So apparently humble bundle is trash.
Click the "update preferences" at the bottom of a humble bundle mail and rewrite the url to your hearts content to randomly update some poor bastards mail preferences.... wow.
[QUOTE=shadowboy303;52684782]So apparently humble bundle is trash.
Click the "update preferences" at the bottom of a humble bundle mail and rewrite the url to your hearts content to randomly update some poor bastards mail preferences.... wow.[/QUOTE]
I remember the hype 6 years ago, then I didn't bother with it because I realized I haven't even touched any of the games I bought (or was gifted) on Steam..
[t]http://horobox.co.uk/u/reag/WP_20170915_20_29_16_Pro%5B1%5D.jpg[/t]
Remember kids, never leave a GPU to sit around for years without checking on it.
This shit was cemented on [B]hard.[/B] Thought I was gonna snap it with the force I was applying.
[editline]15th September 2017[/editline]
[t]http://horobox.co.uk/u/reag/WP_20170915_20_46_29_Pro%5B1%5D.jpg[/t]
Fucking stubborn shit.
So disassemble while in storage, got it!
Heat it up with a hair dryer/heat gun before, it'll get a bit smoother.
Friend once had an AMD Phenom II x2 that he unlocked to 4core. He decided to upgrade to 960T that he unlocked to 6core.
It got unstable and he decided to upgrade the case and PSU and the cooling. maybe mobo etc too, don't remember.
well, we got to the part of extracting the 960T out of the socket. The PC had been off for like 3h so it was all cooled down and the old paste was full on glue. We finally almost like ripped it off and saw that the CPU was stuck to the stock cooler. Then I got the smart idea of using a hair dryer to heat it up to get it off. The CPU/cooler got so hot to touch that the moment the glue didn't hold it anymore, friends hands couldn't hold the CPU and he dropped it to the ground. 10 minutes later we had all the pins straight again and it continued working until his new PSU died again.
[QUOTE=Reagy;52684824][t]http://horobox.co.uk/u/reag/WP_20170915_20_29_16_Pro%5B1%5D.jpg[/t]
Remember kids, never leave a GPU to sit around for years without checking on it.
This shit was cemented on [B]hard.[/B] Thought I was gonna snap it with the force I was applying.
[editline]15th September 2017[/editline]
[t]http://horobox.co.uk/u/reag/WP_20170915_20_46_29_Pro%5B1%5D.jpg[/t]
Fucking stubborn shit.[/QUOTE]
That's how I broke my 8800GT, except it was only a year old by the time the paste cemented and it's also one of many reasons why I vowed to never buy Asus again.
[QUOTE=Protocol7;52684973]That's how I broke my 8800GT, except it was only a year old by the time the paste cemented and it's also one of many reasons why I vowed to never buy Asus again.[/QUOTE]
This is the last Nvidia card I bought because at the time, their drivers always BSOD'd for me.
Been using AMD ever since, but with new parts coming, gonna be moving my 290x into new build and this into my old as the motherboard has no display output.
Eitherway got it cleaned up and with some new paste on it, sadly I've only had AS5 on hand currently (non-conductive but we all know its capacitive as fuck so its a coin flip) so I'm hoping it wont kill it, if it does then I've always got a spare 9400GT as well with a hilarious passive cooler. :v:
[editline]15th September 2017[/editline]
Holy fuck I forgot how small the 9400GT's die is. :v:
[t]http://horobox.co.uk/u/reag/WP_20170915_21_37_25_Pro%5B1%5D.jpg[/t][t]http://horobox.co.uk/u/reag/WP_20170915_22_00_18_Pro%5B1%5D.jpg[/t]
"Dummy dies" AMD said.
[t]https://s.gvid.me/s/2017/09/15/AMD-Ryzen-Threadripper-4-dies-1.jpg[/t]
Those are the 4 threadripper dies sanded down.
:thinking:
[QUOTE=glitchvid;52685726]"Dummy dies" AMD said.
[t]https://cdn.videocardz.com/1/2017/09/AMD-Ryzen-Threadripper-4-dies-1.jpg[/t]
Those are the 4 threadripper dies sanded down.
:thinking:[/QUOTE]
Not Threadripper dice, Zen dice. Ryzen, TR and EPYC are all the same die.
But really, it's fairly clear that Threadripper is faulty EPYC packages, clearly done in the way they are for cost and yield reasons.
What the dummy dice thing most likely is, is when yields are good enough that TR can be its own production line, instead of using partially-faulty EPYC packages, they will have dummy dice purely for packaging reasons and be done with it.
[QUOTE=glitchvid;52685726]"Dummy dies" AMD said.
[t]https://cdn.videocardz.com/1/2017/09/AMD-Ryzen-Threadripper-4-dies-1.jpg[/t]
Those are the 4 threadripper dies sanded down.
:thinking:[/QUOTE]
[t]https://www.helifreak.club/image/20170916012339584.png[/t]
Very nice.
[editline]16th September 2017[/editline]
Oh my referrer control addon turned itself off and fucked everything.
[QUOTE=wingless;52685750]Not Threadripper dice, Zen dice. Ryzen, TR and EPYC are all the same die.
But really, it's fairly clear that Threadripper is faulty EPYC packages, clearly done in the way they are for cost and yield reasons.
What the dummy dice thing most likely is, is when yields are good enough that TR can be its own production line, instead of using partially-faulty EPYC packages, they will have dummy dice purely for packaging reasons and be done with it.[/QUOTE]
Well, technically Zeppelin dies with Zen cores, but that's just being pedantic. Point is they're from a Threadripper 1950X.
The issue is that AMD hasn't been totally forthcoming about their existence. The actual dies would be tested well before being put on a package, so are they just EPYC processors where the actual packaging process kills the dies? (unlikely that number is high enough to shift the volume Threadripper is commanding). So why not make the heat spreader differently and not need the two fully diffused dies?
They're also apparently fully packaged onto the whole thing, so there's a chance they're literally EPYCs with soft-disabled dies.
With how much power the X399 boards are spec'd to, maybe AMD will release a 32-core EPYC once Intel gets their 18 core monolitics out the door. Afterall they did just release an 8-core version, and hadn't mentioned that at all during the initial reveal.
[QUOTE=helifreak;52685771]Oh my referrer control addon turned itself off and fucked everything.[/QUOTE]
Big websites scared of a little hotlinking, I'll just reup on mine.
[QUOTE=glitchvid;52685783]Well, technically Zeppelin dies with Zen cores, but that's just being pedantic. Point is they're from a Threadripper 1950X.
The issue is that AMD hasn't been totally forthcoming about their existenace. The actual dies would be tested well before being put on a package, so are they just EPYC processors where the actual packaging process kills the dies? (unlikely that number is high enough to shift the volume Thredripper is commanding). So why not make the heatspreader differently and not need the two fully diffused dies?
They're also apparently fully packaged onto the whole thing, so there's a chance they're literally EPYCs with soft-disabled dies.
With how much power the X399 boards are spec'd to, maybe AMD will release a 32-core EPYC once Intel gets their 18 core monolitics out the door.[/QUOTE]
Think of this logistically. What's the one thing both EPYC and TR do with Zeppelin that Ryzen doesn't that give it a unique position? PCIe lanes. EPYC has 128 lanes, that's 32 per die, which is every single available one, Ryzen only does 24, TR obviously does 64.
Number of failed cores doesn't matter, because you just adjust your sku's based on that like any other cpu, but it means that for EPYC and TR, you cannot have any IO failure. This is where TR comes in as it simplifies production of EPYC and gives AMD a very price competitive HEDT option that, overall, makes AMD look pretty damn good.
So those dice could very well have 100% fully working cores, but they absolutely do not have working IO to whatever degree. It's quite possible we will see 32-core TR (And I kinda want to just for the sake of it) but I think that won't be happening until they can sustain better yields in general.
In general, Zen has had such an awesome focus on cost-effectiveness and I think TR really speaks well of the success of it all.
Is there a browser extension that can allow you to hotlink on websites that don't allow hotlinking? That would be super useful.
[QUOTE=wingless;52685796]Think of this logistically. What's the one thing both EPYC and TR do with Zeppelin that Ryzen doesn't that give it a unique position? PCIe lanes. EPYC has 128 lanes, that's 32 per die, which is every single available one, Ryzen only does 24, TR obviously does 64.
Number of failed cores doesn't matter, because you just adjust your sku's based on that like any other cpu, but it means that for EPYC and TR, you cannot have any IO failure. This is where TR comes in as it simplifies production of EPYC and gives AMD a very price competitive HEDT option that, overall, makes AMD look pretty damn good.
So those dice could very well have 100% fully working cores, but they absolutely do not have working IO to whatever degree. It's quite possible we will see 32-core TR (And I kinda want to just for the sake of it) but I think that won't be happening until they can sustain better yields in general.
In general, Zen has had such an awesome focus on cost-effectiveness and I think TR really speaks well of the success of it all.[/QUOTE]
My point is that, if you're going to be making a platform based on (as you said) 64 PCIe lanes, that, exactly 2 dies can deliver, why bother adding 2 more whole diffused dies? You could either redesign the head-spreader, make a smaller socket, etc, etc. The real question here is, [I]why[/I] is threadripper on a full 4094 pin socket (The same as EPYC, just with some very slight physical differences apparently). On a full 4094-pin sized package, with 2 fully diffused "dummy" dies? Nothing about ∞ fabric requires that set up, especially since as you've said, X399 is a 64 lane (2 Zeppelin dies worth of I/O) design.
The only thing I could think of is that threadripper is literally EPYC with soft-disabled dies, but physically functional cores, fully wired. Or they're planning on doing 32-core/4die packages in the short future (again, probably after Intel releases monolithic x299s).
As for the dead I/O, as you've said, they can probably bin those and sell them as lower Ryzens, why package binned and usable silicon as dead-weight on a HEDT platform? Furthermore, AMD is apparently getting 80X% (total) yield, and with how tiny the I/O portion of the dies is, I doubt there's any significant amount of Zeppelins that are only K/O because of defects on the I/O.
I just wish AMD was more open about it, since then we could make sense of why there are two, fully diffused dies, going unused, in all threadripper packages.
[QUOTE=glitchvid;52685833]My point is that, if you're going to be making a platform based on (as you said) 64 PCIe lanes, that, exactly 2 dies can deliver, why bother adding 2 more whole diffused dies? You could either redesign the head-spreader, make a smaller socket, etc, etc. The real question here is, [I]why[/I] is threadripper on a full 4094 pin socket (The same as EPYC, just with some very slight physical differences apparently). On a full 4094-pin sized package, with 2 fully diffused "dummy" dies? Nothing about ∞ fabric requires that set up, especially since as you've said, X399 is a 64 lane (2 Zeppelin dies worth of I/O) design.
The only thing I could think of is that threadripper is literally EPYC with soft-disabled dies, but physically functional cores, fully wired. Or they're planning on doing 32-core/4die packages in the short future (again, probably after Intel releases monolithic x299s).
As for the dead I/O, as you've said, they can probably bin those and sell them as lower Ryzens, why package binned and usable silicon as dead-weight on a HEDT platform? Furthermore, AMD is apparently getting 80X% (total) yield, and with how tiny the I/O portion of the dies is, I doubt there's any significant amount of Zeppelins that are only K/O because of defects on the I/O.
I just wish AMD was more open about it, since then we could make sense of why there are two, fully diffused dies, going unused, in all threadripper packages.[/QUOTE]
Cost. The reason why it's the same socket, same package, etc is that it means you have one assembly line for two different products, while gaining the benefits of both. You don't design another socket because that's expensive. You use the same package because then you just throw in slightly lower binned dice into your EPYC line and throw some more caution to the wind, and those that fail testing on enough IO post-packaging get thrown into a TR box. It means for very minimal R&D and very minimal production line changes, they can get a whole new product line out of it.
The only real engineering and production costs you then have is chipset, which compared to the rest is negligible, and chipset is entirely the thing that makes the difference between EPYC and TR at this point.
It is entirely cost, it is entirely logistics, and the way TR works means that it is incredibly simple and close to free to actually produce, just piggy backing off of EPYC.
It also means you get better utilisation of yields. You have a much more consistent, even spread of ASIC quality among your product lines, you get far better yield utilisation, you get far better cost effectiveness. TR has a higher price premium than Ryzen, and it's going to have a higher minimum bar for ASIC quality than Ryzen will, but lower than EPYC. It slots right in the middle. You get a better return on your higher quality dice, you pay back ROI quicker, you make profit quicker. That's what it comes down to.
It's also worth noting that the 80% yield number is probably relative to the minimum bar required for Ryzen, not TR or EPYC.
[QUOTE=garychencool;52685798]Is there a browser extension that can allow you to hotlink on websites that don't allow hotlinking? That would be super useful.[/QUOTE]
[URL="https://addons.mozilla.org/en/firefox/addon/referrer-control/"]This[/URL], and I assume maybe [URL="https://chrome.google.com/webstore/detail/referer-control/hnkcfpcejkafcihlgbojoidoihckciin"]this one[/URL] works if you're using Chrome.
[QUOTE=helifreak;52685849][URL="https://addons.mozilla.org/en/firefox/addon/referrer-control/"]This[/URL], and I assume maybe [URL="https://chrome.google.com/webstore/detail/referer-control/hnkcfpcejkafcihlgbojoidoihckciin"]this one[/URL] works if you're using Chrome.[/QUOTE]
I just have an extension that edits every header
[t]https://s3.wasabisys.com/tenryuu/astrid/2017-09/17-09-16_12-10-22-Add-ons_Manager_-_Waterfox.png[/t]
[QUOTE=wingless;52685838]Cost...
It's also worth noting that the 80% yield number is probably relative to the minimum bar required for Ryzen, not TR or EPYC.[/QUOTE]
I just don't believe that putting thermal pads, or basically [I]anything[/I] else where the two "dummy" dies are more expensive than more or less defective silicon, it wouldn't even require any more substantial change to the production line than would cost to soft-disable fully functioning EPYC packages. I also doubt basically any of the Threadripper packages dead dies are because of packaging related 'deaths'.
As for yield, I believe that's total silicon-> CPU product (Though AMD and GF are understandably tight-lipped). As for binning on Ryzen, IIRC all Ryzen products are full 24-lane enabled. And as I've said, I/O is a tiny part of the actual Zeppelin die, so I doubt they have the quantity of JUST I/O dead zeppelin dies (And even if they did, they'd have to sort them, pick them, and place them specifically on 'Threadripper' bound EPYC packages...).
I mean I guess it worked out for AMD, but they've been typically tight lipped.
[editline]later[/editline]
AMD (before this recent revelation) on the topic:
[quote]The question that many have been asking is what the purpose of the other two dies are, with the answer simply being structural. These extra dies prevent imbalance and allow for simple cooler mounting without any chance of damaging the CPU. [B]These two extra dies are not active or even contain working transistors[/B], [U]they are blank [/U]and as such are not "wasted Ryzen CPU dies".
To make a long story short, the two extra dies in Threadripper are merely structural inserts, with the active dies being placed in a diagonal configuration in all Threadripper CPUs to give them all the same thermal profiles. Threadripper CPUs are not "failed EPYC" CPUs and they certainly are not made in a wasteful way by AMD.
[/quote]
:glare:
[QUOTE=Scratch.;52685863]I just have an extension that edits every header
[t]https://s3.wasabisys.com/tenryuu/astrid/2017-09/17-09-16_12-10-22-Add-ons_Manager_-_Waterfox.png[/t][/QUOTE]
So do I.
[t]https://www.helifreak.club/image/20170916021955431.png[/t]
But I'm lazy and this was easier.
[QUOTE=glitchvid;52685886]I just don't believe that putting thermal pads, or basically [I]anything[/I] else where the two "dummy" dies are more expensive than more or less defective silicon, it wouldn't even require any more substantial change to the production line than would cost to soft-disable fully functioning EPYC packages.[/QUOTE]
This is because you don't understand materials science or manufacturing.
Replacing dies under package:
"Thermal" pads: soft material, differing elasticity compared to packaged die. Causes uneven mounting - the SoC is pressed against the mount a lot more than any other mounting solution. Now you have a CPU where a die will crack as the user is mounting in. Also complicates manufacturing since it has to survive intact through the indium soldering process.
Replacement block: requires a custom manufacturing line that does nothing but builds dummy blocks for mounting purposes. May cause problems in soldering due to difference in material, which maybe cause uneven mounting pressure cracking a die.
Using existing dead dies under the IHS:
* Dead dies are detected in testing before packaging - a sufficient number of dead dies is generated because of symmetry requirements (die requires 2 CCX with equal number of cores, so a 3+2 would never be used for a CPU)
* Same material as functional dies, behaves equally as long as your soldering process is working to spec.
* All packaged dies are tracked individually, so your only change requirement to the Epyc packaging line is an additional configuration to the pick and place machines, where you can load in an extra tray of dead packaged dies, and perhaps interposers. Each MCM is tracked individually on the line, so you can simply separate out Threadripper and Epyc MCMs from each other at the end.
Edit:
I should note, if they get a die in testing with up-to-spec CCX configuration of 3+2, they can either use that for mounting Threadrippers, or laser them into a 2+2, getting them a Ryzen 3, which can be sold to OEMs for low-end business desktops.
This gives them flexibility (get a bunch of dies for Threadripper mounting if Ryzen 3 isn't selling well enough, or turn the dies into Ryzen 3s if there's not too much interest in Threadripper).
[QUOTE=helifreak;52685901]So do I.
[t]https://www.helifreak.club/image/20170916021955431.png[/t]
But I'm lazy and this was easier.[/QUOTE]
Mine has regex support :smug:
Any of you guys got any experience playing 4K blu-ray on PC? It is worth it? My internet is too slow to stream 4K movies.
When you dock too hard
[quote]
[CLOSED] done 413.6 MB Rate: 0.0 / 0.0 KB Uploaded: 0.0 MB [ R: 0.00] Inactive: A
[Throttle 512/102400 KB] [Rate 0.0/ 0.0 KB] [Port: 45566] [U 0/112] [D 0/20490] [H 0/32] [S 0/3/1048064] [F 256/256]
Shutting down...
PS C:\WINDOWS\system32>[/quote]
o7
[QUOTE=Scratch.;52686659]When you dock too hard
o7[/QUOTE]
Question is, why the fuck are you trying to run rtorrent on windows?
[QUOTE=helifreak;52686667]Question is, why the fuck are you trying to run rtorrent on windows?[/QUOTE]
technically
linux
[QUOTE=RoboChimp;52686399]Any of you guys got any experience playing 4K blu-ray on PC? It is worth it? My internet is too slow to stream 4K movies.[/QUOTE]
Don't bother. Only three, I believe, UHD-BD drives for PC even exist at this point, and are a little pricey (Range anywhere from $150USD to $400USD), all made by Pioneer.
Next, in terms of software I believe the only piece of software capable of UHD-BD playback is Cyberlink PowerDVD, so your options are extremely limited.
Then you have the real kicker, minimum hardware requirements. As in, the only thing the ecosystem will let you do any of this on is Kaby Lake minimum (Not even Ryzen, due to lack of SGX), on top of this, making sure you have proper HDMI 2.0a output with HDCP 2.2 (No displayport, HDMI only).
Frankly, you're probably better off just getting a damned Xbox One S and sitting it next to your PC and just switching to a HDMI input on your display. It is such a shitshow.
[QUOTE=wingless;52686738]Don't bother. Only three, I believe, UHD-BD drives for PC even exist at this point, and are a little pricey (Range anywhere from $150USD to $400USD), all made by Pioneer.
Next, in terms of software I believe the only piece of software capable of UHD-BD playback is Cyberlink PowerDVD, so your options are extremely limited.
Then you have the real kicker, minimum hardware requirements. As in, the only thing the ecosystem will let you do any of this on is Kaby Lake minimum (Not even Ryzen, due to lack of SGX), on top of this, making sure you have proper HDMI 2.0a output with HDCP 2.2 (No displayport, HDMI only).
Frankly, you're probably better off just getting a damned Xbox One S and sitting it next to your PC and just switching to a HDMI input on your display. It is such a shitshow.[/QUOTE]Bloody hell, you have to move to kabylake, use that annoying cyberlink software and use HDMI? The monitor I'm saving up for only supports 4:2:0 colour over HDMI @4K. Perhaps I'll just stick to 1080p movies for now.
Thanks, you've probably saved me about $220.
[QUOTE=RoboChimp;52686813]Bloody hell, you have to move to kabylake, use that annoying cyberlink software and use HDMI? The monitor I'm saving up for only supports 4:2:0 colour over HDMI @4K. Perhaps I'll just stick to 1080p movies for now.[/QUOTE]
It gets better too. Kaby Lake-X doesn't work since for some reason intel decided to not include SGX on it, so X299 is not an option. Plus, but I'm not certain on this, it may even have to be motherboard HDMI, not discrete GPU, simply due to DRM paths. Which also means that only some motherboards will work, since not all Z270 boards do HDMI 2.0a, some just do plain 1.4.
Seriously, such a shitshow.
[QUOTE=wingless;52686816]It gets better too. Kaby Lake-X doesn't work since for some reason intel decided to not include SGX on it, so X299 is not an option. Plus, but I'm not certain on this, it may even have to be motherboard HDMI, not discrete GPU, simply due to DRM paths. Which also means that only some motherboards will work, since not all Z270 boards do HDMI 2.0a, some just do plain 1.4.
Seriously, such a shitshow.[/QUOTE]Fuck, it seems like pioneer will sell about 5 of these drives. I can play 4K video files on my PC, why'd they opt for all that insanity just for 4K physical media? It seems like the target market is TV users who unless they have an insanely big TV, wouldn't even notice the difference sitting that far back, that must be why they keep push that HDR thing.
Also, is X299 good for anything? Gamers and casual users will get Z270, workstation users will get AMD or Xeon, who does that leave?
[QUOTE=RoboChimp;52686867]Fuck, it seems like pioneer will sell about 5 of these drives. I can play 4K video files on my PC, why'd they opt for all that insanity just for 4K physical media? It seems like the target market is TV users who unless they have an insanely big TV, wouldn't even notice the difference sitting that far back, that must be why they keep push that HDR thing.[/quote]
Blame the media industry and their insane beliefs on DRM, you know how this goes.
Sorry, you need to Log In to post a reply to this thread.
