Valve Talks More About Nvidia GPUs Inside Steam Machines

[Memnarchon]

I'm just trying to be devils advocate here. Its more than enough for the type of system I would run, and can run a 4770 and a Titan.

But 80+ Gold means its 87- 90% efficient typically. So that equates to a 450W power supply being able to deliver effectively ~392-405W of total Power under max load. That leaves <50W headroom for potential power hungry potentially hardware in the future. Based on recent advances with Haswell and newer GPUs being more energy efficient, you would think that 50W could be enough. But you never know how future hardware will playout (overclocking, more advanced HW etc). You never want to make sacrifices on your PSU.

Since Valve has three tiers of hardware, they can offer a slightly higher wattage PSU for the enthusiasts out there.

I am trying to be the devil's advocate too some times 😛.
But here there is something you misunderstood. Efficiency isn't about how much power will deliver to the entire system. But how much power it will draw in order to deliver this power to machine.
Take an example of Silverstone's 450watt plus gold at HardwareSecrets at load tests. For 449,8watts it draws 525,3watt. Now if it were bronze it would draw more watts for these 450watts which will be delivered to the system.
So with 450watt PSUs there is room more than enough for these machines.

 

[ojas]

Interesting, very few people actually even understand what the announcement means.

 

[somebodyspecial]

Where will this SteamBox stand on the forgotten console list?

Nvidia, did you pay Valve like you did with origin?

http://semiaccurate.com/2013/10/05/much-nvidia-pay-origin-pc-drop-amd/

Well apparently AMD ran out of funds after blowing 8 MILLION on EA's Battlefield 4 Mantle optimizations. So I don't get your point... Everyone cheats? Maybe if people like you started asking for NO FREE GAMES and HIGHER PRICES on AMD cards they'd have more money to blow on things like, I don't know, DRIVERS? 😉 CPU's? You know, that CORE important stuff that your business lives or dies on instead of dying breed crap like consoles. That entire wad should have been spent on MOBILE, that will be revving yearly instead of stuck at 1080p for 8yrs holding back my PC gaming! I don't hate AMD, I just hate the people that currently are running it into the ground. I also hate the fact that because of this, I have $500 low end cards from Nvidia in my future as AMD just keeps getting weaker and weaker.

I already have those idiots to thank for PRICE HIKES on i5/i7 chips. For christ's sake bring back Jerry or Dirk today (dirk left and the cpu's went to crap once his tech ran it's course)! Note the 4770 is MORE than Ivy's and so is the i5. This is what happens when the competition (AMD in this case) bleeds for years on end and can't fund R&D, with no end in sight. The driver situation was caused by the same crap. If they were making money they'd have no driver problems right now as they would have funded the R&D to fix it before shipping the cards.

No consoles won't make them rich, in fact the first $100-200mil they make in the first 12 months on them (tops 2014, assuming at least 10mil each for MS/Sony - and that might not even happen) will barely cover the the Dec 31 payment due to Global Foundry for again, AMD management making stupid deals years ago for FAR too many wafers than they'd need yearly.
http://www.tomshardware.com/news/Wafer-Supply-Agreement-Amendment-Globalfoundries-promissory-AMD,19632.html
This is still due. They've been paying fines (basically, termination fees etc) yearly for their mistakes. When we they be out from under GF? I suspect they'll negotiate again to delay but a good portion will be due (125mil? more?).

At 10% (less actually, they said single digits for a while) on $100 chips (I think they're lower actually), x 20mil units (10mil each sony/ms if lucky), they can potentially make 200mil roughly in the next 12 months. So break even on the GF payment next month if consoles do great for the next 12 months. NOT GOOD. You can see why NV said not worth it. When will it reach 20%? 3yrs? 4yrs? Even at double the current rate that's only 400mil profits. With the current situation (no profits on cpus or gpus) that still leaves you 175mil shy of NV's profits on GPU's alone, and that's taking a total loss for Tegra right now until T4 ramps in the next few months as they switch from T3 fully. Once tegra revenue breaks 1B, they will be profitable (guessing T5 does this, if not T4 once ramped fully Q1 or so). Next year should break 1B with tegra (either T4 or combo with T5) as the desktop meets mobile with kepler. AMD had better start making money on cpus & gpus or they will never catch NV in R&D again.

These numbers are the facts of life, not fanboy crap like your two posts. I'm speaking as a stock purchaser (currently own NV shares) who has owned both over the years multiple times. AMD needs to come up with some big winners and quick. Consoles will take a few years to materialize anything substantial if they get there at all as mobile moves to 20nm gaining power. Nevermind the 16/14nm we'll hit in 2015/2016 that will enable 1080p and likely catch next gen's console power (Intel/Samsung/TSMC all look like they'll be on 16/14nm by end of 2016, Intel probably early). The chips don't have to catch them alone, they will have die shrunk memory with faster speeds, larger caches etc as everything shrinks in a tablet/phone etc.

Quoting Charlie really isn't a good way to establish your point anyway. There is a reason the site is called SEMI-ACCURATE...ROFL. He also hates Nvidia beyond belief.

 

[DjEaZy]

"Nvidia showed a ton of excitement and wanted to get involved" ... hmmm... of course... they are out of the console cycle... the next Nexus devices don't uses tegra... and windows RT is a gigantic failure... so... nVidia is desperate to buy it self in any ecosystem... PC games will be ports from PS4 and XBOne... the architectures will be AMD's GCN all the way... Android games will not be especially optimized for tegra because tegra represents just a small portion of SoC used by manufacturers, who adopt android...
... and another question is... when indi's and big giants will do games for PS4 and XBOne, who of them will it optimize especially for SteamOS, that will use especially and maybe exclusively intel+nVidia? And SteamOS seems to be a specialized Linux distro for high-end gaming... so the challenge is to port the game from one architecture that uses parts from one manufacturer AMD, and optimize it for a system, that uses parts from two different committing manufacturers intel and nVidia and on a different OS... maybe will see...
... maybe i'm the only one, but nVidia seems to be desperate... and they need to do something to stay relevant in gaming... build better cards... adopt some of the AMD technology, because of the console ports... maybe share some some tech or make it open-source... cuda haz small relevance because of OpenCL... maybe open up Physx... and if nVidia haz something like AMD's Mantle, the implementation of that will be not so big, because AMD's GCN is all over the console space...
... ok... i better stop mumbling... the bottom line is, competition and innovation is good..

 

[DjEaZy]

"Nvidia showed a ton of excitement and wanted to get involved" ... hmmm... of course... they are out of the console cycle... the next Nexus devices don't uses tegra... and windows RT is a gigantic failure... so... nVidia is desperate to buy it self in any ecosystem... PC games will be ports from PS4 and XBOne... the architectures will be AMD's GCN all the way... Android games will not be especially optimized for tegra because tegra represents just a small portion of SoC used by manufacturers, who adopt android...
... and another question is... when indi's and big giants will do games for PS4 and XBOne, who of them will it optimize especially for SteamOS, that will use especially and maybe exclusively intel+nVidia? And SteamOS seems to be a specialized Linux distro for high-end gaming... so the challenge is to port the game from one architecture that uses parts from one manufacturer AMD, and optimize it for a system, that uses parts from two different committing manufacturers intel and nVidia and on a different OS... maybe will see...
... maybe i'm the only one, but nVidia seems to be desperate... and they need to do something to stay relevant in gaming... build better cards... adopt some of the AMD technology, because of the console ports... maybe share some some tech or make it open-source... cuda haz small relevance because of OpenCL... maybe open up Physx... and if nVidia haz something like AMD's Mantle, the implementation of that will be not so big, because AMD's GCN is all over the console space...
... ok... i better stop mumbling... the bottom line is, competition and innovation is good..

 

[somebodyspecial]

I'm just trying to be devils advocate here. Its more than enough for the type of system I would run, and can run a 4770 and a Titan.

But 80+ Gold means its 87- 90% efficient typically. So that equates to a 450W power supply being able to deliver effectively ~392-405W of total Power under max load. That leaves <50W headroom for potential power hungry potentially hardware in the future. Based on recent advances with Haswell and newer GPUs being more energy efficient, you would think that 50W could be enough. But you never know how future hardware will playout (overclocking, more advanced HW etc). You never want to make sacrifices on your PSU.

Since Valve has three tiers of hardware, they can offer a slightly higher wattage PSU for the enthusiasts out there.

I am trying to be the devil's advocate too some times 😛.
But here there is something you misunderstood. Efficiency isn't about how much power will deliver to the entire system. But how much power it will draw in order to deliver this power to machine.
Take an example of Silverstone's 450watt plus gold at HardwareSecrets at load tests. For 449,8watts it draws 525,3watt. Now if it were bronze it would draw more watts for these 450watts which will be delivered to the system.
So with 450watt PSUs there is room more than enough for these machines.

Who really cares about this? At 14nm in 2016/2017, only what like 3yrs after it comes out, I can drop in a gpu that will likely smoke a 780 at less watts. I'd consider it a total failure it two die shrinks don't add 50% perf at the same watts in a few years. It's only needing to push 1080p so two die shrinks later those will be some serious upgrades allowing all kinds of fabtastic graphics effects and probably 6GB of memory for super huge textures etc. I'm sure the drivers for steamos will allow changes or at least in game so you can have multiple levels of perf (so old gens can just drop a few effects like on windows to get fps up in newer games in a few years). Fortunately for Valve they are dropping their consoles at the same time a die shrink is about to hit so barely into it's life they'll be upgradable to another die shrink.

I wish consoles would have waiting for 20nm which would have negated the need to stay at 30fps (according to carmack, epic etc, 60fps is out of sight). I might have considered a console with 20nm, but without it they'll be dicking around with features much like I do with a 5850 at 1920x1200 on my 24in (more than a few games end up on my 22in 1680x1050 with everything maxed). Instead I'll wait out T5 tablets or shield 2 and stream my future 20nm Maxwell to the TV rather than buy a console again. PSU's in this steambox won't be a problem on arrival or later. Sure there will be cards you can't pop in the box, but whatever is out in 3-4yrs will be far better than what comes in it. You can't upgrade an xbox1/ps4 at all. I'm surprised going with a more PC oriented box for both xbox1/ps4, they didn't design with an Jaguar rev 3 upgrade in mind. The cpu should have been removable or something. They really needed a way to get to 4K before 8yrs.

 

[DjEaZy]

"Nvidia showed a ton of excitement and wanted to get involved" ... hmmm... of course... they are out of the console cycle... the next Nexus devices don't uses tegra... and windows RT is a gigantic failure... so... nVidia is desperate to buy it self in any ecosystem... PC games will be ports from PS4 and XBOne... the architectures will be AMD's GCN all the way... Android games will not be especially optimized for tegra because tegra represents just a small portion of SoC used by manufacturers, who adopt android...
... and another question is... when indi's and big giants will do games for PS4 and XBOne, who of them will it optimize especially for SteamOS, that will use especially and maybe exclusively intel+nVidia? And SteamOS seems to be a specialized Linux distro for high-end gaming... so the challenge is to port the game from one architecture that uses parts from one manufacturer AMD, and optimize it for a system, that uses parts from two different committing manufacturers intel and nVidia and on a different OS... maybe will see...
... maybe i'm the only one, but nVidia seems to be desperate... and they need to do something to stay relevant in gaming... build better cards... adopt some of the AMD technology, because of the console ports... maybe share some some tech or make it open-source... cuda haz small relevance because of OpenCL... maybe open up Physx... and if nVidia haz something like AMD's Mantle, the implementation of that will be not so big, because AMD's GCN is all over the console space...
... ok... i better stop mumbling... the bottom line is, competition and innovation is good..

 

[DjEaZy]

"Nvidia showed a ton of excitement and wanted to get involved" ... hmmm... of course... they are out of the console cycle... the next Nexus devices don't uses tegra... and windows RT is a gigantic failure... so... nVidia is desperate to buy it self in any ecosystem... PC games will be ports from PS4 and XBOne... the architectures will be AMD's GCN all the way... Android games will not be especially optimized for tegra because tegra represents just a small portion of SoC used by manufacturers, who adopt android...
... and another question is... when indi's and big giants will do games for PS4 and XBOne, who of them will it optimize especially for SteamOS, that will use especially and maybe exclusively intel+nVidia? And SteamOS seems to be a specialized Linux distro for high-end gaming... so the challenge is to port the game from one architecture that uses parts from one manufacturer AMD, and optimize it for a system, that uses parts from two different committing manufacturers intel and nVidia and on a different OS... maybe will see...
... maybe i'm the only one, but nVidia seems to be desperate... and they need to do something to stay relevant in gaming... build better cards... adopt some of the AMD technology, because of the console ports... maybe share some some tech or make it open-source... cuda haz small relevance because of OpenCL... maybe open up Physx... and if nVidia haz something like AMD's Mantle, the implementation of that will be not so big, because AMD's GCN is all over the console space...
... ok... i better stop mumbling... the bottom line is, competition and innovation is good..

 

[jimmysmitty]

"Nvidia showed a ton of excitement and wanted to get involved" ... hmmm... of course... they are out of the console cycle... the next Nexus devices don't uses tegra... and windows RT is a gigantic failure... so... nVidia is desperate to buy it self in any ecosystem... PC games will be ports from PS4 and XBOne... the architectures will be AMD's GCN all the way... Android games will not be especially optimized for tegra because tegra represents just a small portion of SoC used by manufacturers, who adopt android...
... and another question is... when indi's and big giants will do games for PS4 and XBOne, who of them will it optimize especially for SteamOS, that will use especially and maybe exclusively intel+nVidia? And SteamOS seems to be a specialized Linux distro for high-end gaming... so the challenge is to port the game from one architecture that uses parts from one manufacturer AMD, and optimize it for a system, that uses parts from two different committing manufacturers intel and nVidia and on a different OS... maybe will see...
... maybe i'm the only one, but nVidia seems to be desperate... and they need to do something to stay relevant in gaming... build better cards... adopt some of the AMD technology, because of the console ports... maybe share some some tech or make it open-source... cuda haz small relevance because of OpenCL... maybe open up Physx... and if nVidia haz something like AMD's Mantle, the implementation of that will be not so big, because AMD's GCN is all over the console space...
... ok... i better stop mumbling... the bottom line is, competition and innovation is good..

Two years ago is a long time. It probably was well before there were any finalized hardware specs for the PS4 or XB1 meaning NVidia could have wanted in this without knowing if MS or Sony were going to stick with PowerPC/Cell or move to x86.

On top of that, the Surface just came out last year with the release of WIndows 8, so again this was starting with NVidia and VALVe before Windows RT had the chance to flop.

And what about the fact that SteamOS is open source do people not understand? Its not tied to any hardware but if VALVe chooses to only ship SteamBoxes with Intel and NVidia, so be it. That's their choice much like Dell could ship only AMD if they wanted.

Also, people need to stop assuming just because the consoles have x86 arch from AMD that it will instantly run better on PCs with AMD hardware and require optimizations for other hardware. It will require optimizations because in spite of the fact that it has a 8 core CPU, its a low end version without all the same attributes as a normal desktop processor. While the XB1 will use a Windows 8 kernel and DX, it has the EDRAM for the GPU which is not on PCs and therefore is another layer that will have to be removed for the PC version.

I agree that competition is good but too many assumptions get made without facts. The fact is that VALVe decided to go with NVidia (probably due to drivers) and the consoles mean squat when it comes to PC games except that it will be easier to port to PC so we will get more games sooner rather than later.

 

[renz496]

well what i really want to know is what Gabe take on AMD Mantle. there are already speculation that Mantle will be perfect match to this Valve SteamOS and Steam box to the point it might make MS and Sony hostile towards Mantle. so i want to know what Gabe think about AMD Mantle

 

[Mupples]

Valve got themselves into a contract and now they have to justify their contractor's choice in hardware changes. Origin had their hands in two pockets and pulled a dirty, they were originally an AMD company for years and worked with Mass Luminosity, and then they decided just recently to support negative-progress in game industry development. I wouldn't worry though, Valve is making the OS open for everyone and they don't like proprietary technology. So SteamOS will most likely work on AMD hardware. Origin is doing this to make quick cash, fucking greedy company.

 

[somebodyspecial]

"Nvidia showed a ton of excitement and wanted to get involved" ... hmmm... of course... they are out of the console cycle... the next Nexus devices don't uses tegra... and windows RT is a gigantic failure... so... nVidia is desperate to buy it self in any ecosystem... PC games will be ports from PS4 and XBOne... the architectures will be AMD's GCN all the way... Android games will not be especially optimized for tegra because tegra represents just a small portion of SoC used by manufacturers, who adopt android...
... and another question is... when indi's and big giants will do games for PS4 and XBOne, who of them will it optimize especially for SteamOS, that will use especially and maybe exclusively intel+nVidia? And SteamOS seems to be a specialized Linux distro for high-end gaming... so the challenge is to port the game from one architecture that uses parts from one manufacturer AMD, and optimize it for a system, that uses parts from two different committing manufacturers intel and nVidia and on a different OS... maybe will see...
... maybe i'm the only one, but nVidia seems to be desperate... and they need to do something to stay relevant in gaming... build better cards... adopt some of the AMD technology, because of the console ports... maybe share some some tech or make it open-source... cuda haz small relevance because of OpenCL... maybe open up Physx... and if nVidia haz something like AMD's Mantle, the implementation of that will be not so big, because AMD's GCN is all over the console space...
... ok... i better stop mumbling... the bottom line is, competition and innovation is good..

You're under the mistaken impression consoles rule this gen. They don't. Check GDC survey, they are being ignored. Only 11% are even planning a project for next gen boxes. 60% are making games now and planning more for MOBILE. 40% for PC. Devs aren't making this stuff up, this is actually what they are doing. You're comment would have made sense for LAST gen...Not next gen which everyone will NOT be programming for from the get go...It's wait and see due to mobile taking over. Take a good look at what is happening to Vita, 3DS and Wiiu. All having problems getting devs. MS is having the same problem for xbox1 which is why they announce (in desperation as you say) their own 1Billion for their console games. That is because NOBODY but a few big devs are planning anything.

NV isn't desperate, they went the smart way where all the devs are going. SteamOS isn't exclusive to anyone.

Consoles are dead (from MS/Sony/NIn that is). Mantle was DOA.

Cuda is taught in 640+ universities and on the top 200+ pro apps. Some sort of relevance?...ROFL. OpenCL has no relevance as it isn't backed by ANY real money (amd is broke). Cuda has been backed for 7+yrs with billions (that's how you get it into 200+ apps and taught in 640 schools). This is why to show OpenCL you have to run crap like bitmining and Folding@home...ROFL. Maybe soon Adobe which is why toms ignores it now (refuse to run CUDA vs. AMD in adobe), until AMD gets done paying Adobe to actually show something we can USE with opencl...ROFL. Rightmark? etc...Same.. It's not a pro app, you can't make money on it. Almost all of toms crap can be done in Cuda. They just choose to ignore it because they want an open pro app society.

I just want FAST stuff. Don't care who wins the war, just give me the best software. Right now that's Cuda. OpenCL has very little chance of success. Mantle has NONE. It cost AMD 8mil for BF4 optimizations. Yeah, I see them being able to afford that a lot more times...Being broke that should be easy. MS/Sony will block mantle from their boxes. It helps people get off Windows/Consoles/DirectX. They will fight Mantle just like Carmack says.