Predict: The Next Generation Console Tech

Hmm, just spotted a odd Carmack tweet. Working on next gen?

http://twitter.com/#!/ID_AA_Carmack/status/149196752573054976

Later edit: Doom 4?

 

I asked him if he'd like to see 64MB EDRAM. He didn't reply

 

The target specs for the first couple of years for most games will likely still be 512 MB RAM and the current in order XB2 and PS3 processors. They probably won't have the budget to target significantly better than that to create a ground up game for the new systems.

 

Just like they have no money to offer upgraded textures for the PC ports, right? There will be launch titles that look a bit behind, but within a year you'll have major AAA titles making last gen look fairly weak.

 

My bet is MS will go for radical mem system again.

250mm2～350mm2 SOC with 2/4 GB stacked wide-io memchip on a interposer
3 Power7 cores with 4-way SMT,4/8 MB edram L3
16～24CU GCN（1024～1536SP）
unified memory address space
125W～150W TDP (SOC+mem)
Core model HDD only,elite model HDD+BD drive
Launch in H2 2012/H1 2013

 

What are the advantages of having an esoteric memory hierarchy? I don't think cost is one of the benefits.

I think Microsoft will stay [relatively] true to the formula that has worked. If it isn't broken, why fix it?

In both of my systems, the GPU is given prevalence over the CPU. The 7970 is a 250w part with a 1Ghz clock speed on a 28nm node. If we scaled back the v-core, bsclk, and fab it at 20nm then I think a 7970x2 can fit the TDP of a console for 2013. Heck, tack binning priority onto my power-saving parameters and my attributes for a next-gen console sound even more realistic.

If we relinquish this arbitrarily-induced 200w limitation and allow the TDP to swell linearly like it has done every generation prior, then I can definitely see a 300w box in the cards.

If my high-spec is discomforting to you, perhaps you can amalgamate it with the low-spec and get something to the tune of a multi-core Power7 central processor, 2048 SPs, 2Gb GDDR5, and 64Mb of embedded RAM.

I agree with mostly everything Alphawolf has said... When you introduce a new console, you're introducing it to your core audience; the guys that will use their sick days just to stand in line a week before your console launches. To this end, I'm hoping my high-spec is more accurate than my low-spec.

-IK

 

Anyone has info on how cooling much cooling technology has improved since 2005? Would the cooling tech in the slims be suitable for a 200w+ console?

Shouldn't the CPU be power8 at 22nm? Anyways, I seriously doubt we'll see the above specs for a system in 2013, maybe for ps4 in 2015+.
How many Gbytes of ram does 6 Gib translate to?

 

The X360 formula is all about graphics,not suitable for future gaming which need more compute power for AI and physics ect..

 

I think an OoO, quad-core Power7 with SMT is suitable especially as GPUs become more general purpose.

Besides, how much power do you need for AI? If it was truly an intelligent system then it should be no less "intelligent" on an Cortex A5 as opposed an i7 if the term "AI" is inherently relative. I'm tired of just throwing more cores at everything.

 

Most developers don't care that much about improving AI/physics over what is already posible on the X360 and either way they can be improved using GPGPU code rather than CPU code.

 

I don't think you'll see any super revolutionary cooling solution for any new console because on a fundamental level you're limited by the cooling medium, i.e. air. Air is a rather shoddy cooling medium as it's capacity to contain and transport heat is much lower than that of say water or some other more exotic refridgerant fluid. Regardless of how a console is physically designed you still have to ultimately get the heat from the processing cores in to the ambient atmosphere, and simply passing a continuous flow of air over your board is the simplest, cheapest and most energy efficient way to do this.

Liquid cooling solutions are more expensive because they take much more energy to transport and also place too many limitations on the physical size of the unit. Liquids are more viscous, thus requiring more energy to transport around a cooling circuit to overcome the frictional losses causes by the liquid interface moving across the insides your pipes/transport capiliaries. A liquid cooling solution is also much more sensitive to pressure drop around the cooling circuit, and given the inversely proportional relationship between capiliary diameter and pressure drop, it limits the physical size of the overall unit because the smaller the console box (and thus capiliaries) the higher the pressure drop per unit distance around your cooling circuit, and so you need to do more work in your pump (thus higher power consumption) to drive the fluid around the circuit.

All in all liquid systems will be far more complex, thus expensive, and then you also have issues with the potential for leaks and such.

It's most likely why the mindset has been prevailingly towards reducing the amount of heat produced by the chips, rather than trying to engineer some super sophisticated and unavoidably expensive cooling solution to cool foolsihly high power, high heat producing cores.

Edit:
So in short, i don't expect a 300W system next gen, because cooling such a system in a console-sized box (that people would be willing to buy) would not be such a trivial matter at all.

 

I have a few questions:

. Do people in here think the gap between Wii-U and Xbox3/PS4 is going to be similar to DC/PS2 -> GC/xbox like I do?

. I'm expecting 1-1.5 GB RAM in Wii-U. What's the amount of RAM expected from Xbox 3 & PS4 for a 2013 launch?
Do you think Wii-U's GPU will be DX11 compatible and is DX12 even in the equation for Sony and MS' consoles?

. Which one is easiest and least costly to do? Up-porting or down-porting. The reason I ask is because I could see smaller devs developing on Wii-U to port to the others while the bigger ones develop on PS4/Xbox 3.

. It seems Nintendo don't plan to push 3D gaming with Wii-U but I think MS and especially Sony will push it with theirs'. Would this mean the visual gap between wii-U and PS4/Xbox 3 closing a bit more if devs have to use some of the latters' processing power to render games in 3D?

. Last of all, how many of you believe that development costs will rise slightly? Marginally? Significantly?


* Visuals are going to be even more indistinguishable to the average customer next gen. They're not going to be comparing small details, effects and such things like enthusiasts/hardcores . I agree with Squilliam about User experiences being much more important next gen.

Nintendo is going to be using their online system (Which I think is going to pleasantly surprise alot of people) plus their tablet.

MS, Windows 8 + live.

Sony...er...PSN and 3D, I guess.

 

It depends what you want. If you are aiming for a near 1:1 port (only maybe increase resolution and adjust some visual parameters) up-porting is much easier. It's easier to get code running well on a device that is more powerful than the original one. Down-porting requires very heavy code optimization (to get it running as near 1:1 as possible) and often that's not enough so you have to optimize your content as well = more work for the artists as well.

Up-porting can of course take a lot of resources as well, for example if you are porting some older game and want to remake all the visuals. But if you are doing both versions at the same time, you likely have all the textures available at higher resolution (artist work files tend to be at higher resolution). For the lesser platform you can simply drop some LOD levels from the models etc. It of course all depends on the quality required. Teams usually spend most time tuning the platforms that are the most important.

 

IME most teams are directed to lead on the platform they think will yield the most sales, practically there are other considerations (quality of tools, availability of devkits, developer ego etc).
In the US it is unusual for the same team to do both a primary console SKU and other SKU's, usually portable and even PC SKU's will be done by other developers, often external developers.

How much support WiiU gets (assuming there is a significant performance difference) will depend on how much market share it has and how 3rd party games sell on it.

 

I expect it to be more pronounced than that, but nothing like PS360 vs Wii.

I would be surprised at less than 4GBs, but it depends on RAM speed and other memory systems. 2GBs system RAM and 1 GB very fast VRAM could be a fair option.

But those devs would likely lose considerably to games written specifically for PS4/XB3. An upconverted MW5 will look poor versus a true BF3. If Wuu has a massive install base then up-porting makes sense, similar to targeting PS2 and then porting to XB, but that's quite a big if IMO.

Depends how 3D is handled. Those gaming on 2D could potentially have better IQ as a result. Simplest way to do that, if rendering two full viewports, is render the same camera only offset half a pixel in the left and right eyes and average the result - instant 2x supersampling for 2D games. Or render 30 fps in 3D and 60 fps in 2D. But if 3D is achieved through funky tricks, costing all of an extra 10-20% overhead, then the difference won't be so much.

The biggest budget titles will have bigger budgets, but there'll be room for smaller budget titles too. And potentially, better hardware means cheaper efforts creating a moderate quality game. The industry should mature into a wide range of costs and prices.

Sony will be just like MS, with Sony Entertainment Network. They are offering 60 days free Music Unlimited at the moment, for example. They want a flexible content platform on Android as well as PS. Whether they pull it off is to be seen, but the services will be quite discrete between them. That'll put emphasis on exclusives though, and you've also ignored the new controllers. MS will have Kinect 2. Sony should ahve some wand-camera interface. They've had lots of Kinect-like research for years, and one day they might vring it all together into an interface.

So there'll still be plenty of reasons to pick one console over another despite hardware performance, but the latter is going to count if the differences are pronounced, especially in the formative years which are so crucial in establishing user base, word-of-mouth sales, and gaining publisher support.

 

My Q9550 and 6950 manage fine in my P182 with two low RPM 120mm exhaust fans ... that's not a lot of surface area, a console can manage that too. I don't see what the big deal is.

 

How much system RAM is really needed? I was thinking they could get by with 1GB even. Assuming mem arch isn't shared, I think they need to do at least 1GB RAM for the GPU, but 2GB would be real nice

 

I think many see cooling and heat transfer as more difficult than what it really is due to the early difficulties Sony and especially MS had years ago on their consoles. I expect the next consoles to draw at least 200W at launch and it shouldn't be a problem to design the cooling system to handle that with ease.

I just don't think there is room to lower the TPD and still get big enough jump and more importantly I see very little reason to lower it in the first place. The arguments for lowering the TPD from launch of last gen are weak. There are more truly meaningful "forces" pushing the power draw higher than lower imo. I understand this is debatable, but it's the way I see it.

 

Two 12 cm fans would seem to require an awful lot of surface area for a console - that's four times the area that the 360's two fans took up and it's not like you could just mount one of them directly over an optical drive - and you certainly couldn't put one along the edge of the case. Inside a PC case there's a cavern for the air to flow through (even in a mATX case), inside a console the space is much more limited and how the air moves through the case is at least as important as the maximum area of fan(s) you could cover the outside of the case with.

A 2kW low profile fan heater can fit into a space not much bigger than the launch 360, but I don't know how many people would want to put one in their AV cabinet.

 

Sony's solution was excellent, just expensive. That's an expense to be avoided if possible. The YLOD faults are more to do with lead-free solder as I understand it. Although Sony did cheap out with the thermal paste. Hopefully lesson learned, they'll invest the few extra cents per unit for a decent thermal interface and save truckloads on warranty servicing!