|
Arzachel posted:I don't think Ryzen hexa and octa cores are worth it for gaming until they figure out how to keep critical tasks from migrating between the clusters. You can do that manually with a programme like Process Lasso. If what SwissArmyDruid says is true, though, that the 6-cores will have one core disabled from each CCX, that's a disappointment. 2 enabled on one, 4 on the other would be perfect, because you could assign Windows processes to 2 on the same die, and then have all your other tasks running free on one quad core CCX, avoiding that problem of migrating between them. HalloKitty fucked around with this message at 13:43 on Mar 15, 2017 |
# ? Mar 15, 2017 12:33 |
|
|
# ? May 22, 2024 00:55 |
|
https://www.io-tech.fi/uutinen/6-ytiminen-ryzen-prosessori-teknisesti-mahdollista-toteuttaa/ That's the layout these guys said their sources said AMD was looking at. quote:We have confirmed from our sources that AMD’s upcoming Ryzen processor can be technically configured as 6-core model. It is possible to disable each CPU core separately together with dedicated L2 cache from CCX without affecting shared L3 cache. Possible L3 cache configurations are 1/1 (8 MB), 1/2 (4 MB) or completely disabled. Basic rule is that both CCXs should have similar CPU core and L3 configuration for example both should have 3 cores and full 8 MB L3 cache enabled (6C & 16 MB L3). Which means that 4-core Ryzens are likely to be 2 + 2, and not "all on one CCX" parts. SwissArmyDruid fucked around with this message at 12:42 on Mar 15, 2017 |
# ? Mar 15, 2017 12:38 |
SwissArmyDruid posted:http://fudzilla.com/news/processors/43118-samsung-ramping-up-to-7nm-next-year ha, that's exactly what I thought the next step would be
|
|
# ? Mar 15, 2017 13:04 |
|
Is the separate l2 cache going to cause problems with more highly threaded games, going forwards? I don't really see how it can't.
|
# ? Mar 15, 2017 20:23 |
|
Depends on how those threads are scheduled. The grim future is one in which game developers hard core affinity.
|
# ? Mar 15, 2017 20:24 |
|
SwissArmyDruid posted:
Obviously. They are gonna die harvest the 2 ccx non apu core to make these
|
# ? Mar 15, 2017 20:33 |
|
Why is the 1600 a 95W part and the 1700 65W? 'lol AMD TDPs' notwithstanding.
|
# ? Mar 15, 2017 20:34 |
|
Malcolm XML posted:Obviously. They are gonna die harvest the 2 ccx non apu core to make these They could die harvest four good cores all in the same CCX, though. Eliminates switching problems, at the expense of smaller total cache.
|
# ? Mar 15, 2017 20:35 |
|
dont be mean to me posted:Why is the 1600 a 95W part and the 1700 65W? 'lol AMD TDPs' notwithstanding. 1600X is a salvaged high leak binned 8 core, with two cores not working or disabled. It is clocked at the same frequency as the 1800X. 1700 is a low leak binned 8 core, that is clocked much closer to the most efficient state for the architecture (which is a high IPC low clock/TDP design pushed to all market segments).
|
# ? Mar 15, 2017 21:12 |
|
Watermelon Daiquiri posted:ha, that's exactly what I thought the next step would be Yeah but lol at this quote: quote:Samsung will also use extreme ultraviolet (EUV) lithography once it achieves a breakthrough in the technology, Kang added.
|
# ? Mar 15, 2017 23:03 |
ConanTheLibrarian posted:Yeah but lol at this quote: :facepalm: oh wow
|
|
# ? Mar 16, 2017 00:13 |
|
https://www.pcper.com/news/Processors/Ryzen-Locking-Certain-FMA3-Workloads Saved you a click: Ryzen hardlocking on certain FMA3 workloads, a BIOS fix is on the way.
|
# ? Mar 16, 2017 00:41 |
|
SwissArmyDruid posted:https://www.pcper.com/news/Processors/Ryzen-Locking-Certain-FMA3-Workloads quote:This may come at a cost through lower burst speeds to keep TDP within their stated envelope. LOL good one.
|
# ? Mar 16, 2017 00:49 |
|
SwissArmyDruid posted:https://www.pcper.com/news/Processors/Ryzen-Locking-Certain-FMA3-Workloads Phenom III: TLB harder
|
# ? Mar 16, 2017 02:48 |
|
AMD CPU and Platform Discussion: BIOS BIAS
|
# ? Mar 16, 2017 04:19 |
|
Platfrom It's platfrom god dammit
|
# ? Mar 16, 2017 04:27 |
|
dont be mean to me posted:Why is the 1600 a 95W part and the 1700 65W? 'lol AMD TDPs' notwithstanding. The voltage to break 3.7 stable is pretty large. I go from 3.7 @ 1.1875 vs 3.8 @1.25 https://docs.google.com/spreadsheets/d/1wqHz5J4S7tHM5CJKB1zyh8qI52pTEywKws9PcVro2cc/edit http://valid.x86.fr/1410hw http://valid.x86.fr/nggw3f Risky Bisquick fucked around with this message at 04:36 on Mar 16, 2017 |
# ? Mar 16, 2017 04:31 |
|
I was excited about possibly getting a R5-1600X, but reading above, it sounds like it might be a crippled chip?
|
# ? Mar 16, 2017 07:03 |
SourKraut posted:I was excited about possibly getting a R5-1600X, but reading above, it sounds like it might be a crippled chip? Not really, it's just people were hoping for a chip with a 4/2 CCX split instead of a 3/3 one, this is because passing threads between the CCXes is much slower than within CCXes. If you are doing workloads that will benefit from the extra threads the chip should still be very good at that and while still behind Intel's offerings in gaming it should be a decent chip for that too. Ryzen is really a nice chip for mixed workloads, but if you are just gaming Intel is still going to be the better choice for now.
|
|
# ? Mar 16, 2017 07:24 |
|
16MB L3, 6c, 65W for $220? Jesus, sign me the gently caress up. I'm digging how the r5 line is mostly keeping the fuckoff-huge l3 cache.
|
# ? Mar 16, 2017 14:12 |
|
Potato Salad posted:16MB L3, 6c, 65W for $220? Yeah, this parts are much more competitive, and like you said, loving the huge cache. That said, they still need actual clock increases over the 8-core parts, but it seems like their process just won't allow them. Hm. Once they get these 6-core parts to 4.5GHz or so, there's pretty much nothing holding them back. vv It is, I'm just saying that a lot of people will still shy away for gaming reasons because the single-thread performance still isn't up there due to the slightly low clocks. HalloKitty fucked around with this message at 15:01 on Mar 16, 2017 |
# ? Mar 16, 2017 14:28 |
|
Goflo iteration may get us there, who knows. I can only look at the present, and that poo poo is sexy
|
# ? Mar 16, 2017 14:55 |
|
Yeah, the later stepping of these same dies should help them find a couple hundred more MHz. I wouldn't expect anything game-changing but for launch silicon this is pretty great, especially at 65W.
|
# ? Mar 16, 2017 15:20 |
|
HalloKitty posted:That said, they still need actual clock increases over the 8-core parts, but it seems like their process just won't allow them. Called it
|
# ? Mar 16, 2017 15:23 |
|
Truga posted:Called it Who cares, pricing is great, 12 threads at at least 3GHz for $219.
|
# ? Mar 16, 2017 15:24 |
|
Yeah, it's absolutely great, I was expecting 6 threads much closer to $300.
|
# ? Mar 16, 2017 15:27 |
|
Don't care about a 4 GHz wall with mobile parts and to a lesser extent, servers, but I worry the APUs will bring last-gen perf to throw down with Cannonlake, when ideally it should be Zen+ or Zen 2 or whatever the hell.
|
# ? Mar 16, 2017 15:31 |
|
SwissArmyDruid posted:Don't care about a 4 GHz wall with mobile parts That's very true, now we just need some OEMs to make AMD laptops that aren't absolute poo poo in every other respect.
|
# ? Mar 16, 2017 15:41 |
|
Doesn't Zen L2 cache always run at full clock speed? I wonder if that is part of the reason why it won't go far past 4.0-4.1 Ghz. AFAIK Kaby Lake L2 cache frequency (uncore) also tops out around there but it's on an independent multiplier.
|
# ? Mar 16, 2017 15:43 |
|
HalloKitty posted:That's very true, now we just need some OEMs to make AMD laptops that aren't absolute poo poo in every other respect. If ryzen is any good with mobile parts they probably will. The only people buying AMD laptops currently are idiots who can't evaluate what is good/bad for a computer so naturally you're going to make a poo poo laptop with the current poo poo processors. You're fleecing them already by selling them a bulldozer laptop, you might as well fleece them on every other part. If it is an actually good processor, its worth targeting people who know what makes a laptop good when using the processor.
|
# ? Mar 16, 2017 15:49 |
|
HalloKitty posted:That's very true, now we just need some OEMs to make AMD laptops that aren't absolute poo poo in every other respect. Those impressive 30-35W Cinebench results The Stilt posted have me optimistic.
|
# ? Mar 16, 2017 16:09 |
|
Aren't AMD laptops, if you can get one with the best APU offering AMD offers, better GPU wise then say a midrange Intel with a 620 or whatever the low end iGPU is for those? Sure the CPU is going to suck, but if you just wanted a basic media consumption device, is AMD really that bad currently in the laptop realm if you aren't buying a i5+? *Have not done any research because all I have been looking at are 1070-1080GPU powered laptops for replacement of my old i7-720QM-5870M beast...
|
# ? Mar 16, 2017 17:10 |
|
I'm talking about build quality, tiny batteries, poor low res screens, etc. Carrizo is an already existing reasonable AMD solution, but try finding anything that has a build that's on par with the best Intel laptops..
|
# ? Mar 16, 2017 17:14 |
|
AMD is absolutely better for basic media consumption and browsing or whatnot at same price points. The problem is, the kind of laptop OEMs put AMDs in will fall apart after 6-12 months. e;fb
|
# ? Mar 16, 2017 17:14 |
|
HalloKitty posted:Yeah, this parts are much more competitive, and like you said, loving the huge cache. That said, they still need actual clock increases over the 8-core parts, but it seems like their process just won't allow them. Hm. Once they get these 6-core parts to 4.5GHz or so, there's pretty much nothing holding them back. If I'm not doing serious gaming but occasionally playing games, I'd imagine it's still fine? Just not "THE BEST IT COULD BE PLAYED"?
|
# ? Mar 16, 2017 17:16 |
|
So for most people except those trying to squeeze out every single last FPS, the 1600 is going to become the best value CPU right? With cooler included it's a pretty solid proposition.
|
# ? Mar 16, 2017 17:32 |
|
They need to fix memory compatibility with their controller. Can they patch this with microcode or a bios update? As is it will create a problem when the more numerous R5 users can't get over 2400 with their 3000+ dual rank kits they already have.
|
# ? Mar 16, 2017 17:35 |
|
Truga posted:AMD is absolutely better for basic media consumption and browsing or whatnot at same price points. You are right about a lot of crap laptops, though if you sift through enough of them, there are at least a few that have 1080P screens and aren't made of the cheapest plastic you might find on a low end Acer or something. Asus made one... Still pretty sad but bang for the buck its not terrible. Though hopefully if mobile Zen becomes a thing then the OEM's might take notice. Only time will tell currently as the iron out all the bugs and chips get made.
|
# ? Mar 16, 2017 17:42 |
|
EdEddnEddy posted:You are right about a lot of crap laptops, though if you sift through enough of them, there are at least a few that have 1080P screens and aren't made of the cheapest plastic you might find on a low end Acer or something. I hope someone sticks an 8 core chip in a laptop. Would be nice to have a semi-portable workstation, even if it gets as bulky as some of those gaming laptops with dual 1080s inside.
|
# ? Mar 16, 2017 17:55 |
|
|
# ? May 22, 2024 00:55 |
lol still no gaming benchmarks on anandtech 2 weeks out. Sure hope someone at least got a nice beach house from that agreement!
|
|
# ? Mar 16, 2017 18:03 |