What about intel broadwell, which hardware level is its GPU at?
Tyvm!
Don't get your hopes up, I believe the next Intel gpu was supposed to be feature level 12_0 (if I remember the slide correctly) not broadwell.
All of the language IHVs have used so far is with respect to supporting the DX12 API, not new feature levels. Microsoft has not yet announced new feature level requirements so you guys will have to stay tuned on that front
All of the language IHVs have used so far is with respect to supporting the DX12 API, not new feature levels. Microsoft has not yet announced new feature level requirements so you guys will have to stay tuned on that front
One IHV (Intel) has decidedly mentioned DX12 Feature_Level_12 at least for Skylake's integrated graphics. Technically, Intel said, next gen, but next = Skylake, in contrast to future gen.
http://www.pcgameshardware.de/Direc...rafik-erst-echte-DirectX-12-Hardware-1135515/
That of course does not change anything you said wrt to Broadwell.
Intel just removed Skylake device IDs from the Haswell/Broadwell Win10 driver. And I just looked, Skylake beta drivers also don't have Haswell support, only Broadwell.
So it seems Intel is going to continue the policy of only fully supporting the two latest generations, no matter what their hardware capabilities or similarities.
This means if I plan to keep my laptop for 4-6 years, I have to buy it with a discrete GPU, even though I do not need the performance.
So it seems Intel is going to continue the policy of only fully supporting the two latest generations, no matter what their hardware capabilities or similarities.
This means if I plan to keep my laptop for 4-6 years, I have to buy it with a discrete GPU, even though I do not need the performance.
6 years is quite a while for a laptop, that keyboard's going to be mighty gunky after such a long time. Anyway, if you want a long-lived laptop, buy a Macbook, then you're not a victim of Intel's fickleness. Apple is still actively supporting systems with Core2 processors in them...
I'm not sure that I understand the complaints in the last two posts.
You may not be able to get drivers directly from Intel for your "old" chipset, but even the most modern OS still supports them. I loaded Windows 10 on a Dell Mini10V, which came with the Intel 945GME video chipset. That chipset was launched in the first quarter of 2006, and it now reports a WDDM 2.0 driver under Windows 10.
Why do you somehow think your Haswell is going to be out of date in five years?
More to the point, why do you think Intel-direct video drivers are somehow going to matter much versus the OS-included drivers, after your integrated video is six years old?
You may not be able to get drivers directly from Intel for your "old" chipset, but even the most modern OS still supports them. I loaded Windows 10 on a Dell Mini10V, which came with the Intel 945GME video chipset. That chipset was launched in the first quarter of 2006, and it now reports a WDDM 2.0 driver under Windows 10.
Why do you somehow think your Haswell is going to be out of date in five years?
More to the point, why do you think Intel-direct video drivers are somehow going to matter much versus the OS-included drivers, after your integrated video is six years old?
Could you please check Device manager for the driver date&version and device ID?
I have an equivalent laptop, and it only got a WDDM1.0 driver from Windows update.
The OS included drivers are, IMHO, "bugfix so that they don't crash on a new OS" drivers.
Three reasons why I would like to have drivers from the main branch in 3-4 years:
Features like Miracast that might not be hardware-dependent
If I buy a GT4e laptop, the hardware capabilities might actually stay relevant in games for 3 years, but if I buy it half a year after Skylake is released, it will stop receiving serious optimizations in 1.5 years, while the equally powerful GT840M will still be receiving full support.
DmitryKo
Veteran
And what is the problem, exactly? GMA950 in the 945GM chipset is a DirectX 9.0c part, it doesn't support new features exposed in WDDM/DXGI 1.1... even GDI acceleration requires hardware "cache-coherent GPU aperture segment" to access video card memory from the CPU.
Only Nvidia supports WDDM 1.1 on DirectX 9 hardware, according to MSDN documentation:
Windows 8.1 x86 and x64 graphics driver availability status
The problem is that somebody else with the same GPU claims to have a WDDM2 driver.
It's WDDM2 in name only, that doesnt' mean the featureset has somehow been upgraded to that of a 9-year newer chipset. Hell, it's the same as it claiming any sort of WDDM1 feature capability, to be quite frank. Sure, the driver level might indicate WDDM1, but many of the necessities of that feature level will be emulated by the CPU itself.
I'll grab you a screencap tonight when I get home.
Regardless of ANY of that, what is the point of complaining about Intel not providing their own most-modern driver for it? Nobody in their right mind is going to use a GMA945 for anything harder than desktop compositing and perhaps a solitaire or minesweeper game. Even 480p YouTube videos were occasionally challenging for it, back when it was brand new.
The same point applies to your iGPU seven or eight years from now. I find your original complaint to be trivial at best, or purposefully contrived at worst.
I'm not complaining about GMA950 not having WDDM2. Though do please grab that screenshot.
I'm complaining about Haswell GT3e being out of the main branch already.
And, more importantly, about Skylake GT4e probably getting only two years of performance improvements.
Both Fermi and Radeon HD5xxx cards seem to still be on the main branch (5 years in).
I'm complaining about Haswell GT3e being out of the main branch already.
And, more importantly, about Skylake GT4e probably getting only two years of performance improvements.
Both Fermi and Radeon HD5xxx cards seem to still be on the main branch (5 years in).
If you're worried about your iGPU being "out of the main branch" after only two or three years, then you've never been happy with any Intel iGPU since they started making them. Haswell graphics are a solved problem at this point. The performance you might eke out of iterative driver revision isn't worth the effort, as Intel has obviously decided as such.
The AMD 5000-series of video chips is still in production, in the form of APU's, which is why they continue to get newer updates. The original Fermi series of video cards might still be included in the regular NV driver releases, but when was the last time a graphics performance or bugfix update was included for Fermi that wasn't part of a general solution for the entire driver stack? In other words, the time and money spent on driver-iterating those earliest NV cards isn't being spent on those cards at all, it's simply leftover code that NV has politely declined to remove from the SVN root.
The AMD 5000-series of video chips is still in production, in the form of APU's, which is why they continue to get newer updates. The original Fermi series of video cards might still be included in the regular NV driver releases, but when was the last time a graphics performance or bugfix update was included for Fermi that wasn't part of a general solution for the entire driver stack? In other words, the time and money spent on driver-iterating those earliest NV cards isn't being spent on those cards at all, it's simply leftover code that NV has politely declined to remove from the SVN root.
Not quite. I wasn't that unhappy when those were just GPUs meant for desktop composition. But since Ivy Bridge, they're actually rather usable.
Fermi is getting DX12.
I don't think "DX12" means what you think it means. Here's a hint: the hardware-enabled features that exist for the original Fermi architecture aren't going to suddenly change.
I know that. I was replying to
And in the case of Intel's APUs, DX12 is going to give them more performance no matter what hardware features they have, due to being power-constrained.
I can deny that. I was happy. After two years Intels Gen7 is outdated though.
We are not necessarily talking about Haswell, we are talking about Intels 0-2 years active feature & bug fixes driver policy in general. Some examples from the past. The Sandy Bridge low clock bug (stuck at base frequency) never got a fix for SB because Intel fixed it in a driver branch for IVB and later. Base and turbo frequency is a massive difference, in such games performance is more than doubled. Ivy Bridge also suffered from being dropped from the active driver branch. The AF placebo bug in directx games has been fixed in HSW+BDW driver branch, IVB still can't enable AF in the driver. Also all the software features introduced in 15.36 like CMAA or newer OpenGL extension support. It's a combination of bug fixes, performance, features from a new driver branch. At least Intel should extend it to 0-3 years starting with Skylake.
And on another note, one can understand why Intel does it this way. They have (had?) very limited development resources, and giving more features to older GPUs would probably be done at the expense of less features overall. So up until GM45, I didn't really mind, and with Sandy, I was only irritated by not fixing the aforementioned low clock bug.
The thing is, times are changing. Intel GPUs are staying relevant longer, so I had just been hoping that with Haswell, which is the first to have a GT3e variant, Intel's policy would be changing. And the first DX12 driver, being common to Haswell/Broadwell/Skylake, had seemingly confirmed that.
The thing is, times are changing. Intel GPUs are staying relevant longer, so I had just been hoping that with Haswell, which is the first to have a GT3e variant, Intel's policy would be changing. And the first DX12 driver, being common to Haswell/Broadwell/Skylake, had seemingly confirmed that.