PDA

View Full Version : hd7970 and oc catleap



cirthix
05-15-2012, 09:12 AM
I was playing around with flashing my catleap to an edid wihich only supports 2560*1440@80+hz, and found a few things.

1) 100Hz is displayed properly in bios, thugh I don't know if the mode is actually applied. I'll know tonight when I hook up the oscilloscope to the internal control lines. edit: 100hz is easy, so is 120hz. timings need to be adjusted slightly.

2) 120Hz results in a black (but on) screen. This remains the same when using a 15cm cable (very short). I will try tonight to run this mode again with tighter blankings (was testing ~500MHz pixelclock). edit: 120hz needs slightly adjusted timings. my normal 2meter cable works fine.

In either 100Hz or 120Hz case, once windows loads, the entire screen is essentially rejected, being blocked from being enabled. When including a 1280*720 mode alongside the fast native modes, the fast native modes are ignored, but the system will happily use the lower mode.

cirthix
05-16-2012, 09:49 PM
SUCCESS. Tested with the scope tonight, some interesting notes:

1) Timings are picky (as usual when dealing with an effectively direct-to-tcon interface) with horizontal, but the vertical timings don't seem to matter as much.
2) Pre-windows drivers, the 7970 will use whatever the primary timing mode is, even if you tell it to give you 650MHz pixelclock).
3) The combination of my 7970, dvi cable, and ep269 (the DLDVI receiver used by the q270 monitor) were able to operate stably up until about 630MHz pixelclock (not a typo, not an error, verified with the scope watching the reliability of the sync signals (they are encoded in the same data stream as pixel data)). Of course the panel wasn't able to keep up, but that bodes very well for my weekend project: attaching a 2560*1600 panel.
4) Using near-default timings, I was able to get 120Hz using 7970, as verified in the pictures. This works in bios and ubuntu, but the amd drivers ignore the display in windows, not even listing it for use.

Here is the EDID which I made after tonight's testing:

EDID BYTES:
0x 00 01 02 03 04 05 06 07 08 09 0A 0B 0C 0D 0E 0F
------------------------------------------------
00 | 00 FF FF FF FF FF FF 00 0C 34 00 00 00 00 00 00
10 | 28 15 01 03 A5 3C 22 78 22 6F B1 A7 55 4C 9E 25
20 | 0C 50 54 00 00 00 01 01 01 01 01 01 01 01 01 01
30 | 01 01 01 01 01 01 83 BA 00 A0 A0 A0 18 50 30 20
40 | 35 00 55 50 21 00 00 1A 00 00 00 0F 00 0A 0A 0A
50 | 0A 0A 0A 0A 0A 0A 0A 0A 0A 0A 00 00 00 0F 00 0A
60 | 0A 0A 0A 0A 0A 0A 0A 0A 0A 0A 0A 0A 00 00 00 FC
70 | 00 51 32 37 30 20 31 32 30 48 7A 20 4F 43 00 E4


Some pictures in my picasa:
https://picasaweb.google.com/112908198380555779242/Overclocking25601440LcdTo120HzEp269Hd7970

End results: EP269 is an amazing DVI reciever, capable of operating well beyond the the minimum requirements.
HD7970 is perfectly capable of running this display at 120Hz, driver limitation needs to be fixed.

Arnotts
05-16-2012, 10:19 PM
This... is very interesting. Thanks for your work!

I'd love to be able to use a 7970 while getting 100hz+ with the Catleap. Imo the 7970 is better value once you start overclocking the gpu.

cirthix
05-16-2012, 10:22 PM
catalyst 12.5 betas do not remove the lock

HyperMatrix
05-16-2012, 10:43 PM
Wow. Very exciting news. If this were somehow to work in windows...

cirthix
05-16-2012, 10:44 PM
It works in safe mode. Better than nothing.

HyperMatrix
05-16-2012, 10:46 PM
So as long as the AMD drivers themselves aren't interfering, windows can pick up the data from the edid itself. But the AMD drivers somehow have an issue with it as it is still over the 330mhz (or 400mhz?) limit.

I wonder if the AMD driver would accept the edid info if it were set to around 100hz refresh rate with a sub-400mhz pixel clock.

cirthix
05-16-2012, 10:47 PM
tried that, didn't accept.

cirthix
05-16-2012, 10:48 PM
I'm uninstalling all amd driver stuff and seeing what happens with the default-loaded drivers. next step is to install the drivers without CCC to see if it is CCC that interferes

edit: after uninstalling all but the driver, the problem persists. uninstalling driver now too.

siberx
05-16-2012, 10:56 PM
630MHz pixelclock

630MHz pixelclock

630MHz

Yes. I'm pretty damn excited. Now if only we could get the powerstrip guys to properly support the card or (even better) get AMD to fix the driver restrictions imposed on the card... 120Hz on 7970 should be easy pie then.

cirthix
05-16-2012, 11:02 PM
Consider that the DP1.2 output shares the same pins (reconfigurable, point is that they have identical electrical characteristics) and that runs at 5.4Gbit, so seeing the card pass data at 6.3Gbit isn't all that surprising. I would expect the gtx680 cards to do 540MHz pixelclock at least.

BTW, I was able to get 120Hz 2560*1440 in windows non-safemode by uninstalling the driver, but that is sort of pointless. I've installed them again and am back on the zr30w.

Now, do I feel like trying LM300WQ2 with the tcon and ep269 board from the catleap tonight or do I wait for the weekend. HMMMmmmmm

BTW, I passed along the driver limit request to a reasonably high up person at AMD, we'll see what he says.

HyperMatrix
05-16-2012, 11:11 PM
Regarding the GTX 680 max I was able to get it to was a 537mhz pixel clock. Beyond that hit a big old solid red screen. But that's related to the horizontal refresh rate hitting 200khz. It doesn't like going beyond that point. Not sure if that's limited on the card side or monitor side.

cirthix
05-17-2012, 02:38 AM
Just took apart zr30w and catleap, tcon boards look incompatible, so I did not try swapping driving electronics. I'll trade 160 pixels vertical for doubling the refresh rate.

TheJesus
05-17-2012, 03:10 AM
Yo, 7970, I'm happy for your 630MHz and all, but Kepler does 680MHz

Lol, just kidding. Well, if this really is a driver issue, I'm gonna beat the shit out of the support I've been talking to.

HyperMatrix
05-17-2012, 03:47 AM
Yo, 7970, I'm happy for your 630MHz and all, but Kepler does 680MHz

Lol, just kidding. Well, if this really is a driver issue, I'm gonna beat the shit out of the support I've been talking to.

Lol. Yeah. Make sure you let them know that it works without the AMD drivers installed. So the card can handle it. The system can handle it. But the drivers ruin it.

Brian
05-17-2012, 04:06 AM
Would you expect this is unique to the high end 7-series? I have a 6850 and am wondering if these developments affect me (although I don't have a 2B right now so it's all speculative).

HyperMatrix
05-17-2012, 04:13 AM
Would you expect this is unique to the high end 7-series? I have a 6850 and am wondering if these developments affect me (although I don't have a 2B right now so it's all speculative).

Well the 6850 can already do between 101-105hz. Problem with the 7000 series was that it was limited to 82-85hz. Whether any of this results in AMD taking action and fixing their drivers...who knows. It could very well affect you as well. But regardless, at 100hz+ on your card as it is, you don't have a lot to worry about.

Brian
05-17-2012, 04:15 AM
Perhaps this would alleviate the need for staged refresh rate bumps? ie 60 - 85 - 100?

HyperMatrix
05-17-2012, 04:49 AM
Perhaps this would alleviate the need for staged refresh rate bumps? ie 60 - 85 - 100?

The bump from 85-100 is non existent, apparently. My guide is wrong in that regard. :( But yes if AMD fixes it, then we'll be in a much better spot.

Sn0_Man
05-17-2012, 08:11 AM
Mad props to Mr. Cirthix for this excellent research and info. Too bad I already caved and bought the 670 :P

Not that there is any guarantee of this ever being fixed.

cirthix
05-17-2012, 01:25 PM
Btw hd5970 does not display to the catleap with 120hz edid. I don't have hd6000 series to test with.

cirthix
05-17-2012, 01:30 PM
Can someone with gtx 680/670 flash my EDID and report back? Looking at you hyper :D.

Qiko
05-17-2012, 02:49 PM
What about modded amd drivers? worth a try maybe?
http://forums.guru3d.com/showthread.php?t=322616&page=153

HyperMatrix
05-17-2012, 10:16 PM
Can someone with gtx 680/670 flash my EDID and report back? Looking at you hyper :D.

Don't have any hardware to flash it. And can't get DDC to work in softmccs or edid writer. And can't detect i2c to aux translation support from the display driver.

cirthix
05-17-2012, 10:39 PM
use old powerstrip and an old video card. old drivers of course.

HyperMatrix
05-17-2012, 10:50 PM
I have a 4870 in my other PC. I'll try it out there.

Davion
05-18-2012, 12:08 AM
U go Cirthix!! I want to do 100-120Hz on my 7870 too =)

chuckin_beast
05-18-2012, 04:04 PM
Sooooo......my crossfire 7970s on water....won't be able to take advantage of this Catleap gloriousness?

cirthix
05-19-2012, 12:00 AM
In the past (ATI time-frame), the items you are interested in have been per interface options enumerated in the Windows registry. There were entries which described: the maximum permitted pixel clock, max display resolution, enable/disable the display with default modes if there were none that fit other restrictions, force adding of EDID described modes, restricted modes, use a fake (emulated) EDID contained in the registry, etc., etc.

It is likely that this type of information is still contained in the registry.

These were located in multiple places within the registry. At that time they could be found at:
HKEY_LOCAL_MACHINE\SYSTEM\CurrentControlSet\Servic es\ati2mtag\Device0
HKEY_LOCAL_MACHINE\SYSTEM\CurrentControlSet\Contro l\Class\{4D36E968-E325-11CE-BFC1-08002BE10318}\0000\Settings
NOTE: Device0 and 0000 could also be, at least, Device1 and 0001.

If they are not under the above keys, I would look for them in the registry by searching for the strings:
DALRULE
DALOPTION
DFPRULE
DFPOption
GCOOPTION

For some of the less used functions/options (e.g. things that would only be used in testing/development) I found them by searching for strings in the driver files. These were some options for which the entries were not put into the registry by the installer. It was possible to enable/disable them by adding new keys into the registry. Using keys which were not in the registry under a normal installation was not required for resolving the issues you describe below. In other words, try changing the keys which are already there prior to looking for other possible entries which do not already exist.





Someone please look into this. I poked around a bit but didn't see an obvious limit.

TheJesus
05-19-2012, 12:36 AM
I'm looking pretty intensely through here, but not seeing anything directly related. Did you try editing "3D_Refresh_Rate_Override_DEF" under HKEY_LOCAL_MACHINE\SYSTEM\CurrentControlSet\Contro l\Video\{Random numbers/letters, just look at them} and change it to 1?

Also, DAL_DFPEXTTMDSRestrictedModesBCD may be of interest. There used to be a registry value named "ExtTMDSMaxTMDSClockSpeed" which sounds like exactly what we need, but I can't find it :/

I also saw this article, no idea if anyone tried it: http://www.ehow.com/how_7649449_add-custom-resolution-ati.html

I'll keep looking but that's the most obvious one I've seen.

cirthix
05-19-2012, 09:27 PM
The key:
GCOOPTION_ExtTMDSMaxTMDSClockSpeed
Does not seem to help.

ToastyX
05-19-2012, 11:06 PM
I looked through the driver files and found a bunch of options that look like what we want, but none of them seem to do anything:

GCORULE_RestrictTMDSFreq
GCOOPTION_IntTMDSMaxTMDSClockSpeedIncoherentLow
GCOOPTION_ExtTMDSMaxTMDSClockSpeed
GCOOPTION_IntTMDSMaxTMDSClockSpeed
GXODFPEXTDualLinkMaxTMDSClockSpeed
GXODFP5DualLinkMaxTMDSClockSpeed
GXODFP4DualLinkMaxTMDSClockSpeed
GXODFP3DualLinkMaxTMDSClockSpeed
GXODFP2DualLinkMaxTMDSClockSpeed
GXODFPDualLinkMaxTMDSClockSpeed
GXODFPEXTMaxTMDSClockSpeed
GXODFP5MaxTMDSClockSpeed
GXODFP4MaxTMDSClockSpeed
GXODFP3MaxTMDSClockSpeed
GXODFP2MaxTMDSClockSpeed
GXODFPMaxTMDSClockSpeed
GCOOPTION_MaxTMDSFreq

cirthix
05-20-2012, 12:07 AM
I added all of those, didn't let me enable the display. The value that I'm using is 60000, as they seem to use it as tens of KHz.

I tried a fresh install of win7x64 and then specifying the driver to install through device manager, rather than the amd provided installer. Still locks the display out, so the problem isn't with CCC or any of the associated software, it is within the driver itself.

Qiko
05-21-2012, 09:37 AM
Is this helpful?

How to Use an INF to Override the Monitor EDI
http://msdn.microsoft.com/en-us/windows/hardware/gg487330

cirthix
05-21-2012, 09:21 PM
Is this helpful?

How to Use an INF to Override the Monitor EDI
http://msdn.microsoft.com/en-us/windows/hardware/gg487330

No point in an override, it is flashed to the monitor.

cirthix
05-23-2012, 09:03 AM
Tried with ubuntu12.04 x64 desktop and 12.4 catalysts. Same problem, monitor only works before drivers are loaded.

waperboy
05-23-2012, 12:11 PM
Just want to mention that nVidia's stable Linux drivers (295.40 and 295.49 on Ubuntu 12.04) refuse > 400MHz pclk across the board, it seems - tried with GTX 680, and I'm stuck at the same wall as with the 460.

Qiko
05-23-2012, 03:10 PM
" flashed to the monitor." So would the refresh rate be available to any other system? Like a ps3?

cirthix
05-23-2012, 10:26 PM
@waperboy: sucks, find the limit and see if you can modify it.
@quiko, yeah, but ps3 can't handle it so no point (you'd get a black screen)

Interestingly, with frglx installed, xrandr otuputs:

xrandr: Output DFP10 is not disconnected but has no modes.
get-edid only finds the active display

cirthix
05-27-2012, 03:02 AM
Interesting find.

I've already contacted Alex Deucher, please don't harass him.
http://lists.freedesktop.org/archives/dri-devel/2011-June/011909.html

I see the solution being one of the three:
1) change the max pixelclock as stored in the bios of the card (if it is there) to something higher, but reasonable (~720 would be good).
2) change the atombios layer to report a max pixelclock that is higher than what the card says it is capable of.
3) disable the check in the detection code to not care about the pixelclock.

IMO #3 is the easiest as #1 would have other complications, like invalidating the bios signatures.

spikestabber
05-27-2012, 07:10 AM
I don't know why its up to the CODE to artificially limit it. What are they going to do once we start seeing higher resolution displays for instance? Simply adding all this hackjob of hardcoded values that will eventually need fixed, bright idea.

TheJesus
05-28-2012, 01:23 AM
I don't know why its up to the CODE to artificially limit it. What are they going to do once we start seeing higher resolution displays for instance? Simply adding all this hackjob of hardcoded values that will eventually need fixed, bright idea.

Welcome to my current discussion with AMD. Nvidia is already preparing (kind of). Until the other displays that require higher pixel clocks start arriving in large scale, they have no need to increase it.

cirthix
05-28-2012, 07:46 PM
After more digging:

ADL_Display_PixelClockCaps_Get
ADL_Display_PixelClockAllowableRange_Set
in atiadlxx.dll C:\Windows\System32\atiadlxx.dll

getting closer!

TheJesus
05-29-2012, 03:10 AM
FWIW, here's atiadlxx.dll at ADL_Display_PixelClockAllowableRange_Set content: (open it in a new tab if in Chrome, it'll be fullsize)

http://i174.photobucket.com/albums/w118/winnah9000/asdf-4.jpg

siberx
05-29-2012, 02:25 PM
FWIW, here's atiadlxx.dll at ADL_Display_PixelClockAllowableRange_Set content: (open it in a new tab if in Chrome, it'll be fullsize)

Shouldn't it be possible to hex that dll and skip that subroutine (or force it to return an alternate higher value) then just run it as an unsigned driver? I've dealt with assembly, but not for x86; anybody well-versed enough to give input?

cirthix
05-31-2012, 12:16 AM
The limit is NOT applied to displayport displays. I tried an apple minidp->dldvi adapter and while artifacting due to the apple converter, the drivers did not reject the 340MHz pixelclock mode.

HyperMatrix
05-31-2012, 12:50 AM
Who was it earlier who said they had a DP adapter that went over 400mhz pixel clock, even. This would be very interesting to test. :)

cirthix
05-31-2012, 12:55 AM
Who was it earlier who said they had a DP adapter that went over 400mhz pixel clock, even. This would be very interesting to test. :)
That adapter will max out at ~425MHz, they are probably using the ST dp reciever which supports 3.24Gbps datarate.

Anegon
05-31-2012, 09:16 PM
The limit is NOT applied to displayport displays. I tried an apple minidp->dldvi adapter and while artifacting due to the apple converter, the drivers did not reject the 340MHz pixelclock mode.
So, if there was 2B with a DP port, it would be easy to set 120Hz with 7xxx?

TheJesus
06-01-2012, 04:26 AM
So, if there was 2B with a DP port, it would be easy to set 120Hz with 7xxx?

Theoretically. However, that requires different wiring and what-not. I think someone mentioned it was essentially a eDP or something and that the pin layout is similar so if someone hardwired a DP connector to it, it would work (in theory).

siberx
06-01-2012, 07:30 AM
Theoretically. However, that requires different wiring and what-not. I think someone mentioned it was essentially a eDP or something and that the pin layout is similar so if someone hardwired a DP connector to it, it would work (in theory).

That is not true for the 2Bs; the non-OC monitors use an embedded displayport connection internally, but the overclockable ones go from DVI straight to LVDS and then to the panel itself. The only way to use the Displayport to drive an overclockable monitor that would be known to work would be doing it through an active DP-DVI adapter that could handle a high pixel clock. Does anybody have more information on that DP-DVI adapter capable of higher pixel clocks?

HyperMatrix
06-01-2012, 07:36 AM
That is not true for the 2Bs; the non-OC monitors use an embedded displayport connection internally, but the overclockable ones go from DVI straight to LVDS and then to the panel itself. The only way to use the Displayport to drive an overclockable monitor that would be known to work would be doing it through an active DP-DVI adapter that could handle a high pixel clock. Does anybody have more information on that DP-DVI adapter capable of higher pixel clocks?

Someone posted they had one that tested at somewhere higher than 400mhz pixel clock. Which...would be phenomenal news for HD 7000 series card owners if it works.

cirthix
06-01-2012, 09:21 AM
No 4lane hbr2 receivers on the market yet.

cirthix
06-01-2012, 09:24 AM
Someone posted they had one that tested at somewhere higher than 400mhz pixel clock. Which...would be phenomenal news for HD 7000 series card owners if it works.

Tested an apple adapter at 340MHz pixelclock, autodetect was fine. the apple adapter itself wasnt stable at that speed though.

the 3.24Gbps dp mode is not fast enough for 120Hz.

cirthix
06-05-2012, 11:51 PM
Using ToastyX's driver modification, everything works fine at 120Hz, except for one thing:
The display becomes unstable when the gpu goes into it's low power state, so you have to keep the gpu active.

Vega
06-06-2012, 12:14 AM
Using ToastyX's driver modification, everything works fine at 120Hz, except for one thing:
The display becomes unstable when the gpu goes into it's low power state, so you have to keep the gpu active.

Could you be more specific? When the GPU voltage drops? When the frequency drops? Both? What do you mean by unstable?