X61 - Proposed hacks under $500: Dual HDD/SSD both bootable

[flyingfishfinger]

E350: What politics? I may have missed a joke someplace... oops.
I won't have any updates until next week at the earliest, since we're in the process of moving offices and our workbench is currently packed up until further notice. I may also learn the hard way how often one can connect and disconnect the internal keyboard cable from the motherboard, since it's required to turn on the thing, but I don't store it set up at the moment.
I have no input on the wireless AC since I don't know anything about it, and my HDMI experiments are currently stalled pending the successful use of an internal USB port, with which I haven't had much luck. I could use a hand there: I tapped into the Ultrabay USB port by stealing power from elsewhere, but I can't get anything to enumerate. Devices detect just fine, then fail. Any ideas?
I also stripped one of the mPCIe slot screws, so my Wifi card is stuck at the moment. ...

R

[axur-delmeria]

UPDATE: Tried out the mPCIe SATA controller again that I initially played with (linked in first post) and it seems to work this time around

Which SATA card did you use? The StarTech or the PM362?

[flyingfishfinger]

I added a power switch to the PM362. Not for compatibility issues, but rather for temperature problems. It functions fine, but the chip becomes almost too hot to touch. This will let me choose when I engage the 2nd hard drive.

The modification:
- Cut all the 3.3V traces (there are 5, turns out two of the "reserved" pins are 3.3V as well)
- Solder two wires: One to the connector side of any of the traces you just cut, and one to the board side of any of the traces you just cut. This is a 4-layer PCB, one of the internal layers is power.
- Add a switch and find an appropriate place to stash it. I removed the IEE1394 connector (who uses that anyway?) and put it there.

The power pins I chose are on the bottom side, since the ones on top are very close to a decoupling capacitor, which means I fed the wires to the switch through the mechanical key. It still fits, but a little more tightly than before. Also, to remove the giant SATA plugs, you need to drill out the mounting posts...at least I wasn't able to unsolder them.

https://drive.google.com/uc?id=0B3TxYx- ... ExweVp1ekU
https://drive.google.com/uc?id=0B3TxYx- ... m9oTVJMdkk

Hot glue everything in place and you're good to go. Next up, modify the hard drive interface card to split off data pins for this...
Cheers,

R

[E350]

Dang, your images are not showing for me.

Also, would you mind showing or listing the soldering (iron?) which you are using? I have never soldered anything as delicate as a motherboard before and when I was thinking about soldering pins on a X61s to install the second fan, I decided neither my skills nor my soldering gun would be up to the task.

(Finally, Bay Area/Central Valley political rivalry - no missed joke, just a little ribbing. The Central Valley of California is a lot more like Texas than the California Coast just like Austin is a lot more like the California Coast than the rest of Texas.)

[flyingfishfinger]

Ah, fixed permissions. People should be able to see them now.
There isn't any soldering on the motherboard, this is adding the switch to the CARD, not the socket. Thankfully, the traces on the card, especially the power traces are much wider than signal traces anywhere else. I used an Aoyue Int 906 rework station, but I don't think it matters here. You'll see that the wires are pretty thick (26AWG), that's because I wanted to at least try to maintain the ability to draw lots of current. If you read the mini PCIe spec, devices can draw a peak of 2.7A (!) out of the 3.3V supply. I'm probably already pushing it by using just one supply pin through the switch, whereas the card originally had 5...

R

UPDATE: Whatever I did to the card, it looks like I broke it. Maybe any of the 5V lines is some sort of status indicator. Either way, it no longer shows up at all, whether the switch is on or off. I'll do some experiments.

[vavet]

Any news from the mods. How is the overclocked doing? is it stable?

[flyingfishfinger]

I haven't load-tested it yet. It runs fine at 2.9Ghz without overvolting as far as I can tell. The machine is still in pieces, the last few weeks at work have been brutal. I just received the PCHUB cable in the mail yesterday, so I'll be doing the screen mod this week. today.

Result: There seems to be something off with the EDID detection, besides the broken LCD cable. If I hold the plug just right, I can get it to work without color bars and static, however the resolution is off. I can't adjust the resolution past the usual 1024x768, so I figure the panel specs aren't being detected properly. I feel like I've someone else have this problem before, but I haven't been able to find it recently.... Found it. Yup, I need a new cable assembly. This is what's going to gobble up the rest of the $500, replacing things I break in the process...

R

https://drive.google.com/uc?id=0B3TxYx- ... WRScnBudUE

[flyingfishfinger]

New LCD cable did the trick. On to putting the SSD / HDMI cards next to each other in the PCMCIA slot, it's tight in there.

R

[E350]

Finally...

[flyingfishfinger]

Sorry! If this was my full-time job I'd have it done in a week. I've only been able to put in maybe half an hour every few days lately. When I'm done with work I usually don't have motivation left to be in the office any longer, even to use the workbench...

[E350]

Ditto that. Just kidding you. I pretty much collapse when I get home from work. You deserve compliments for what you are doing and how you are documenting it. And you are certainly way more ambitious and able than I.

[flyingfishfinger]

Some updates:

- HDMI adapter works on my hacked-in USB port from the PCMCIA socket (1080p is a little laggy, but that's expected on USB2).
- Wired SSD in the way I want it, it physically fits next to the HDMI adapter where PCMCIA used to be
- Fixed SXGA issue
- Running stably at 2.9Ghz AFAIK.

I'm debating whether I want to break out the overvolting cable from under the heatsink before I screw it back together just to have the option of getting to 3.4Ghz. That's my hesitation on putting the body back together. I'll probably mount the SSD and HDMI where they're supposed to go over the next few days.

R

Teaser pics:

https://drive.google.com/uc?id=0B3TxYx- ... 3NmTWJldTA
HDMI + hacked USB

https://drive.google.com/uc?id=0B3TxYx- ... nhGZktvb28
SSD harness

https://drive.google.com/uc?id=0B3TxYx- ... lBYc3hFRTg
HDD adapter modification

https://drive.google.com/uc?id=0B3TxYx- ... FZvaUFNbUU
FSB selection switch

MOD edit: pictures too big, changed to links.

[xiphmont]

EDIT: I started a new thread here to play more with overclocking. You can mostly disregard this original post, you can in fact do some fun overclocking with these boards!

Original post:

I am trying to find a guide for overclocking an x61s (L7700). Can you suggest one?

BTW, I actually got around to playing with overclocking the L7500 and L7700 on some 'spare' (not really spare, just available) motherboards using the pin-7 FSB mod described on the German forum. This was all using memory tested to operate well at PC2-1066.

None of the four motherboards I modded was stable overclocked.

With only the pin-7 mod (FSB 200->266, stock voltage) the two L7500 mobos appeared stable; they did not crash at any point. But linpack's result checking indicated they were returning sporadic math errors. Even at this point, the processor would throttle hard after about a minute of load. Adding more voltage would stabilize the math results, but also caused the processor to throttle harder. In general, the L7500 only got to 2.0GHz without a hardware voltage boost mod, and even then it was drawing more voltage/power than a stock L7700 just using dual-IDA. Adding more voltage caused it to throttle sooner such that the overall MFLOPS actually dropped overall, even if it was faster for a burst at the beginning.

Neither L7700 mobo would even boot overclocked without a hardware voltage boost or locking to a low multiplier in BIOS. Even at 10x and 1.1v, both would throw sporadic machine check exceptions under heavy load. At that point, I was already well past what even the spiffier heatsink could handle and did not push further. More voltage was only going to hurt performance even if it made things 'stable'.

So, sadly, that appears to confirm there's really no point to hardware OCing the X61s or X61t. It could perhaps be useful to get the faster RAM/FSB throughput, but only if limiting the CPU to lower multipliers that don't much exceed the original operating range.

[wileE]

That mod with the second drive in the X61 looks very very interesting. Looking forward for your experiences with that.

@xiphmont I would say you were particularly unlucky with your motherboards.
The L7700 is known to frequently cause problems, but the L7500 usually works.

The linpack calculation errors you got were worrying enough to try that benchmark on my X61 with a T7300. It runs at 2.66GHz, no voltage adjusments, and I dont use throttlestop. Under light loads the cores do go to 2.92GHz.

Tried several settings for nearly 1 1/2 hours and never had an error. Temps were around 80° average, peaks at 90°. But it alternately throttles one of the cores to a multiplier of 4x.
Advantage of the overclock on that machine is that, together with an SSD, it is very responsive and snappy. And its not my first choice for serious numbercrunching.

[xiphmont]

@xiphmont I would say you were particularly unlucky with your motherboards.
The L7700 is known to frequently cause problems, but the L7500 usually works.

Very possible. I didn't try any T-series on a full X61 motherboard, only a few low-voltage processors. And I'm also sure they'd have been stable with more voltage, no problem-- but given the thermal limitation of the smaller slim/tablet heatsink, it wouldn't have gained ground. I've been cursing those tiny heatsinks for years, no reason to stop now

That said, I have been thinking more about running a custom multiplier setup with a Linux PCH kernel as the _memory_ was perfectly stable at 266FSB over four hours of memtest86, and that should be a big boost to the integrated graphics.

Oh, and thanks for rechecking your own machines with the linpack tests. I did wonder how many of the glowing overclock reports on the german forum were thoroughly tested. Knowing yours passes the tests that mine failed keeps some of the interest alive (though I currently have no personal interest in non-tablet machines )

EDIT: I dug out one more machine that was busted (yes, I have a small pile) and extracted the L7500 mobo. Figured it was worth one last try. This one, with a straight pin-7 mod, is not only stable with no voltage boost, it passes a linpack test significantly undervolted! Thanks to WileE for convincing me to try again! EDIT2: Not stable with dual IDA however-- insta crash.

(the sad part is, of course, it still throttles )-:

BTW, I noticed something very odd. On some of my mobos, I did the pin7 mod and speedstep still worked as normal. On others, the BIOS locks the mobo to lowest-multiplier-only (though I was able to set higher multipliers with some direct register manipulation in Linux). Is this a known phenomenon? Is it a specific BIOS version?

[flyingfishfinger]

Sadly, I haven't had any luck getting the SATA -> mPCIe reawakened. As I posted earlier, I tried adding a power switch to solve the heat problem, but that seems to have broken it.

I'm going to try another thing that should also help with the HDMI port, which is USB3. I ordered one of the SYBA USB3 cards as well as as a USB3 -> SATA adapter. The card presents 2 USB3 ports, so I should be able to run the HDMI adapter off one of them and the SATA adapter off the other one. It'll live in the space where the dial-up modem used to go.
I'm fully aware that the mPCIe ports on the X61 won't get me full USB3 speeds, but plain math says I should get about a 4x improvement over USB2, which should hopefully help the 1080p lagging issue and run the 2nd HDD on the side. It arrived today, I should be able to report back with speed tests on the HDMI tomorrow.

R

[E350]

I have nothing to add. Only posting to encourage you to continue your explorations! USB3? Everybody has wondered about it. Borrowing from the Aussies: GoodOnYa' M8!

[flyingfishfinger]

Quick update: Installed the Syba USB3 card, which works (as in, drivers install and 2 new USB ports present). I can even run the HDMI adapter off of them.

Issue: I don't think I'm getting any improvement in speeds at all. I'll bring in a USB3 HDD tomorrow and see if that works any better, but the HDMI video is just as laggy as before and USBView crashes when I try to select the new ports

R

[xiphmont]

Oh WileE, you have opened an interesting little rats-nest of observations here... I'll have to break this out into its own thread soon. I have noticed a few things playing with the overclocking that have surprised me (and may be of interest to your OCed machine too)

1: All the OCed numeric faults are happening on power state transitions. Eg, over half an hour of running linpack on an pin-7ed L7500, it may return an errant result only once or twice, and when it happens, it appears to always happen when the machine is switching power states. I think it is happening when unthrottling, ie, going from 0x880b (throttled SLFM) -> 0x081c (full speed/full voltage), but it might be happening at the point of throttling (transitioning the other way); I don't have enough resolution to see that with my current test. Something tells me that if I force constant P-state switching, I would see a lot more failures. WilE: in your testing for stability, does the machine ever throttle or transition?

2: Different stock motherboard revs appear to have significantly different idle power consumption. Eg, a naked, idle 42W7178 (48.4B401.011) with Middleton and 8GB booted to console linux draws 6.75W doing nothing. An identically set-up 43Y9032 (a later 48.4B401.021) draws 5.5W. (Interestingly, the 42W7178 is stable when pin-7 OCed, the 43Y9032 is not). The sample space on this is still small; I've tested three 7178 boards (behavior was consistent), but only one 9032 board. I've also not checked that voltage reg is in spec on the 9032, nor compared loaded draw. Another interesting tidbit-- the 7178 board advertises different, lower VID multipliers than the 9032 board despite the same processor rev (SLAET) and BIOS.

3: More RAM *does* consume measurably more power. Each additional 1GB of RAM adds approximately 1/8W to idle power consumption. This holds well across mobo revs, different BIOSes, and different manufacturers' sticks. It does not matter if the memory is split across two sticks or all on one.

4: Just having ethernet plugged in consumes 1.5-2W. Ouch.

5: Different BIOSes cause significant changes in idle power consumption. It seems that the Middleton BIOS causes a straight up 1.5W additional draw on earlier mobo revisions. I have not tested this carefully on later part numbers, but Middleton appears not to dramatically increase draw on the 43Y9032 board. This might be the answer as to why SATA-II was disabled on these machines-- that's a huge hit to battery life even if it only affects some mobos. Or... the additional draw may be due to something completely else, or even a change in an earlier BIOS rev that Middleton simply incorporated-- I obviously need to bisection search all available BIOS revisions

[EDIT: After much more careful testing on a set of motherboards, four different BIOS revisions, and better calibrated measurement hardware, it is *NOT* the BIOS causing power consumption differences. Middleton also does not increase power consumption measurably. Testing reveals that different motherboards simply have different inherent idle power draw. Some draw as low as 4.5W, some as high as 6.5W. I've found no pattern by model number, planar revision, processor stepping, or even processor microcode. Some mobos are just 'better' than others. Median is 5.7W across 16 tested motherboards.]

Oh, and the locked multiplier question I had earlier is answered.... somehow a 65W adapter snuck into my cache and I was testing both with and without battery 65W + no battery == limited to 6x.

[wileE]

Quick update: Installed the Syba USB3 card, which works (as in, drivers install and 2 new USB ports present). I can even run the HDMI adapter off of them.

Issue: I don't think I'm getting any improvement in speeds at all. I'll bring in a USB3 HDD tomorrow and see if that works any better, but the HDMI video is just as laggy as before and USBView crashes when I try to select the new ports

R

USB 2 is not your limiting factor when you play HD video with an adapter that most certainly uses a DisplayLink chip. At least not as long as you do not try for some crazy 4k vid.
You can try a different software player. Sometimes helps. But basically your CPU cant handle it. And the OC only helps a little. My T7300 2.66 GHz can play most 1080p 25fps videos on a 1680x1050 screen, using the Lenovo USB to DVI adapter.
But that is nearly the limit of what it can do. A 50fps video will not play without lagging.

@xiphmont
Yes, you should start another thread in the X6x forum.
With linpack my CPU throttles one core to multiplier x4. It switches cores every few seconds. With wPrime 1024M it never throttles.

[flyingfishfinger]

It seems like you're right. I DO get a speed improvement with storage, but not with the HDMI.

My HDD went from 16MB/sec on a USB2 to around 50-60MB/sec on the USB3! At least now I have USB 3 ports.

Updating the first post with some results.

R

[flyingfishfinger]

Well, looks like I broke something (again).

Stress-testing the overclocking with slightly over-volted CPU (20KOhm) at 3.2Ghz. At 96C on Prime95 the machine shut down.
Attempts at reawakening it haven't succeeded, plus it seems like the DC power connector is arcing every time I plug it in now (between the inner shell and the contact that receives it).
I've been checking fuses, but the fuselist doesn't correspond to my motherboard very well and I can't find about 1/3 of them...

R

[wileE]

http://thinkwiki.de/X61_Sicherungen will help. Hopefully.

[flyingfishfinger]

Ah, that's the right one. All fuses seem to check out, however...

[Pete B]

Quick update: Installed the Syba USB3 card, which works (as in, drivers install and 2 new USB ports present). I can even run the HDMI adapter off of them.

Issue: I don't think I'm getting any improvement in speeds at all. I'll bring in a USB3 HDD tomorrow and see if that works any better, but the HDMI video is just as laggy as before and USBView crashes when I try to select the new ports

R

USB 2 is not your limiting factor when you play HD video with an adapter that most certainly uses a DisplayLink chip. At least not as long as you do not try for some crazy 4k vid.
You can try a different software player. Sometimes helps. But basically your CPU cant handle it. And the OC only helps a little. My T7300 2.66 GHz can play most 1080p 25fps videos on a 1680x1050 screen, using the Lenovo USB to DVI adapter.
But that is nearly the limit of what it can do. A 50fps video will not play without lagging.

@xiphmont
Yes, you should start another thread in the X6x forum.
With linpack my CPU throttles one core to multiplier x4. It switches cores every few seconds. With wPrime 1024M it never throttles.

My experience is about the same in terms of a T7300 and 1080p 25fps. Is it me or has IE11 gotten
better at decoding MPEG in the last 6 months or so? I often play videos as a mild stress test and
it seems that CPU utilization has gone down in IE and it seems to do significantly better than
Chrome at MPEG decoding.

[flyingfishfinger]

Got a new mobo. Will resume assembly this week, probably going to skip the volt mod this time...

EDIT: I decided to do one thing at a time and shelve the SXGA mod for a while as well, at least until I get the rest running well enough to swap lids (I did a cable mod and the SXGA panel works, I'm unsure about whether I actually need to trim the metal lip which I'm a little afraid of doing). Also investigating replacing the Wifi with this and putting a webcam in place of the BT module.

R

[PandorasThinkpad]

Well, flyingfishfinger,

good luck that it will work this time, all you're looking foward to mod on this machine.

Some steps seem to be very risky....but that's quite the same with modding anything.

Me by myself, i have completely given up the X61 models.....and instead of that i use X201 machines with I7.

Maybe, when i will be able to get an not so expensive X61 with T8100 or T8300 and already done display mod...this will be the only moment for me returning to X61 again....and i have had in both ways the most powerful machines, an X61 with t9300 and an X61s with L7700.

So again good luck in anything you are going to mod again.

Sincerly,

PAnDoRa

[wileE]

Also investigating replacing the Wifi with this and putting a webcam in place of the BT module.

You do know that the X61 WLAN slot does not have USB lanes? So if the Bluetooth part of that card uses USB...

Where did you grab the 5V for the additional internal USB 2.0 connection? And did you cut the two traces on the back?
I did the mod on two boards, and found that without the cut, some of my USB sticks will not boot on that port. Otherwise no differences.

[flyingfishfinger]

For the HDMI, I took 5V from the gold-fingered port I've mentioned before somewhere (front center of the motherboard, near the WiFi switch) and ran the wires around the 2nd fan.

I just got up the courage to do the frame modification for the SXGA screen, but I have to head home from work now so I didn't finish yet...

I still have some thinking to do about making the 2nd HDD bootable or not (switching the SATA port with an 8-way switch? Seems legit. Not sure where to stash them).

Meanwhile, I made a triple-boot Hackingtosh out of an old HDD. OS X Lion, Win 7 and Ubuntu 14.10

R

[flyingfishfinger]

Update: Here's the USB 3 card and the USB3 -> SATA adapter installed. I haven't wired it to the HDD bay yet, but both the USB and the SATA adapter report and install fine (in Windows at least). Also, the HDMI card next to my SSD for size.
Slowly but surely it's coming along....

Image on Gdrive

R

EDIT: Seems like the fingerprint reader is in the way where the SATA card currently is. Exploring alternate placements