Wow, never heard of a bracket breaking like that unless it was traumatized by excessive mechanical force. Looks like you figured out a shade-tree engineering fix for the mean time. I wonder if the material keeping the screw from pulling through is going to be sufficient considering the plastic tab already has broken under the spring tension. You might look around for scavenged old AMD socket boards to see if you can liberate another socket tension bracket while you wait for Zen. I haven't heard of a timeline for AM4 motherboards and CPUs that soon though. I think they won't hit until 1Q 2017 at the earliest. Great pictures of the event though.

---

Seti@Home classic workunits:20,676 CPU time:74,226 hours

A proud member of the OFA (Old Farts Association)

Yeah, I'm not 100% confident on what little material is holding that screw in-place. Hence the safety wire I tied to the cooler. I don't move the rig around, and I don't make it a habit to take the heatsink off frequently. In fact.. the last time I touched the latch on the tension strap to lock the heatsink down was in January 2012--when I built the rig, and it's been in the same place on the desk ever since, with the exception of carrying it out to the garage about once per year to blow it out with the air compressor.

I guess it's just one of those things where plastic breaks down with heat + age, and isn't helped by having an untold amount of tension on it.

Since I have some wiggle room between the main heat block for the cooler and the bracket itself, I was thinking I could put another screw through down lower (where there's still a lot of meat) and wrap some baling wire around both screws (so that the lower one pulls the current one down toward the PCB). Or something like that.

If it breaks again, that's probably what I'll try next.





As far as the timeframe for Zen..

- In July, PC Advisor says October, though that is according to this WCC article which is from 6 months ago.
- PC Games N seems to reference that WCC article for all their info, too.
- a few others reference the press release info and WCC article

I would have figured boards would be on the market now, but I think it would be a case of "why buy a board when there's no CPU to put on it yet?". Apparently WCC got it's street date of October from a mobo manufacturer.

But from the press details AMD has released so far, 8c/16t Zen and a 8c/16t Haswell-E (i7-6900K) matched at the same GHz, Zen wins, and uses 40w less power. [it was close, but.. Zen did win: https://youtu.be/oQS8s7TOXsE ]. 32c/64t server CPUs look tasty too.

Exciting times are coming.

---

Linux laptop:
record uptime: 1511d 20h 19m (ended due to the power brick giving-up)

Yeah, I'm not 100% confident on what little material is holding that screw in-place.

Artic has some really decent coolers for ~$20 at Amazon. I've used 5-6 of them, and been very pleased. And the parts that are under tension in their design are steel, not plastic.

---

I went ahead and used a strand of cat-5 to put a check-strap on the heatsink onto the PSU support brace so that at least if it lets go again, the 800g heatsink isn't going to plummet down onto the GPU again.

I've got ideas for how to improve that fix.... but I ran out of motivation.

Both of my systems are lying on their side, so the heatsinks are vertical.
Tower case in desktop orientation.

---

Grant
Darwin NT

I went ahead and used a strand of cat-5 to put a check-strap on the heatsink onto the PSU support brace so that at least if it lets go again, the 800g heatsink isn't going to plummet down onto the GPU again.

I've got ideas for how to improve that fix.... but I ran out of motivation.

Both of my systems are lying on their side, so the heatsinks are vertical.
Tower case in desktop orientation.

I've often joked to my clients about the three people who designed computer assemblies.

The motherboard designer, who mounted a circular DIN or PS/2 socket on the backplate, in a standardised orientation.
The keyboard designer, who moulded the word 'top' onto the circular plug casing, in the matching orientation.
The case designer, who put the rubber feet on the short side of the case...

I got most of my information from the slides presented at the Hot Chips conference earlier last month and all the summations from the tech magazines. Noticed this article today about the Vega and Zen production estimations. Posted a day ago.

I have hope that they won't pull a "Bulldozer" in expectations. I hope they will provide a good "bang for the buck" product that is better, faster than their current and very old technology, and even if it doesn't compare to Intel product, at least they price it very appealingly.

---

Seti@Home classic workunits:20,676 CPU time:74,226 hours

A proud member of the OFA (Old Farts Association)

There may be certain portions of Zen coming in Q1'17, but WCC seems pretty confident that there's going to be an FX CPU out in October. I did read with the articles I provided that the 16-core SOC APU that is expected to go into the xbox won't be ready until at least January, and most of the APU line-up won't come until Q1, but the enthusiast CPU (FX) is still going to be "Q4", but more like October, according to "reliable sources" within at least one mobo manufacturer.

Of course, maybe October is when mobo manufacturers will get their engineering sample so they can make sure the products they've built and designed are actually going to work. That is a real possibility.



But other than that, it seems pretty promising that the pricing scheme will be similar to the past. I read in one article that the 8c/16t FX model is expected to be in the same price range as a 4c/8t i7.

And yeah, I've been hoping for six+ months that Zen doesn't end up falling on its face like Bulldozer.. when Bulldozer was touted as being the "i7 killer" and when it came out.. it was on-par with a mid-range i5 at best. I read in some press release earlier this year that AMD said they are "for sure definitely" going back to a single FPU per physical core and abandoning the "shared FPU" design--which is what made Bulldozer fall on its face in the first place.

---

Linux laptop:
record uptime: 1511d 20h 19m (ended due to the power brick giving-up)

The article I linked in my previous post has slide captures which show pretty much the entire microarchitecture of Summit Ridge. Each core gets its own FPU register so no sharing like Bulldozer.

The engineering samples are already long in the hands of the motherboard manufacturers as well as the software developers. I have seen a screen shot already of a SIV application window running a ES Zen 4 core @2.2Ghz from SIV's developer.

---

Seti@Home classic workunits:20,676 CPU time:74,226 hours

A proud member of the OFA (Old Farts Association)

Hmm... this doesn't look too promising...

Microsoft is slamming the door on PC builders and upgraders who might have hoped to use the new Intel Kaby Lake or AMD Zen chips for Windows 7 or Windows 8 PCs. Sorry: Both chips are officially supported only by Microsoft’s Windows 10.

Microsoft's mandate is discreet rather than secret. In January, the company tried to shorten its support lifecycle for Intel Skylake PCs running Windows 7 and 8, a policy the company subsequently abandoned after much outcry. But Microsoft’s statements have also consistently included a critical caveat: The latest generations of silicon—specifically Intel’s Kaby Lake chip, Qualcomm’s 8996, and AMD’s Bristol Ridge silicon—will all require Windows 10.

Here's the obvious question: What would happen if a naïve or not-so-naïve user attempted to run Windows 7 or Windows 8 on a Kaby Lake or Zen system? Without actual chips to test, the answer is unknown.

One source privately guessed that the processor would boot, though without driver support and security updates the experience would be “a bit glitchy.” Without specific support for a chip’s features—such as the dedicated video processing logic within Kaby Lake, for example—certain apps, if not the OS itself, might crash, another said.

Microsoft has laid out its rationale for encouraging users to adopt Windows 10: In short, it’s a more manageable, secure operating system with better collaboration across users and devices, the company claims. Nevertheless, a substantial chunk of users can’t see past Microsoft’s attempts to force Windows 10 upon them.

Of late, the carrots Microsoft has used to entice Windows 10 adoption—the Insider program, free upgrades from older operating systems, and synergy with Windows phones, the Xbox One, and even iOS and Android phones—have been largely ignored. Instead, Microsoft has come under fire for its ongoing program of forced upgrades, locking down Cortana to exclusively use Bing, and what some users see as a concerted attempt to mine personal data for advertising purposes.

We don't know whether terminating support for older operating systems on Kaby Lake and Zen actually means they won’t work. What seems more certain is that Microsoft’s latest strategy will be seen as just another “stick” wielded to force customers to upgrade to Windows 10.

---

Linux laptop:
record uptime: 1511d 20h 19m (ended due to the power brick giving-up)

I see that the donation/green star boondoggle has not been corrected whilst the kitties were off on walkabout.....

Meow?

---

"Freedom is just Chaos, with better lighting." Alan Dean Foster

I got back a "lost task" (from over a week ago) on one of my scheduler requests this evening, without having to go through the usual convoluted rigamarole to get it. Perhaps that feature was turned back on late in the [Berkeley] day. Anybody else with ghosts see this on their boxes? It would appear as a line in the Event Log like:

SETI@home | Resent lost task blc2_2bit_guppi_57403_HIP11048_OFF_0007.16487.831.22.45.153.vlar_0

EDIT: Ah, I think I figured out what happened. My crunch-only machines shut down automatically on weekday afternoons, to avoid peak electric rates. The one just happened to shut down in the middle of a scheduler request, apparently after it reported a task but before it received the acknowledgement. Then, when it started back up again 5 hours later, it ended up reporting the same task, thus triggering the "resend lost tasks" event. Much easier than what it usually takes to fool the scheduler to resend those suckers!

Very interesting. PFB splitter on the Vader is working, and on the Lando are definately stuck...

---

It certainly does appear that the 3 splitters working on 23ja09aa have been stuck for a day or so, now. Maybe somebody in Berkeley needs to give them a kick in the morning. I've gotten very few Arecibo tasks today, just lots and lots of Guppi VLARs.

There should be 8 pfb splitters, only four is running.

It seems that 4 splitters are Disabled by staff. Two in LANDO and two in VADER.

Panic Mode On?????

Not sure where to put this.. so I'll put it here.

Zen CPU pins and AM4 socket pictured, with explanation of why it needs 1331 pins.

and.. Zen is coming in Jan/Feb, with the debut/launch being at CES. Vendors and OEMs are supposed to be getting the mobos starting in October.

I read somewhere a few weeks ago that the top-end FX processor is likely to be in the £225 (US$295) range, and based on that press demo that compared it to an i7-6900K (US$1095), I think that bodes very well for AMD this time around. They don't necessarily have to stomp on top-end i7's, but if they can be pretty much on-par for a third of the price, that puts AMD in a very good position.

I'm excited about this. I figure that a month or two before the CES debut/launch, some independent benchmarks from various sites should come out and that should paint a real picture of how Zen compares. One thing I'd like to see is instead of how it performs at doing 4K resolution on three games, and compressing various archives... I'd like to see one of these review sites run it through Cinebench. That's where you'll be able to see just how efficient it is at a multi-threaded task. (My FX-6100 gives me 3.95x versus a single-thread, which means you effectively lose 1/3 of the cores due to the shared FPU. I'd like to see what the multi-processor speed-up is for the 8c/16t Zen.)

---

Linux laptop:
record uptime: 1511d 20h 19m (ended due to the power brick giving-up)

I'd like to see one of these review sites run it through Cinebench. That's where you'll be able to see just how efficient it is at a multi-threaded task.

Both AnnandTech & the Tech Report do that, with both single threaded & multithreaded runs.

---

Grant
Darwin NT

A lot of Astropulse tasks finished after 2 sec. What's the matter?

---

A lot of Astropulse tasks finished after 2 sec. What's the matter?

We're trying to work that out in the Anything relating to AstroPulse tasks thread.

Two possible suggestions so far:

a) Some recently-processed Arecibo recordings were made during a radio-astronomy survey of the Moon. It's possible that these were strongly contaminated by radio-frequency interference (at the sensitivity at which we search for ET), perhaps even by reflection of earth-based interference.

b) For a long time - several years - one of the Astropulse channels (B3_P1) has been plagued by an error, probably an intermittent hardware failure, somewhere in the antenna - A-D converter - receiver chain. A similar failure may have started affecting other channels as well.

It would be interesting to know which it is - or maybe it's something else entirely.

This morning I got a whole bunch of non AP from Arecibo that are all blank.

Mike

---

Is anybody having a problem with CPU units ?

At the moment I'm doing CPU units in 14sec yes that is correct 14sec

I now have over 360 valid units , only thing is I had stopped seti from downloading and forgot it was set like that and about 1 hr ago I let it start downloading and for some reason I keep getting CPU units being done in 14 sec I'm doing them faster than I can download them .

"Something is wrong Scotty I need more power "" I'm giving you all I can Capitan" "then there's something wrong look into it Scotty" "right away Capitan"

---