|
3.7GHz Zen 8c/16t will be worse in a 4t game than a 4.6GHz Skylake 4c/8t. But if you could clock your 8c/16t Zen at 4.2GHz, I think people wouldn't mind being 500MHz behind in exchange for twice the cores.
|
#
?
Dec 17, 2016 05:46
|
|
|
|
| # ? May 16, 2024 04:20 |
|
|
SwissArmyDruid posted:the window will close by the time the back half of 2017 comes around with Cannonlake. Hell that'd even be true even if AMD 'only' got Haswell levels of single/multi thread performance. Haquer posted:Just worse than Skylake, intel's current offering. If they've really pulled off Broadwell-esque single/multi thread performance then its a pretty big deal. Intel will still win the synthetic benches by significant margins but in real world apps the difference will be negligible. If they price it right they'll sell a ton of them. And if those rumored price lists turn out to be accurate I'll almost certainly end up buying the 8C/16T part. EDIT: \/\/\/\/\/\/ Here is to hoping they're setting clockspeeds within the power or temp limits (with the stock HSF) they set for themselves and they're not actually process or design limited with what they can reach. Its quite possible that Zen will get over 4Ghz, maybe even well over that, if so but it'll probably turn into a furnace. PC LOAD LETTER fucked around with this message at 08:35 on Dec 17, 2016 |
|
#
?
Dec 17, 2016 06:35
|
|
|
Have hope for Zen overclocking though, someone delidded a Bristol ridge APU and they use either sodder or some other liquid metal as TIM. No way they use an inferior material for Zen.
|
|
#
?
Dec 17, 2016 06:56
|
|
|
SwissArmyDruid posted:My point is that even as Intel has left an opening for AMD to catch up, they have not done so, and even then the window will close by the time the back half of 2017 comes around with Cannonlake. given how much kaby lake has "improved"  , i wouldn't get excited about intel jumping ahead all that much on the same node
|
|
#
?
Dec 17, 2016 09:36
|
|
|
Yera Intel is already looking ahead at 2018 with desktop Cannonlake, which gives me the impression that if not performance competitve, Intel thinks AMD will be price competitive with Haswell-E/Broadwell/Skylake.
|
|
#
?
Dec 17, 2016 09:43
|
|
|
Wonder how effective AMD's power saving stuff actually is. And how well this does in games at 1440p.
|
|
#
?
Dec 17, 2016 15:21
|
|
|
Anime Schoolgirl posted:current intel roadmap says cannonlake 10nm will be almost exclusively mobile chips that are 2 cores, coffee lake will stay 14nm which means only uarch improvements and little else Funny thing about that, I actually think that Intel hosed up Kaby Lake's TIM again, and not because Kaby Lake is bad. https://forums.anandtech.com/threads/my-i7-7700k-has-arrived-insights-benchies-overclocks-inside-now-with-delid.2493250/ That's an overclock to _5.3 GHz_. (Yup, we're doing THAT again.) edit: Oh hey, WCCFT finally picked up the article. SwissArmyDruid fucked around with this message at 06:41 on Dec 18, 2016 |
|
#
?
Dec 18, 2016 05:39
|
|
|
Whoah, if 5GHz is doable for the majority of these delidded. I wonder if the memory controller will clock better, also.
|
|
#
?
Dec 18, 2016 09:19
|
|
|
That is much better (~25% OC) but if you have to delid to do it probably not many will get that. I think before the delid he was getting to ~4.7Ghz stable which is much more ho hum. It may be that if Intel just went with a shim/bare die instead of a mediocre IHS/TIM they'd really have a interesting item on their hands. edit: actually those rockit 88 delid tools look pretty easy to use. https://www.youtube.com/watch?v=bwg0HRw17lY aren't too expensive either http://rockitcool.myshopify.com/collections/all Razor method still works I believe but if you're terrified of killing your chip that tool seems to be the way to go. Vise method might be a bad idea since the CPU package with newer Intel CPU's is thinner so you're more likely to break it that way. You can pay places like Silicon Lottery to delid for you but it costs $50. It'll be interesting to see if delidding takes off but I strongly suspect only a few will really go through with it. Too many people terrified of bare dies or torquing on their CPU. edit 2: actually reading more of the delid OC threads with current Skylakes getting to 5Ghz or a bit above that is already fairly common. This might not be as big of deal as I thought. PC LOAD LETTER fucked around with this message at 10:38 on Dec 18, 2016 |
|
#
?
Dec 18, 2016 09:43
|
|
|
SwissArmyDruid posted:Funny thing about that, I actually think that Intel hosed up Kaby Lake's TIM again, and not because Kaby Lake is bad. He still can't get it stable at 5 GHz even after the delid: quote:The journey to rock solid 5Ghz continues. 1.392v is already high for a 14nm chip, and he can't really go higher. 1.4V wasn't recommended even for (32nm) Sandy Bridge, much less 22/14 nm.
|
|
#
?
Dec 18, 2016 12:02
|
|
|
Why can't intel make thermal paste to save themselves?
|
|
#
?
Dec 18, 2016 12:15
|
|
|
I would blow Dane Cook posted:Why can't intel make thermal paste to save themselves? The problem before (at least in the past) was not thermal paste, but the gap between the die and the heatspreader, caused by the black glue around the edge of the heatspreader. Someone did extensive work on testing different scenarios. The paste was OK, the gap was not. But it doesn't compare at all to when they soldered the drat things on. I think AMD should definitely do that with Zen, and advertise it as being a selling point.
|
|
#
?
Dec 18, 2016 12:30
|
|
|
HalloKitty posted:The problem before (at least in the past) was not thermal paste, but the gap between the die and the heatspreader, caused by the black glue around the edge of the heatspreader. They're doing it for Bristol Ridge, I'd expect them to do it for Zen as well.
|
|
#
?
Dec 18, 2016 12:42
|
|
|
sincx posted:He still can't get it stable at 5 GHz even after the delid: I feel that when a man must run Prime95 for SEVEN HOURS before they have a failure on a single LOGICAL core, that OC is basically stable.
|
|
#
?
Dec 18, 2016 12:49
|
|
|
SwissArmyDruid posted:I feel that when a man must run Prime95 for SEVEN HOURS before they have a failure on a single LOGICAL core, that OC is basically stable.
|
|
#
?
Dec 18, 2016 13:28
|
|
|
PC LOAD LETTER posted:For a gaming desktop you might be right but generally many consider that a OC has to be torture tested for 24 hours without errors or BSOD's before being called "stable". For me once I get close to 24 hours I call it good. I back up my stuff and I don't mind the occasional BSOD or reboot. Why though, is something like a 100mhz more really worth the BSODs? I cannot imagine it brings any significant performance increase over a little lower and stable OC. Also i had a an OCd 6600k that would pass prime for a day but would randomly lock up in WoW only, that was weird.
|
|
#
?
Dec 18, 2016 13:50
|
|
|
Prime probably hits other instructions than wow more
|
|
#
?
Dec 18, 2016 13:52
|
|
|
LooKMaN posted:Why though, is something like a 100mhz more really worth the BSODs? LooKMaN posted:Also i had a an OCd 6600k that would pass prime for a day but would randomly lock up in WoW only, that was weird.
|
|
#
?
Dec 18, 2016 15:28
|
|
|
Boiled Water posted:Prime probably hits other instructions than wow more Prime will only really test the arithmetic units so I'm not surprised the more complex stuff that a game will exercise aren't stable when over clocked
|
|
#
?
Dec 18, 2016 16:22
|
|
|
Yeah, the SIMD units seem to be the real killers here. Intel CPUs hit the heat ceiling pretty quickly when running a lot of AVX code and have to throttle themselves. If you're overclocking with a lot of voltage offset, no surprise it'll poo poo the bed eventually.
|
|
#
?
Dec 18, 2016 16:56
|
|
|
I would blow Dane Cook posted:Why can't intel make thermal paste to save themselves? because they dgaf about overclockers
|
|
#
?
Dec 18, 2016 18:57
|
|
|
FuturePastNow posted:because they dgaf about overclockers This is a good move.
|
|
#
?
Dec 18, 2016 19:08
|
|
|
I would blow Dane Cook posted:Why can't intel make thermal paste to save themselves? Because it's hard to get the heatspreader/die/TIM/glue interfaces to all be perfect. The glue lifting one edge of the heatspreader a tiny faction of a mm can gently caress up the thermal conductivity pretty badly. No glue dries perfectly without shrinkage or expansion, so you run into some issues no matter how good your TIm or process is. Solder+sealant doesn't have that issue, because you have a whetted bond between heatspreader and die before you fill in the cracks with a sealant to stiffen it up. Here's hoping we get more soldered chips, despite the extra cost to do so.
|
|
#
?
Dec 20, 2016 00:41
|
|
|
To temper the love for solder: it's known to form super tiny cracks after many heat cycles in some use cases. Which could be bad. At least, that's what Intel says. E: was just about to refer to Gwaihir's post in the Intel thread. drat have we been doing a lot of crossposting, to the point where I feel like people have been putting up rumors in the wrong thread. Sidesaddle Cavalry fucked around with this message at 04:05 on Dec 20, 2016 |
|
#
?
Dec 20, 2016 03:56
|
|
|
Crossposting for relevance!! http://overclocking.guide/the-truth-about-cpu-soldering/ TLDR, once CPU dies get small enough it breaks the solder after enough heat cycling. (But de-lidding/using liquid metal TIM performs the same as solder does) ((Unless you're an extreme OCer doing sub zero cooling))
|
|
#
?
Dec 20, 2016 03:57
|
|
|
Time to start a rumour that AMD's new chips will come embedded directly in heat pipe fluid channels for ultimate OCability. WCCFT would surely write it up.
|
|
#
?
Dec 20, 2016 23:03
|
|
|
Stock price: AMD (NASDAQ) US$11.50 +0.55 (+5.02%) Dec 20, 4:00 PM EST - Disclaimer 
|
|
#
?
Dec 20, 2016 23:06
|
|
|
ConanTheLibrarian posted:Time to start a rumour that AMD's new chips will come embedded directly in heat pipe fluid channels for ultimate OCability. WCCFT would surely write it up. You should write it and submit it.
|
|
#
?
Dec 21, 2016 00:13
|
|
|
Wistful of Dollars posted:You should write it and submit it. I'd be game to do this. Can someone provide convicting pictures of "internal" slides we could leak?
|
|
#
?
Dec 21, 2016 00:18
|
|
|
No Gravitas posted:I'd be game to do this. Can someone provide convicting pictures of "internal" slides we could leak? IBM has some nice illustrations of on-chip watercooling, it would be a shame if someone were to photoshop them into AMDs slide template 
|
|
#
?
Dec 21, 2016 00:34
|
|
|
No Gravitas posted:I'd be game to do this. Can someone provide convicting pictures of "internal" slides we could leak? Just type something up with Platfrom in it in big red letters and you should be good.
|
|
#
?
Dec 21, 2016 00:37
|
|
|
repiv posted:IBM has some nice illustrations of on-chip watercooling, it would be a shame if someone were to photoshop them into AMDs slide template I spiced it up a bit. https://imgur.com/a/ciyas
|
|
#
?
Dec 21, 2016 01:07
|
|
|
Please, do you actually think AMD has the money to do poo poo like that? They're more likely to cut channels in the heatspreaders and mount the rest of a waterblock on top of it. "It's like having a bare die, without having to delid it yourself!" And they can get away with it, because it's still using a ZIF socket, there's no piece of metal that has to go over to push the chip down into the LGA socket. C'mon guys, get your act together.
|
|
#
?
Dec 21, 2016 01:23
|
|
|
Risky Bisquick posted:Stock price: AMD (NASDAQ) US$11.50 +0.55 (+5.02%) Everyone, stop buying this stock.
|
|
#
?
Dec 21, 2016 02:17
|
|
|
No Gravitas posted:I spiced it up a bit. 
|
|
#
?
Dec 21, 2016 02:53
|
|
|
My favourite bit is the Chernobyl reference. https://en.wikipedia.org/wiki/Chernobyl_liquidators
|
|
#
?
Dec 21, 2016 11:35
|
|
|
No Gravitas posted:My favourite bit is the Chernobyl reference. I didn't know what that badge was, but it's even funnier now. I can actually see those dimwits running with it.
|
|
#
?
Dec 21, 2016 16:44
|
|
|
I did not realize that today is the 22nd. Holy crap, where has the time gone. Is January 17th still the expected launch date for Zen? I got my dates for CES and SHOT Show mixed up, that seems like critical delay between CES and launch, what with Kaby Lake dropping the 5th. SwissArmyDruid fucked around with this message at 14:39 on Dec 22, 2016 |
|
#
?
Dec 22, 2016 14:31
|
|
|
Q1 2017. I would guess closer to Computex than CES but we'll see.
|
|
#
?
Dec 22, 2016 15:00
|
|
|
|
| # ? May 16, 2024 04:20 |
|
|
I'm stoked to see what Zen and Vega are like, buuuuuut it won't be till fall/winter that I could actually afford to buy anything so.  Guess on the plus side I'll actually dodge the early adopter tax for once.
|
|
#
?
Dec 22, 2016 17:51
|
|