Posted on by mwhybark

Cable

Came back from picking Viv up at the airport to find a neighbor’s delivery truck had popped the cable-company line to our house. Since we haven’t used cable services since we bought the house it was mostly a hassle in figuring out how to deal with the line in the road.

The original install had routed the house-side stay around a telephone line that appears to have been routed under our vintage vinyl siding, so I had a tiny cardiac moment in fear our DSL and land line were out, but they proved to be fine.

The real hassle started when I tried to get ahold of someone at Comcast (er, Xfinity) to report the downed line so that they would come and get it. There was no way to interface directly with Seattle-local CSRs, and the offshored CSRs could not file a report unless I had a customer number. I just hung up on the nice Ukranian lady who was telling me this in the middle of her family’s night.

I hopped in the car and drove up to the cable offices, not too far away, where I took a number and waited for forty minutes for a two-sentence interaction with the in-person CSR. She told me unambiguously that the downed line was being reported and that a technician would take care of it.

The interaction was so short I was not sure if she had even actually recorded my contact information, so after I got home I called Seattle City Light and discussed things with them too. I ended up filing a ticket with them as well, although the phone rep there said she thought SOP would be that the City truck would come by, the dudes in the truck would verify that it was not a City line, and that would be the end of it.

Instead, the City truck rolled up just after five and the dug in the truck was awesomely helpful and immediately took the line off just below the hanging lines on the pole. We joked about copper theft and in the end decided against putting a sign on the cable advertising it as free.

A few minutes later, a Comcast truck actually showed up! The cable guy was bemused but not unhappy that the City had already taken care of it.

Sadly I had to miss going to see Bill Vollmann read from his latest, “The Dying Grass.” Viv was not ready to eat when she normally is and felt her blood sugar was running high, so I just waited it out before serving. Unfortunately, that meant we did not finish eating until seven, when the event was slated to begin in downtown Seattle. There was no realistic way to expect being able to arrive there before seven-thirty and of course parking is another matter entirely.

The phone rang three times yesterday when I was unable to get to it and as it rang I knew it was Bill; it rang once more today at five or so and again I was unable to reach it. So I suppose I’ll drop a quick note. I am genuinely bummed about missing the reading.

Posted on by mwhybark

Noted

Continuing the theme of entertainment hardware failure, I experimentally fired up the Dumpsterama on Thursday and found to my sadness that the projector appears to have failed.

Naturally, I posted to AVSForum.

Today, trying to work through some configurations for the sake of completeness, a weird image which appears to be native to the projector partially displayed on the screen. The device seems hosed but shows signs of life. Given that crappy consumer projectors are at $100 on Amazon it seems foolish to spend much time worrying about something I rarely use. Undoutedly I will continue to do so.

Here are the AVSForum posts, and below them a link to a community with lots and lots of LP350 hacks. I suppose I should pick up on the Epson 76c too while I’m at it.

InFocus LP350 weirdness

Hi all – I know this projector is practically stone-age by now, but I love keeping old stuff running as long as possible. Anyway, I had this set up and operating for a few years sharing inputs between an S-video out pre-HD DVD player and an HDMI-out ATV2. It worked with both inputs just fine (up to a point – the ATV2 enforces HDCP on some media and the projector predates HDCP, which is another tale). Our use patterns changed and it’s been hanging from the ceiling unused for a couple of years. Today I got a wild hair and hooked it up to a PC running Win 7 with both onboard VGA/HDMI out and a GTX 470 with dual DVI.

Windows recognizes the projector when it is attached to the HDMI out. I did not get as far as testing the 470 DVI connection due to some oddities.

First, I did not note any visible startup screen. I didn’t think anything of this as the machine always powers on with the lamp running quite dimly and since I was shuffling cables around I wanted the room illuminated.

Second, I was only able to note the menu displaying for a brief instant before I had set the focus.

Third, after realizing there was clearly something going on with the projector, I was able to get it to display the HDMI out at the correct and expected resolution of 1024×768 – but again, just for a couple of seconds, and the display went blank.

The bulb illuminates. It does not appear to be reaching full brightness. The output on the projection screen is a bordered square of indeterminate color. If I look directly into the projection beam and sweep my eyes across it, I can see the rainbow shifts associated with the display technology of this era.

The bulb was replaced some time ago after the typical series of bulb failures associated with this model. The last time I had occasion to check the use hours on the bulb, it was along the lines of 20 at the most.

The keypad appears to be operating correctly, with the standby button placing the unit in standby and waking it, the menu button illuminating the menu navigation keys (and as glimpsed the menu proper).

I am stumped. I have attempted a bulb-timer reset. What the heck is going on here?

Second post in-thread:

Continuing to mess around with the projector. I hooked up an old VCR to the RCA S-in and started running a tape just to see if I could get anything. Yesterday late and at first power-on today I noticed a faint black bar appearing across the projection area, something like the frame gap that would occasionally occur in non-digital broadcast days if your reception was slightly out of sync. The bar appears in different locations each time the projector comes out of standby. After one of the standbys I noticed the the bulb flickering more than usual, accompanied by some faint clicks, and then it began to ramp up to normal brightness.

A bar began to appear at the top of the screen, but bright white with a black stippled pattern. After it stabilized, it appeared to be the top eighth or so of a bitmapped image which looked somewhat like it might have been originally taken from a 19th century anatomy text or equally possibly a 1980s punk rock poster. There was a strange form on the left of the image which resembled a large-mammal rib and in the center there was a form which seemed possibly to be the top of a drawing of a person’s head. However it was objectively impossible to identify the subjects of the image as there was simply not enough shown on screen.

The image was a 2-bit bitmap with characteristic gridlike dithering, very closely resembling the sort of thing one might have drawn in the mid 1980s with MacPaint or a similar 2-bit graphics program. The image was displayed on screen regardless of the inputs connected to the projector, so my operating assumption is that it is an image stored in the device’s ROM or equivalent and used in assembly testing and verification at component assembly.

It was also reminiscent of cracker credit screens and the like seen on warez or malicious virus incidents. I don’t think I have ever attempted to use the device’s USB interoperability features. I don’t really think it’s too likely that there’s a virus in the machine, the image just sort of resembled that kind of thing.

Since noting the image I have standby-cycled the projector several times and the image has not recurred. I did take a couple of pictures with my cell phone so I will try to make them accessible here in time.

DIY Audio search results for LP350.

Posted on by mwhybark

Ramp

Logan snapped a knee ligament, the vet says, and requires surgery. As I posted in the morning, this blows our planned California road trip away. Viv will go alone via air instead while I manage a hobbled dog.

I built a 12 or 13 foot dog ramp out of to-hand scrap today – so far he hasn’t used it, which is to be expected. After I had the basic design and wood manifest I felt that I needed to check it via more reliable generators of measured dimensions than my head and my pen so I redrew the plan in SketchUp. My wood manifest was fine, as it turns out.

Every time I use a CAD-family tool, I’m taken aback at the obtuse design. For example, one may enter exact dimensions on a newly-created object but I was unable to discover a way to copy an extant object and then enter new dimensions, so for example instead of creating an 18x6x2 cross brace and editing a copy to be 17x6x2 I had to create a new object with the desired dimensions. Some of the UI decisions presumably stem from long years of CAD feature heritage, but another feature I could not locate was align-to-center, a basic building block of 2D illustration and page layout over the exact same period of time that CAD has existed.

I’m not sure what it means, other than I was as frustrated by the experience as I am when using accounting software, which includes features that inhibit wholesale copying of entries, the consistent UI location of specific information elements such as dates, and other sandpaper for the mind.

Posted on by mwhybark

trace

Today, as I was carded for beer to my surprise, a drunk man with a strong resemblance to Vladimir Putin shared my amusement at the event. He had a faint accent which I could not place, but his resemblance misled me to assume Russian.

Posted on by mwhybark

When things break

In late June, the signin on my old Apple TV 2 for MLB stopped working. This was not entirely unexpected because I run it jailbroken to allow the use of non-Apple software on the device, and the jailbreak software service (Firecore’s ATV Flash (black)) I use requires that the Apple-provided operating system not be updated beyond an old version, 5.3. Therefore as application providers change aspects of their services, including Apple, bits and pieces of the software on the device will fail as well.

I previously had set up XBMC on the device with an add-on installed called MLBMC that allows XBMC to access MLB.TV streams, but similar issues are affecting both XBMC and MLBMC – XBMC is now called Kodi and has a different suite of supported devices, and the volunteer developer for the MLBMC add-on has left the XBMC development community for reasons unknown.

I sighed and grouchily set about researching my options, eventually concluding that the best way to re-establish login to the MLB app was by un-jailbreaking the device and letting Apple update it. The plan was then to use AirPlay re-streamers to get local content on to the Apple TV, which is Apple’s preferred way of letting consumers use non-Apple content on the ATV family of products. It’s exactly what the Amazon Instant Video app does, and the developers of Firecore even have an app that does exactly that, Infuse. It’s been some time since Apple’s ATV version of iOS has been jailbroken, and it seems likely that there won’t be future jailbreaks, so these restreamers are the best we can do with the devices.

After playing with that setup, I found that none of the re-streamers I had in hand did a good job transcoding 1080p local content to the ATV2’s max rez of 720p and decided to give jailbreaking another go, this time with the idea of a fresh install of Kodi and some hacking around in the underbrush of MLBMC in the hopes that a new maintainer or version had appeared. In the end, this approach succeeded, though with some caveats. MLBMC was no longer, and may still not be, an officially-approved add-on for Kodi and consequently has been removed from the default ‘repository’ that is the most easily accessed source of software add-ons for Kodi. Therefore I had to hand-install it in the ATV2 implementation of Kodi.

I took the time to document this in a couple of posts at Firecore’s generally-moribund forums and wanted to take the time to grab that and publish it here too.

First, I responded to a fellow-user looking to get MLB working again.

Chris, I see this too. I have gone ahead and rerun all the steps (restore to factory, re-jailbreak) and can add a couple of observations. First, after a full factory restore to Apple’s current iOS, MLB.tv did work, as far as logging in was concerned. I did not try it with a geospoofed IP, but my understanding is that with a geospoofing DNS proxy service you can get around blackouts in the app on a non-JB ATV.

My prior setup had an old version of XBMC installed with divingmule’s abandoned MLBMC XBMC add-on and the add-on was able to connect and stream MLB content even with the MLB.tv app no longer functioning. However, the livestreams were stuttering and failing badly enough that I decided to tear it all down and start over.

Once I was on Apple’s official release I tried using several AirPlay-based restreamers to send my local content to the ATV and found that lo-rez content (480SD and up to 720p) played well, but HD 1080p content did not. I therefore determined to try an install of Kodi.

I just finished re-jailbreaking with seas0npass. I can confirm that seas0npass successfully rolls the base system on the ATV2 to 5.3 and that reinstalling Firecore and in 5.3 MLB.tv is no longer functional. I do not yet know if I can re-add MLBMC to Kodi even though it is no longer maintained.

Then, a couple of days later, with everything running smoothly, I dropped by with an update.

I was able to reinstall MLBMC in Kodi, and it seems to work OK. It’s somewhat convoluted to do. divingmule has stopped maintaining the add-on and another volunteer, Youioup, has been working on it. Since the add-on is not in the official repository any longer, you must install it another way.

The MLBMC thread at kodi.tv:

http://forum.kodi.tv/showthread.php?tid=104391&page=67

Youioup’s github for the add-on:

https://github.com/Yuioup/kodi-mlbmc

When you download the release (the ‘download’ button on the right of the Github page), you’ll need to locate a zip file inside the zip file you download called ‘plugin.video.mlbmc.2.0.75.zip’. the numeric string may differ; it’s the release number. That’s the install-ready zip that Kodi needs to see to install the add-on.

I followed these instructions to upload the zip file to Kodi / XMBC’s ‘Home’ directory on the Seas0npass/Firecore JB 5.3 ATV2:

http://www.appletvhacks.net/2012/04/06/how-to-install-xbmc-add-ons-and-p…

The article was posted in 2012 and uses screenshots taken with a much older version of XBMC. The basic procedure: Copy the zip file to Kodi’s Home Folder as described using SFTP. Navigate to “Settings > Add-Ons > Install from zip file”. When the file source dialog displays open the ‘Home Folder’. I am not sure if it’s labeled ‘Home Folder’ or ‘Home’. Once you’ve selected it, you should see the filename of the zip file you uploaded earlier along with some other info and folders. Select it and Kodi will install the add-on.

You’ll need to configure it with your MLB.tv account info. My install went without a hitch and I have not had problems with playback or logging in or anything. Your experience may vary; posting to the MLBMC thread at kodi.tv for help and with questions is your best bet if you run into problems.

A note to consider: I found that I had to restore and jailbreak the ATV a couple of times. The first time the restore-to-current went fine but then performing the JB failed as the device became stuck in DFU mode. I was able to restore it again and the second time everything went smoothly. However, other users here have sometimes reported that getting stuck in DFU mode can make the device unrestorable. I also had to pay for the Firecore software again, which was aggravating, but does provide access to what must be considered the final release of the software, including an implementation of Infuse which provides excellent downconversion of 1080p content to the ATV2’s 720p.

Finally, in January, the development lead for the Apple TV 2 implementation of Kodi announced that the current release is the final release for the platform:

http://kodi.tv/farewell-to-future-appletv2-development/

As absurd as it seems to relegate these less-than-ten-year-old devices to history, that is the reality we are facing. Given that MLB.tv is the subject of active and well-executed technical development and that MLBMC is both not an official MLB licensee nor officially supported under Kodi, one must anticipate the eventual loss of MLB.tv playback functionality on these jailbroken devices at any time.

Hope this helps.

An addendum: on July 19th, a user noted that MLB has started working for them again without their having changed the setup. I look forward to trying that.

Posted on by mwhybark

burned

The intolerable heat drove us from the hearth tonight in search of sustenance. Our stated destination was a fantastic and extraordinarily inexpensive Ethiopian restaurant nearby called Lucy, after the famed fossils of Olduvai Gorge.

I pause here to celebrate my just-concluded assassination of a mosquito.

When we arrived we were saddened to see the place shuttered. We walked around the corner and had an unremarkable meal of tacos and a burrito before strolling home. We stopped by the nearby market and picked up a couple of paletas to eat while headed home.

We crossed the nearby deadly arterial and watched three small kids play in a dusty chainlink lot next to a shack labeled “Blessed Event Counseling Center,” which pissed me off. A few houses later Viv found a “for sale” flyer box in front of an awful-looking full-lot crime against architecture.

I wasn’t really paying attention, because the house was so ugly, when Viv told me I had to see what they wanted for the place. Obediently, I came over to look at the hand-reduced asking price.

$600,000.

Free of volition, I yelped “WHAT!”

From behind a nearby bush, an amused voice said, “That’s for interested negotiators only.”

Viv and I burst into laughter and had a brief conversation with the fellow reclining in his lovely, gardened and detailed front yard in which we both expressed hope that the absurd asking price would pan out.

We turned home and on entering the nearest precincts were struck dumb by the mighty colors spilling across the sky for the next hour. As we entered the house we let our neighbor Tony know that now was probably a good time to gaze slackjawed at the sky.

The walk from our house to the nearest commercial center is not that long, and on a good day it holds a mix of good and bad, with the good predominating. But the bad is so bad, and so clearly the effect of structural forces within our economy and society that it defies reason to expect that an individual citizen can resist and remedy them. Therefore, we, and our neighbors, prefer to drive.

Posted on by mwhybark

Video card follies

As I’ve noted a couple times here in the past month, I impulsively bought a contemporary-architecture video card, a GTX 960, with the intent of using it in my Windows PC. On seating the card I was frustrated to note that it failed to operate, and have been engaged in a technical odyssey since then trying to troubleshoot the issue or to determine the extent to which the card may or may not be compatible with the machine.

This entry is going to be a link-and-data-dump consolidating my online discussions regarding the unit.

The card proper is the GeForce GTX 960 SSC 2GB, purchased from Amazon for under $200. The intended destination machine was a stock Inspiron 570 with an upgraded 500w power supply – the stock PSU for the machine was 350w and I bumped it when I bought my last set of GPUs, GTX 470s.

Here is the long narrative doc I wrote to clarify what tests I’ve run a few days ago. By the time I wrote this I had begun to try he card on one of my Mac towers as well, the current best-fit implementation but one with its own extreme limitations that need resolution as well.

—
(composed July 10, 2015)

GTX 970 narrative notes

First attempted install into i570

Useful link:
http://en.community.dell.com/support-forums/desktop/f/3514/t/19490818

Thread appeared to validate card as OK based on power reqs, but card would not power on boot.

http://en.community.dell.com/support-forums/desktop/f/3515/t/19638117

response notes probably higher power requirements

750w Corsair CS750M suggested PSU replacement.

Have add’l supercompact drive-bay PSU in hand originally sourced with the idea of driving 2x GT 470 in Mac Pro, not needed for that purpose. Mac Pro install is non-standard. i570 install may be wiser choice, certainly an acceptable test case.

After hitting power wall, began to experiment with GTX 960 in Mac Pro. In the meantime, correspondence with Nvidia established that the card was unsupported on the Mac hardware and that flashing it would void the warranty.

Kernel panic at boot with GTX 960 seated under Mavericks 10.9.x. Card is known as ‘Maxwell’, Maxwell wholly unsupported under pre-Yosemite OS X. Vanilla Yosemite install produces non-native video resolution to 3x monitors via DisplayPort to DVI cables, no reason to expect more out would also be supported.

Please note this is with stock GT 9500 / 120 seated. 120 is a fully Apple supported Nvidia card and as such displays what is known as the ‘boot screen,’ the apple and progress bar at startup or the boot drive selection at option-key startup.

No readily apparent downside to a Yosemite migration on the Mac Pro 4,1; even iPhoto is said to remain operable despite the hype of Apple migrating iThing services to new software packages.

Initial installation of Win 7 to discrete internal boot drive went smoothly. Both cards were recognized in Device Manager. GTX 960 appears with error triangle and is disabled. Error reports as Code 43. Diagnostic is step-through of disable, reenable, roll back driver, uninstall, reinstall. No resolution.

After several rounds of this the GeForce installer reported a problem with “MSI”, the remedy for which was to either rebuild MSI through a series of registry-type edits or via a clean install of Windows. I opted to start over.

Once back and up, the same issues generally presented themselves.

Using downloadable GeForce installers, installer numbered 3.41 is installed. On reboot, whichever card was disabled at GeForce install reports as enabled but video is available ONLY via GT 120. Additionally, Device Manager consistently installs more recent drivers for the GTX 960 than the GeForce installer.

(Update, a few days later the GeForce installers are fetching and installing the same drivers as WU, .5330)

In a Hail Mary, I rebooted without the GT 120 seated. The 960 came right up.

So to summarize:

  • To boot in Mavericks 10.9.x, the 960 must be unseated.

  • To boot in Yosemite 10.10.x, both the 960 and the 120 may be seated, and best results are obtained via Nvidia’s web drivers.

  • To boot in Win 7, both cards may be seated, but only the 120 will output video even when it is shown as disabled in device manager.

  • To produce output from the 960 under Win 7, the 120 must be unseated during a power-down cycle.

  • To observe the boot screen in Yosemite, the 120 must be seated.

This is clearly untenable; the two cards must be harmonized under Win 7 in order for this configuration to be acceptable.

SIDE NOTE: Bootcamp cannot maximize PCIe data transfer rates (it is limited to PCIe 1.x) and therefore the i570 is likely to produce higher throughput to the card. It would still be worthwhile to benchmark and experimentally run RoF on the MacPro prior to moving to the power upgrade procedure on the i570. if the i570 power upgrade fails, retrench with the expectation of selling the 960.

A random benchmarking site rates both the Mac Intel Xeon and the AMD stock chip in the i570 as roughly equivalent.

I need to post in on a Mac site regarding the conflict between the GT 120/9500 and the GTX 960 in Windows. I really think there should be a way to get the two cards to run concurrently.

Framerates in RoF with a direct-feed triple-screen setup at 3x 1080p (in either orientation) run mostly in the 50s and 60s with very occasional slowdowns under certain high-information-density conditions (flying low over an in-game town through mist with gfx settings on max). Overall a significant performance improvement is immediately apparent.

—

Shortly after I wrote this, I initiated a post on Dell’s community forums regarding the attempt to bring the card up on the i570. The same forums host a very detailed, insanely ambitious, highly-informative, but somewhat dated LOOOONG post on upgrading and modding the i570. The thread was written quite some time before the 900 series cards came out and does not address the issues I have been experiencing.

In my thread, a user suggests another power upgrade, to a 750w system that offers 20w on the PCIe power feeds (‘rails’). This sounds plausible but something about the suggestion led me to more carefully look at the power specs of the 960. As it turns out, the 960 has a lower power budget than the 470 it’s intended to replace, so while that 750w power supply *might* do the trick, it seems to me more likely that the problem is elsewhere. I’m reposting my contributions to the thread here.

—
(posted June 23, 2015)

Hi, after scrubbing forums here and elsewhere (such as Tom’s Hardware) I have not found a specific user testimonial to success running a GTX 960 in a power-bumped i570.

Is this card supported in the i570? Failing official Dell support, should it work anyway?

The master upgrade thread was posted prior to the 960 and is light on GPU details.

In general, user advice seems to indicate that given sufficient power, which 500w appears to be, BIOS update, and successful gfx driver uninstall, it should work.

What actually happens is: system powers on, all fans spin up, GPU fans spin up, no POST code beeps are heard, boot procedure halts at ‘checkpoint NN’ (A2, I think). I found a list of the Dell checkpoint codes and it indicated something about failure to start, I assume indicating a failure to initialize the GPU.

I use DDU to strip the gfx drivers. BIOS is current (A06, iirc).

I can boot the system without the card seated, with an additional antique VGA card in the lowest slot on the mobo and the 960 seated, and with the current GPU (gtx 470) seated. when the system boots with the VGA card and the 960 seated the 960 fans do not spin. I have not taken the time to examine the 960 status in device manager in this configuration.

I do not have another wintel machine to test the card in.

Signs seem to point to a bad card. GeForce appears to be willing to authorize an RMA. However, if the card won’t work in the machine, it may be wiser to send it back to the merchant rather than the mfr.

—

SpeedStep’s concrete analysis reads in part,

“You need an EPS12v 2.92 power supply. 500W is not sufficient.

You need 170W on the 3.3v/5v rails and 15 to 20W on the +5VSB rails and Minimum 375W on the 12v Rails.”

He also provides a detailed chart of the EPS12v 2.92 spec and a specific product recommendation, extremely helpful.

After I ran my power experiment with a discrete 450w unit, I posted an update to the thread.

—

Further progress or lack thereof.

First of all, thanks to SpeedStep for pointing at power as an issue. I’m not sure that’s the solution, however, after doing some investigation.

First, although the Mac does have a 900w PSU, and the actual max draw for cards has not been well-documented, the standard recommendation for dual hi-power cards is the implementation of a second PSU in parallel with the Mac PSU. I can’t recall the estimated top draw off hand but I think it was something like 300w, I could definitely be wrong. Given that the 960 comes up, and is described was a low-power card in comparison to others in the 900 series I wondered what its actual draw and PSU reqs were.

This PC World review notes that “The GTX 960 has a scant 120 watt TDP”. The GeForce 960 specs state that the minimum PSU is 400w.

The previous card I had seated on the i570 was a GTX 470. The linked specs suggest a min PSU of 550w, so it appears I was just lucky getting it to roll with the 500w unit. The wikipedia roundup on the 400-series cards indicate that the 470’s TDP is 215w.

(Update: I was misstating the specs, as will be discussed below. Nvidia’s specs do not call for a minimum PSU, they call for minimum available system power, which can be roughly estimated as PSU total power availability minus allocated power consumption prior to the card’s desired power draw. So that 400w min above more likely calls for a 550w PSU or more despite the card’s stated maximum power usage of 120w.)

So to summarize the data, the newer card appears to have a specification that calls for it to have considerably less need for juice than the predecessor card.

The 500w PSU currently installed does also have some deficiencies with regard to the specific available power that SpeedStep recommended, however. In particular, the PCIe jacks are rated up to 18w, not 20w. For the sake of experimental thoroughness, I poked around in my stack of parts and came up with a 450w PSU that had been acquired with the idea of implementing the dual-PSU mod to the Mac to enable dual 470s. This particular PSU was designed for use in the ride-along role and offers dual PCIe power feeds at 20w each. Given that the card would be the only draw on the additional PSU, I felt it should help to eliminate or confirm power as an issue.

As before, with the 960 seated and the on-board video disabled prior to seating and reboot, the system does not complete a boot sequence, and on next restart provides the “System halted at boot checkpoint [A2].” Slotting an antique VGA card into the PCI slot does allow the machine to boot with the 960 slotted, and consulting device manager results in the 960 being reported as active, installed, and working properly. Sadly, there is no video being produced by the card in this state.

There are some quirks in this i570’s startup sequence, the most notable being that it seems the USB ports are not being activated until after the initial startup screen, offering F2 and F12 setup options, has zipped past. This of course makes it impossible to get into these startup menus to examine them for video-device options.

In short, while it’s still possible replacing the 500w PSU with a beefier unit might resolve the matter, it does not appear that the card in and of itself requires the additional power by spec, as the unit in this power configuration exceeds the spec’s maximum power requirement and can be shown to do so with a higher TDP card.

I still have not found a positive anecdotal report that the 960 can work with the i570 under the current (and probably final) BIOS update, which was released in 2012. There are many instances of people asking this question online and being advised that with sufficient power the card should work; none of these interactions have produced a troubleshooting thread like this one, so I am inclined to think it should be possible.

Thanks again to SpeedStep for prompting me to take a closer look at the power issue.

—

So currently my suspicion is that the motherboard is experiencing some sort of issue which is preventing the keyboard from being recognized at boot. Earlier in this project I was very definitely able to get into both the F2 and F12 startup screens. The USB sketchiness is very odd – it seems to affect certain ports more than others, which could I suppose indicate intermittent physical connectivity. However visual inspection of the ports does not seem to support a bad connection from the port I/O unit to the mono. The port cluster is modular and does rely on a slotting system for mono contact, however.

One aspect of the i570 that I had not noticed previously and which normally is NOT a bad thing is the apparent lack of a built-in speaker. In this instance, it means that there are no boot-cycle status beeps, which is impeding my troubleshooting.

Just after I posted this SpeedStep replied with more information elucidating the writer’s perspective on power supplies and power specs. The upshot is that the manufacturer’s suggested minimum system power spec line refers to available power to the card after all other draws are accounted, not to the total available power provided by the PSU. So my 500w PSU may have as little as 354w available for the card after accounting for all other draws. This does not invalidate the power test conducted with the ride-along PSU, however, which certainly did have 450w available to the card.

As I was reading this and thinking about it I realized I had another thread to pull in here, concerning getting the 960 up on the Mac. MacRumors forums has a super-long and active thread about working with Nvdia non-EFI cards in Mac Pros, and I posted this in thread:

—

Hi, thought it might be beneficial to share experiences slotting a GTX 960 in a Mac Pro 4,1 late ’09.

First, as well-documented here, I had to install Yosemite to get any video out on the 960. Vanilla Yosemite provided video out even prior to installing the Nvidia web drivers, but not at my monitors’ optimum resolution. I don’t recall what the available resolutions were but they were all higher than the 3x 1080p I wanted; the video was scaled and usable if ugly. With the web drivers installed, 1080p was available. I have not tried installing the web drivers on the extant Mavericks boot volume and do not expect that to succeed but intend to try it for the sake of completeness.

Using Bootcamp to install a Win 7 boot volume has also succeeded, although the process was hairy and took several days of troubleshooting to master. I went through two complete clean installs of Windows before I could get the Nvidia installer to successfully implement the current drivers for the 960. I was able to get the correct drivers (.5330) for the 960 both by using Windows Update and by using the Nvidia installer in the end.

However, there are some major issues that remain unresolved.

First, the stock GT 120 (or GT 9500, the OEM EFI card), *must* be removed prior to booting into Windows or Device Manager will become confused and direct video out ONLY the GT 120. Disabling and/or uninstalling either card in Device Manager does not affect this; Device Manager can report either card as active and the other as disabled and the only video out is the 120. On reboot, one of the two cards will be reported as functioning normally while the other is yellow-triangle halted with, I believe, Code 43. It does not matter which card is shown flagged, there will only be video out from the 120. The only way I have noted to get video out of the 960 is to unseat the 120, which makes booting back and forth between Mac and Windows less than convenient.

Second, the Bootcamp / Win 7 audio issue which is customarily resolved by reinstalling the Realtek drivers under Windows does not successfully restore normal audio operability. I have been able to implement an audio solution by plugging in an external USB audio in/out module. This is actually an acceptable long-term solution for me even though it’s inelegant.

Hope this helps others running similar configuration experiments.

—

In-thread, another user suggested that preventing install of any Nvdia drivers for the GT 120 / 9500 would resolve the disabled 960 issue. I haven’t run a controlled experiment to verify that but am skeptical as I very definitely have seen the 120 come up as a generic VGA and with the 960 reported as active in Device Manager but in actuality with no video out. I should still see if I can get it to work as just keeping the 120 around in a static bag in case of emergency or booting back into Mavericks seems problematic at best. Reading the thread it does seem to be the most common response.

Posted on by mwhybark

featherball

Today while walking the dog I happened through a majestic three-way furball, a large wing of violet-green swallows and their allies the dragonflies against what must have been a considerable swarm of dastardly and delicious gnats.

I would guess there were thirty or so birds. The combatants were using rising air up from a large work pit bordering the gravel road I take to get to the dog’s accustomed playfield. The net effect was to place the zenith of the swallows’ upward trajectories just above my eye level about six feet out beyond the edge of the embankment I stood on, looking down into the action. It was as fantastic a session of aerobatics as one can ever hope to witness. Occasionally a patrol of two or three birds would wing out and around me until they saw my placid and uninterested large dog, where up they’d wheel back into the thick of it.

I did not notice any dropped frames. There was some sort of processor-intensive time-compression effect whereby the moment both lasted forever and was gone in an instant.