Graphics
Intel® graphics drivers and software, compatibility, troubleshooting, performance, and optimization
20498 Discussions

Bug in Intel Vulkan driver

attx
Beginner
3,792 Views

It is currently impossible to use a reverse-z technique with your Vulkan driver as it completely ignores minDepth and maxDepth fields in VkViewport structure.

The driver always defaults to minDepth = 0.0f and maxDepth = 1.0f, even if both fields are set to minDepth = 1.0f and maxDepth = 0.0f, producing wrong depth buffering results.

This technique is mentioned in one of your tutorials:

https://software.intel.com/en-us/articles/api-without-secrets-introduction-to-vulkan-part-3 API without Secrets: Introduction to Vulkan* Part 3: First Triangle | Intel® Software

More info on Khronos pages:

https://www.khronos.org/registry/vulkan/specs/1.0/man/html/VkViewport.html https://www.khronos.org/registry/vulkan/specs/1.0/man/html/VkViewport.html

Drivers affected:

-15.45 and 15.46 on Windows 7 x64 / Windows 10 x64

Hardware used:

-Intel HD 530 (Intel Core i7 6700k)

NVIDIA / AMD hardware works as expected, only Intel is affected.

SSU scan attached.

0 Kudos
20 Replies
RonaldM_Intel
Moderator
2,096 Views

Hello thom36,

Thank you for reporting this issue to us. If it is not asking too much, can you add the issue details using the table format as detailed here:

Once we have this info I'll report it directly to our development team.

Kindest Regards,

Ronald M.

0 Kudos
attx
Beginner
2,096 Views

Thank You, Ronald.

Category

Questions

Answers (N/A if not applicable)

Description

Provide a detailed description of the issue

Intel's Vulkan driver ignores minDepth and maxDepth fields in VkViewport structure, always defaults to minDepth=0.0f and maxDepth=1.0f disallowing to use a reverse-z technique

Please place an X to the right of the option showing how often you see this issue using specific steps. (Ex: 'Every few times a game is started it flickers.' <- This would be "Often")<p> 

Always (100%): X

Often (51-99%):

Sporadic (20-50%):

Very Sporadic (<20%):

Hardware (HW)

Brand and Model of the system.

Custom built desktop PC

Hybrid or switchable graphics system?

 

ie Does it have AMD or NV graphics too?

PC also has GeForce GTX 960 in PCI Express slot, monitor cable connected to Intel HD 530 (i7 6700k)

Make and model of any Displays that are used to see the issue (see note2 below). 

LFP = Local Flat Panel (Laptop panel)

 

EFP = External Flat Panel (Monitor you plug in)

Tested on 3 monitors:

ASUS MX239H EFP

SAMSUNG 743BM EFP

SAMSUNG 797MB EFP

but the issue can be observed on any display

How much memory [RAM] in the system (see note2 below).

16GB

Provide any ...

0 Kudos
RonaldM_Intel
Moderator
2,096 Views

Hello thom36,

I apologize for not replying sooner, I try to stay away from the "we're working on it" kind of updates.

With the information you provided, I filed internal bug ID 1406462374, and our devs were able to reproduce the issue.

Today I was informed that we have a fix checked-in, so please look forward to our next 15.46 and higher driver releases.

Thank you for taking the time to report this issue to us.

Kindest Reegards,

Ronald M.

0 Kudos
attx
Beginner
2,096 Views

That's an excellent news, thank you for doing a fantastic job.

I will be waiting for the next driver update.

0 Kudos
attx
Beginner
2,096 Views

I installed the latest update that has been released yesterday (15.47.02.4815 (22.20.16.4815)) but I see no changes whatsoever.

Could you please check if the fix has been included in the mentioned update?

With viewport.minDepth=1.0f and viewport.maxDepth=0.0f, depth that should be written for a triangle with depth=0.95f is 0.05f, not 0.95f.

I use VK_FORMAT_D32_SFLOAT_S8_UINT format.

0 Kudos
RonaldM_Intel
Moderator
2,096 Views

Hi again thom36,

Thank you for your reply. Let me check with our devs, I think the fix was checked-in after this particular driver package was finalized, but need to confirm it first to be sure.

I'll let you know once I have more details.

Best Regards,

Ronald M.

0 Kudos
RonaldM_Intel
Moderator
2,096 Views

Hi thom36,

I got confirmation from my dev that this fix is going to be included in the upcoming 15.60 driver baseline, which should be coming out sooner rather than later. Unfortunately I cannot provide any ETA since release dates can shift at any time.

As I have more information I'll post it here.

Best Regards,

Ronald M.

0 Kudos
attx
Beginner
2,096 Views

Thanks.

Will 6th Gen be still supported in 15.60 baseline?

0 Kudos
attx
Beginner
2,096 Views

I just tested the latest 15.60 release and I can confirm the bug I reported has been sucessfully fixed.

6th Gen Core processors are supported in this baseline.

0 Kudos
RonaldM_Intel
Moderator
2,096 Views

Hi again thom36,

Ahhhh you beat me on the reply time!

I am happy to hear that this is working fine now with 15.60.

Once again, thank you for taking your time reporting this issue and helping us make a better product.

Kindest Regards,

Ronald M.

0 Kudos
Parapele
New Contributor I
2,096 Views

Hi Ronald_Intel, Xavier told me i should asking help in this thread. Yups and i have the same issue, after installation 15.60.0.4849 driver, Vulkan reports 1.0.0 version API. All programs that check level support of API says that Vulkan driver is 1.0.0 instead 1.0.61, even Intel HD Graphics Control Panel reports 1.0.0. I tried everything, nothing helps. All system scans, you can take from this /message/507600 thread.

0 Kudos
idata
Employee
2,096 Views

Hello Parapele,

 

 

Let me do further research about this and come back to you. I will post information on this thread as soon as I have it.

 

 

Regards,

 

Fred D.
0 Kudos
Parapele
New Contributor I
2,096 Views

Hello Fred! Just quick update. You probably have noticed, some Intel driver files different and wonder why. That's because first time i installed driver 11/11/2017, found that Vulkan broken, then downloaded and installed driver again - nothing change. Today i decided check MD5 and found that it different. Driver zip package was updated when Bryce made official post 14/11/2017. Old zip 4849 MD5 - 081FA73B48D6F4B6A1F1DE5F0540D2E2, new zip 4849 MD5 30EBDB2ADD1D8778F180A544522B842E. Difference between driver package.

Installation_Readme.txt says only difference is Intel(R) Display Audio Driver: 10.00.00.1 (old) and Intel(R) Display Audio Driver: 10.24.00.01 (new) So, in the world now exist two driver package and it can be very confusing. After installing new driver nothing change. I attached new scans files, just in case..

0 Kudos
idata
Employee
2,096 Views

Hello Parapele,

 

 

I believe we are targeting for a new 15.60 release in the next couple of weeks, so I would suggest checking our site and let us know once this driver is available if it has any issue. I tried here internally and so far it looks good, no conflicts in the versioning.

 

 

Can you check how many different Vulkan packages you have installed under "Programs and features"? This can help me compare with my systems here just in case I'm missing something.

 

 

Regards,

 

Fred D.
0 Kudos
Parapele
New Contributor I
2,096 Views

Hi Fred.

It continues with VulkanRT1.0.54.1-2 and finally VulkanRT1.0.61.0

So driver not can install properly, because version conflict? Should i deleted something?

0 Kudos
idata
Employee
2,096 Views

Hello Parapele,

 

 

Thanks for the information you sent to me.

 

 

There is no need for you to delete anything. I will be posting information on this forum thread as soon as I have it available.

 

 

Regards,

 

Fred D.
0 Kudos
Parapele
New Contributor I
2,096 Views

Hello Fred! I happy to report, that new driver 4877 indeed gives Vulkan support back. Xenia and RPCS3 still the same, but i think that emulator problem. Case can be closed.

Fred, can you help me get response in , i left alone there and problem still persist in 4877 driver.

0 Kudos
idata
Employee
2,096 Views

Hello Parapele,

 

 

I am glad to hear Vulkan issue has been fixed.

 

 

I will check what happened with the other thread, but the updates will only be posted on that thread.

 

 

Regards,

 

Fred D.
0 Kudos
Parapele
New Contributor I
2,096 Views

Fred, i still not get response in that thread.

0 Kudos
RonaldM_Intel
Moderator
1,856 Views

Hi Parapele,

An update as been posted in thread

I'm sorry for the delay.

I'm locking this thread since the original author confirmed the issue that is about is already fixed.

Best Regards,

Ronald M.

0 Kudos
Reply