Pimax 8K/8K X 90Hz hurdles [for tech savvy]

Pimax 8K Series
1

I didn’t find clear answers so far on one place so I spent some time decoding bits of information on this forum. In apology letter we got some clues, but not real explanations why HMD can’t hold 90Hz even if LCD panel used is proven stable using Qualcomm platform(mobile) at that frequency.

What I found out, is that panel is proven stable at 90Hz only by using data compression technology（Qualcomm FBC 2.0 or VESA DSC 1.1）source . I was also confused why would SW engineer say that they need 2xDP because ANX7530 support DSC on MIPI but has only 2 slices source. So I guess original idea was to use 2xDP on Pimax 8K, to go around limits of the chip, but it was proven not possible(pixel clock(HBR2), and DSC limits).

Why did Pimax made decision to try using 1xDP on 8K, and 2xDP on 8K X is probably to reduce cost, and since ANX7530 can’t provide even with 2xDP solution option to have native 2x4K@90 input, because Pixel clock limit of 720MHz and lack of support for DSC on DP input, it was regarded acceptable to make more sacrifices, refresh rate being in 75-85Hz range, and using scaler is a must.

My guess is that Pimax 8K was envisioned much like the 8K X, but due to limitations of current tech, it is decided to split solutions into 2 products. If you wonder why 8K X is postponed to Q2 2018, solution to support native 2x4K@90 is by using DSC on input, Analogix is still developing chip with such feature source.

Also to me is unknown if using DSC(on MIPI) prevents scaler implementation, that can explain why not use 2xDP on 8K to solve 90Hz issue(with 2xANX7530).

To note, DSC 1.2 is part of DP 1.4 standard, so only latest cards are capable of taking advantage of this tech to drive 2x4k@90 on 8K X HMD. Also to note, there is DSC on MIPI and DSC on DP, ANX7530 has DSC encoder for MIPI, don’t support DSC on DP input.

I don’t blame Pimax for anything, maybe it is unfortunate to find out that without DSC/FBC panels can’t operate at 90Hz, but I expected them to start campaign with Prototype V1 that is using 1xDP why even bother with 2xHDMI I am yet to decode :slight_smile: . So V3 had to be V1, that is my point, are there too many compromises with final 8K HMD we are yet to find out - when backers get their units.

About myself, former CS student. Not a backer, but interested in VR development, would love to see how Pimax solves some of tech challenges on road to next gen VR.

Maybe I am wrong, because there is also one part that I don’t understand in apology letter, it is noted that 3840x2160@80 is native res per eye tested and stable, we know that V3 is using 1xDP and ANX7530 with scaler, so I guess for that testing one screen is only tested with input res of 3840x2160@80(scaler off), but they wouldn’t be able to test 90Hz since chip is limited to 720MHz pixel clock, and when transferring 4k@90 pixel clock is 811.44 MHz. So not sure if there is way to test my theory that LCD without DSC on MIPI can’t hold 90Hz, or maybe it is just Pixel clock that is limiting factor.

Maybe Pimax didn’t want to share this because it would bore backers to death, who knows, I wrote warning (for tech savvy) xD

9 Likes
2

Much appreciated - thank you.

3

Why would they advertise 90hz if they weren’t sure they could do it, out of curiosity. (?)

It seems as though they’ve known about the limitations of the hardware and known exactly which hardware they were going to use. To us this stuff seems complex, but to an engineer it’s fundamentals. I personally think they have a game plan, they said 90hz with a game plan, they decided to advertise 180hz brainwarp because they have a game plan. They wrote whoops apology letter because they have a game plan. You get the picture. While they may be a mom and pop virtual reality company compared the likes of the giants, they arent totally new to business, I’d bet money on the fact they’ve been getting help from valve too.

Side note, why care about 90 vs idk…84? Im concerned about how openvr games already made have been developed, steamvr reprojection algorithms i dont think are flexible and I have concerns about brainwarp being able to work on ANY game period. So 90 I think is important. 90 is also scary. It’s pushing the boundaries of the displays “happy zone”. So getting your shiny new pimax and marathoning a ten hour vr party and whoops the screens fried. Its a catch 22 seemingly. But maybe this is hopefully part of a plan. Im truly hoping we dont have the full story at all.

4

Same reason manufacturers like Gigabyte advertises theorectical specs. If all componets run at ideal specs you will get peak performance

Good audio makers for example advertise amps at rms instead of peak power.

Now @SpecsReader did miss the post that tge 5k is stable at a higher refresh than the 8k. Which to me suggests the bridge chip starts to be become unstable earlier but the upscaler is amplifying the instability resulting in lower stable refresh than the 5k.

We see things like this often in pc as you cam have 2 identicle pcs but one performers better than the other & is more overclockable.

5

That’s the panel that Pimax used in the Pimax 4k. We don’t even know for sure which panel Pimax is using for the 8k. A pimax employee told a visitor in Tokyo that they got the panels from a Japanese firm, so it’s very likely that it’s Sharp again. In fact, Sony used a new Sharp panel in their latest Xperia 4k phone, so I’m guessing that Pimax is using that exact same panel. But again, it’s just speculation and either way we don’t have any specs for it at all. So we don’t know if the panel supports 90hz uncompressed.

1 Like
6

@Sjef
To be honest there are a lot of assumptions, due to lack of feedback from Pimax regarding details. But it’s safe to assume it is same panel used in 4K(Sharp) just tuned(by Pimax or other company to get to CLPL state).
Problem is debugging or testing process, to pinpoint exact cause of instability, is from my understanding very tricky, because you can’t test with native input(on DP) to panel at 90Hz, because DP to MIPI chip is limited at 720MHz, can’t push 4k@90 to panels, scaler is used to feed panels native res(via upscaling) but in that case you won’t know if limit is scaler or panel. One user pointed out that since these panels are made for mobile, they are not designed to handle native input res without compression(hard to expect mobile to push native 2160p@60 let alone 90, due to power constraints).

As @Heliosurge pointed out, I didn’t take into account 5K having issues as well to reach 90Hz, that can mean scaler is not the issue to reach 90, maybe 82 is stable on 5K not on 8K but that is far off 90Hz. We have even less details about that screen, does is support DSC on MIPI with 1 slice per panel etc. I will assume it is taking native res from ANX7530 without DSC on MIPI, so at 2x1440p@90 pixel clock is close to 720MHz probably reason for instability.

Will we get more details from Pimax this week about this issue, I hope, just to keep transparency promise. @Matthew.Xu ?
EDIT: Update about V4 published, no progress on 90Hz, we will probably not get more details until V4 demo time(2018) . No easy fix possible probably, as expected.

2 Likes
7

MIPI BW analysis:
Theory work, may or may not have implications with 8K HMD.

Native 4k@90/eye input:
3840x720@90Hz input per eye(DSC 3:1 on 3840x2160@90)
2xMIPI per panel
4xLanes per MIPI port, 4x1.5Gbps = 6G/s

3840x360@90Hz at top mipi: = 138240090 =124416000 pixel/s = 125Mp/s = ~3Gbps(2.985984Gbps (1244160003RGB8*bit/px) )
With DSC, since data rate is <3Gbps, we can send this data faster to DSC decoder, how?
Use the fact that we have 4 lanes(per MIPI), with 1.5Gbps each. Split 3840x360@90 into 2x3840x180@180, so DSC decoder can get data in 1/2 of time so it can drive whole panel in native 4k@90Hz.

1440p@90hz per eye input(upscaling no DSC on MIPI)
2x mipi ports per panel/eye
12801440@90hz at top mipi: = 184320090 = 165888000 pixel/s = 166Mp/s = 3.7078857421875 Gbps
We need at least 3 lanes so we can’t use double frequency to give Driver IC data fast enough for 90Hz refresh on panel.

ANX7530 has DSC encoder, but DSC on MIPI is not supported if you use chip in quad-MIPI mode, as it is with 1DP solution.

Idea is to therefore use 2xDP to enable DSC on MIPI at least, since we can’t use DSC on DP input. And that way we can feed the Driver IC faster data to enable 90Hz refresh rate even with upscaled input.

1440p@90Hz per eye input (upscaling) with DSC on MIPI
2560x1440@90 into DSC encoder, 2560x480@90Hz out
2560x240@90Hz per MIPI, we should be able to use 4 lanes to reduce time to transfer data to DSC decoder then we feed the scaler with 2560x1440 .

My (wrong?) assumption was that in Pimax 8k V1 and V2 by using 2xHDMI bridge chip, DCS was not in use(on MIPI), and that input was 60Hz output async 90Hz(per eye). If I read apology letter again, they say games did run at up to 90Hz, games are rendered in FPS, but nevermind, input was probably 90Hz to HMD, and panel according to specs can run at 90Hz only with DSC or in Async mode.
I couldn’t find specs on chip used for HDMI to MIPI in V1-V2, so I guessed they were saying 90Hz/eye async. Something like HMD with Brainwarp ON should work.
They also wrote that panel was proven stable at 90Hz with Qualcomm platform, and Qualcomm has version of DSC called FBC. I was wary that it is 90Hz sync, but after thinking about DSC on MIPI, how it uses 3Gbps and 6Gbps is available, should in theory give headroom to handle 90Hz input and 90Hz output(sync - real 90Hz).

Would like someone from Pimax team with tech knowledge fill in the blanks. If not, this is for people who want to find out why is 90Hz not as stable with 1DP comparing to 2xHDMI versions - and are tech savvy :slight_smile: .

3 Likes
8

Since you seem to be a tech expert, can you please explain how the async refresh rate would work in a configuration with 2 screens? From what I understand, the async 90hz featured in the Pimax 4K was accomplished much in the same way as brain warp has been described, except using only 1 display rather than 2 as the 8K uses. If this is the case, the async refresh rate concept in relation to the 8K is completely ridiculous and makes no sense.

So, let’s hear your expert explanation now about a subject you can’t just copy paste information from spec sheets on.

http://community.openmr.ai/t/what-does-asynchronous-90hz-mean/80/2?u=brian91292

BTW- [quote=“SpecsReader, post:7, topic:4562”]
and panel according to specs can run at 90Hz only with DSC or in Async mode
[/quote]

Source? You are making a lot of wild claims with no source. Has Pimax even disclosed the exact display they are using? How are you pulling specs for hardware that has yet to be revealed?

2 Likes
9

Async 90Hz in 4K and async 180Hz in 8K (they call that mode Brainwarp) is same principle per eye. You feeed in signal at 60Hz, output at 90Hz per eye, you get async frames per eye, turn of backlight/use shutter to close 1 eye for 1/180ms, and they assume brain will perceive 60Hz input as 90Hz input. You don’t need brainwarp if have sync(native) 90Hz input and 90Hz output per eye.

Source(regarding 90Hz and DSC/FBC), well I just connect dots, Qualcomm platform, Qualcomm FBC, or you expect mobile system to feed native res to panel at 90Hz(no DSC) ? This is mobile screen after all, not designed for VR, 4K screen was designed for Sony Xperia Z5 Sony Mobile Phones . About 90Hz and async, from 4K, it safe to assume if that model of panel could have 90Hz async so can new one.

If my theory is proven wrong, by releasing specs(by Pimax team), then ok, but you don’t prove my theory wrong, just claim I don’t have actual specs to make theory. I have some specs, some data from PimaxSWD, some data from Apology letter, some data about 4K HMD, about VR in general, and about computers in general (former CS student). You make new theory, prove me wrong if you want :smiley:

2 Likes
10

No reason no to be civil fellas.

1 Like
11

What is not civil about our comments? He thinks I can only read and tell specs, not interpret missing data from Pimax about real reasons 90Hz is not as stable with 1x bridge chip version(1DP) comparing to 2x bridge chip version(2xHDMI). I have every right to guess/theorize about it until Pimax give us explanation. Not just throw in 10 reasons, as cable/pixel clock/EMI etc. PimaxSWD - actual engineer, not marketing person, said they plan 2xDP(on 8K) because of no DSC on MIPI with 1xDP. Maybe they decide to stick to 1DP and try roll without DSC on MIPI but run into issues above 82Hz, and this will not resolve 90Hz issue.
You can wait for their explanation, or accept mine, or disregard mine. I don’t force anyone on anything. Just write my understanding of known tech limits to native 4K/eye VR.

4 Likes
12

You must be thick skinned because I have definitely seen some not so civil comments aimed in your direction as of late. I really appreciate your posts, even if I don’t fully understand everything you are trying to explain. I hope you continue to post from the computer science and theory side of the coin. Makes for interesting reads and as of late some colorful commentary.

5 Likes
13

Yea, I get used to it by now, I throw in actual tech perspective on the problems, and get attacked by people without counter-argument, they justify Pimax lack of sharing the tech details to public about problems. And justify putting marketing on place where you need real number(spec) like 75/82/90 not like they did, 90/eye(8K), or 90/async(4K) . So by the looks of things, we will get actual specs from backers not from Pimax :smiley: .
I think that is a bit sad, in internet era, if you don’t share some info, it’s usually because you want to hide it. Why hide it? They use excuse like NDA, but we don’t ask about panel maker(who is), but actual specs of panel/driver IC. I hate hiding the specs, hence my name xD

1 Like
14

I tend to lean this way as well. At this point, I’m already a backer so it does not matter. If there is some less than appealing news, I would rather they just break it to me rather than hoping I do not notice. Nothing worse than a bait and switch whether literally or actually.

1 Like
15

8K or 8K X ? 8K X can still turn out decent, from what I read 8K X has chance to be 4k@90Hz (with DSC) per eye input. Now does DSC(compression) effect image quality we don’t know really, in 2D world, where panels are not magnified VESA claims no impact, in VR not so sure… I consider 8K X as DK1 of Gen 2 VR, if it delivers real FOV, and native input res@90Hz . It seems well though off, price is a bit high, but that is price for getting first latest tech.

Pimax 8K so far is changed from original idea too many times. Original idea was 2xHDMI, then 1xDP, then they found out it need 2xDP, then stick to 1xDP and use 2xDP for 8K X(cause 2xDP is probably more costly than 2xHDMI). Compromise, so far is stable refresh rate(82 vs 90). Is Brainwarp put out as idea to compensate for this compromise, time will tell. Because Pimax for sure won’t xD

16

I picked the 8K considering the overall hardware demands for the 8K-X sounded like a challenge. I’m already planning to upgrade to 1080ti or Volta depending on final specs. I already have an HTC Vive and not looking forward to shelling out a few large for the next gen experience. Hopefully I won’t have to.

17

Ok, maybe 8K will be mid gen upgrade like ps4 pro, xbone x, they claim 4K gaming but it’s 30fps or upscaled(mostly), looks like Pimax has same problems, upscaling input and reaching 90fps/hz :slight_smile: . I already wrote somewhere if you have Vive, 8K(only HMD) can be interesting addition even replacement if MTP is low enough for 200FOV to not make owner sick when moving a lot.
Edit: And if you have 1080Ti or plan to buy one, 1070 they recommend is not something I recommend for good MTP in most games. Just from math when comparing rendering at 1.4k/ eye in Vive vs ~4k/eye in Pimax 8k.

1 Like
18

So far no one has ever talked about “async” anything in Pimax 8K. In order to claim all the conclusions your wrote above you need some solid supporting info. Where did you get this info?

19

The problem with @SpecsReader posts is that he tries to disguise his wild theories as the facts.

1 Like
20

I asked Pimax about that . And said if they don’t give info I will assume is same method used in 4K 90Hz async as with 8K 90Hz/eye (async) source. If you read specs about 8K, they put 75/ 90 per eye, then put brainwarp in same line. With brainwarp you can mask async input 60Hz displayed on 90Hz per eye. If you can’t understand my logic, sorry. We just don’t have other explanation, I think mine is good enough.

Explain why bother with per eye, then 180Hz and Brainwarp if it’s 90Hz sync as Rift/Vive/Win MR/PSVR etc? Give me your theory. I listen. I don’t say all my theory is right, I was wrong few times, and corrected over time. With new data, but Pimax provide very little new data so far.