maintaining better source texture data, and then improving lighting/shadowing all around is going to make a big difference without killing budgets. Big titles are going to cost an arm and leg like they always have. My worry is, after spending some time doing frame analysis of a game like FM4, **console games have a lot to improve**.

PS- In general, to throw this out, are the consoles really holding back PC graphics? It is all about money and whether it is cost effective to make a game on a platform with dedicted features and design. Because the dominance of PC sales with robust CPUs and crap GPUs over the year and the meager install base of even mainstream GPUs the market just isn't huge. At the turn of the gen middleware, shovelware, last gen cross development, etc help back the current consoles for a couple years. And while consoles are currently on the lower end they still offer more bandwidth, more processing cores, and stable GPU targets compared to the PC market. There are over 100M HD consoles in the market. How many active gaming PCs are above that in the consumer space (i.e. subtract business PC sales)? I love me some PC action but it just doesn't appear to be an option. And just wait because the new complaint come 2013/2014 will be the weak APUs with crap bandwidth will be the baseline target on the PC for most games (it makes no sense to cut that market out once the newer models come out imo). If that is the target 8-10 years post-360 it looks like the 4 core CPU w/ IGP sales from Intel and the continued dominance of essentially IGPs (which, btw, I think IGPs would be better than the integrated stuff we are seeing now) just shows how the market could use a big kick in the rear.

I think MS or Sony, one of them, will at least push a little under the hood. It won't be a 300W+ GPU or a 128 real core CPU with 8GB of memory. In fact I bet one of them goes with crap CPU cores, a very modest GPU, and 2GB of memory. But I think one of them will think a little bigger and it will make a difference on screen... unless someone decides to trojan their next BDR or Kinect at the expense of the platform.