|
Alder Lake will be on 10nm
|
#
?
Mar 6, 2021 00:39
|
|
|
|
| # ? Apr 27, 2024 03:54 |
|
|
If this really is final performance, thats really terrible. The AVX-512 temperatures I don't really care about, because there really isn't any reason for gamers to care about AVX-512 at this point, but there are multiple games where it is straight up slower than a 10700k. Intel pushed the launch back 2 weeks supposedly for BIOS reasons, and there are a lot of really bizarre results. You gotta think something is busted in the firmware. If not, Rocket Lake will make Broadwell look like a smooth and normal launch in comparison. Guess I should start looking for a 5950X again.
|
|
#
?
Mar 6, 2021 00:52
|
|
|
The BIOS reason is that CPU temps melted the spi
|
|
#
?
Mar 6, 2021 00:57
|
|
|
Looks like Intel is having a bit of an FX-9590 moment
|
|
#
?
Mar 6, 2021 01:05
|
|
|
Alder Lake is supposed to be 10nm
|
|
#
?
Mar 6, 2021 01:05
|
|
|
Why the gently caress are they cutting away L3 cache?! EDIT: L3 latency is apparently also higher. BlankSystemDaemon fucked around with this message at 01:23 on Mar 6, 2021 |
|
#
?
Mar 6, 2021 01:11
|
|
|
Intel learned absolutely nothing from the 5775C. Their ace in the hole should be gigantic caches, but they're still making Skylakes.
|
|
#
?
Mar 6, 2021 01:55
|
|
|
I really really hope the final numbers are significantly better. Otherwise I'll be grabbing an zen3 if its ever in stock, and hoping all the weird compatibility issues are worked out.BlankSystemDaemon posted:Why the gently caress are they cutting away L3 cache?! 2MB cache / core, 2 cores less -> 4MB cache less?
|
|
#
?
Mar 6, 2021 02:49
|
|
|
BIG HEADLINE posted:Intel learned absolutely nothing from the 5775C. Rocket Lake isn't Skylake... the caches are actually worse than Skylake in the Cove cores with latency specifically, but we already kinda knew that from Tiger Lake: https://www.anandtech.com/show/16084/intel-tiger-lake-review-deep-dive-core-11th-gen/4 I'm just surprised that they are even bothering to release this part at all if those numbers are final.
|
|
#
?
Mar 6, 2021 03:44
|
|
|
Feeling pretty good about getting Comet Lake instead of holding out for Rocket Lake Seems like they had a... failure to launch
|
|
#
?
Mar 6, 2021 03:47
|
|
|
Holy poo poo, I'm glad I have Zen and not crappy Bins
|
|
#
?
Mar 6, 2021 03:53
|
|
|
gradenko_2000 posted:Feeling pretty good about getting Comet Lake instead of holding out for Rocket Lake You still got a bad space heater CPU in the end.
|
|
#
?
Mar 6, 2021 04:18
|
|
|
wtf, intel?
|
|
#
?
Mar 6, 2021 05:18
|
|
|
I'm thinking the sky-high thermals are also why they ~mysteriously~ ended their questionably-worthwhile overclocking insurance program: https://www.tomshardware.com/news/intel-kills-off-performance-tuning-protection-plan-overclock-warranty
|
|
#
?
Mar 6, 2021 05:22
|
|
|
Hats off to the poor engineers who spent months sitting in a lab next to some heat sink fan blasting away at 20k rpm
|
|
#
?
Mar 6, 2021 05:26
|
|
|
Well, Alder Lake will hopefully be generally available before the end of the year.
|
|
#
?
Mar 6, 2021 05:29
|
|
|
Inept posted:Alder Lake is 10nm and supposed to be out later this year. We'll see. In before 10nm doesn't clock anywhere near as well as 14nm+++++++ and it ends up being a performance regression
|
|
#
?
Mar 6, 2021 05:31
|
|
|
Fame Douglas posted:You still got a bad space heater CPU in the end. It's only the i5-10400F, so not that warm really
|
|
#
?
Mar 6, 2021 06:20
|
|
|
repiv posted:In before 10nm doesn't clock anywhere near as well as 14nm+++++++ and it ends up being a performance regression i'm with ya. just because they're going to finally hit 10nm doesn't mean that sucker's gonna be any good.
|
|
#
?
Mar 6, 2021 06:53
|
|
|
even for Intel this is shockingly bad, yikes
|
|
#
?
Mar 6, 2021 07:09
|
|
|
B-1.1.7 Bomber posted:i'm with ya. just because they're going to finally hit 10nm doesn't mean that sucker's gonna be any good. I know intel is working on 7nm but mannnnnnnnnnnn the fab side of intel is failing so hard.
|
|
#
?
Mar 6, 2021 07:35
|
|
|
This feels like a Pentium 4 moment...
|
|
#
?
Mar 6, 2021 07:50
|
|
|
The only possible silver lining would be making the subsequent i5-F a better value that it currently is, and even that is highly suspect IMO.
|
|
#
?
Mar 6, 2021 07:57
|
|
|
TheDemon posted:This feels like a Pentium 4 moment... 10Ghz by 2005 at <1v! No, seriously: https://www.anandtech.com/show/680/6
|
|
#
?
Mar 6, 2021 08:00
|
|
|
Lol this sucks
|
|
#
?
Mar 6, 2021 09:43
|
|
|
a plus too many added to 14nm. End of the road. Unless you want avx512, it improves nothing except its ability to test vrms and heat a room. No wonder they didn't want to add more cores The guy earlier who mentioned the 5775c was right, we saw what big caches could do, but then Intel did nothing with the idea. Now we have less cache and higher latencies. Great job
|
|
#
?
Mar 6, 2021 10:03
|
|
|
Ika posted:2MB cache / core, 2 cores less -> 4MB cache less? One way this is noticeable to the end-user is if you're working in a high-level language like C++, and if you're designing large-scale code - for example video games where you're effectively dealing with a soft real-time requirement - you need to hand-optimize the code so that it keeps what it's working on in L2, and and retains the ability to keep as much overflow as possible in L3. Going L1 to L2 is less than an order of magnitude, and going from L2 to L3 is about an order of magnitude, whereas hitting DRAM is several orders of magnitude slower. Data-oriented design was publically described by Mike Acton was around 2014, so it's a sure bet that it's being used by engines now, especially because he now works at Unity. Here's the video in case anyone's interested: https://www.youtube.com/watch?v=rX0ItVEVjHc
|
|
#
?
Mar 6, 2021 11:28
|
|
|
BlankSystemDaemon posted:That's not really how it works, because L3 is shared. There's a web site! https://www.dataorienteddesign.com/site.php
|
|
#
?
Mar 6, 2021 12:41
|
|
|
I'm so glad I just went with a bargain binned 9900k mid last year instead of waiting around for whatever was gonna come next. It's not going to be another 2600k but it feels like some kind of high point lately, and for gaming it apparently wont be lacking for a looong while yet.
|
|
#
?
Mar 6, 2021 13:52
|
|
|
Inept posted:The heatsink slayer has arrived Peaking around 290W and 104C for AVX512?!? Has there been any desktop retail CPUs even close to that kind of peak power? Maybe that ridiculous AMD FX-9590 did it, but that's still more power draw than my GPU  At least gaming and other normal workloads require a lot less power than that, but those are still not impressive in performance or power draw.
|
|
#
?
Mar 6, 2021 15:39
|
|
|
Bofast posted:Peaking around 290W and 104C for AVX512?!? Has there been any desktop retail CPUs even close to that kind of peak power? Maybe that ridiculous AMD FX-9590 did it, but that's still more power draw than my GPU An FX-9590 could draw up to 370w when OC'd to 4.6Ghz all-core (at stock settings it could draw about 230w)
|
|
#
?
Mar 6, 2021 16:31
|
|
|
Heat spreaders are becoming a catch-22 on modern CPUs. They are necessary because dies are fragile and they also provide a thermal buffer to save CPUs from instantly burning up from improperly installed coolers, but they also are basically insulators at this point and prevent any cooling system from being able to remove heat from the die as fast as it can be produced.
|
|
#
?
Mar 6, 2021 16:52
|
|
|
BlankSystemDaemon posted:That's not really how it works, because L3 is shared. I know its shared, but a lot of the SKUs have L3 cache scaling with number of cores. Possibly having to do with die area allocation. Another thought: with regression to 14nm cache takes up more die and is easier to cut than CPU logic.
|
|
#
?
Mar 6, 2021 17:09
|
|
|
Indiana_Krom posted:Heat spreaders are becoming a catch-22 on modern CPUs. They are necessary because dies are fragile and they also provide a thermal buffer to save CPUs from instantly burning up from improperly installed coolers, but they also are basically insulators at this point and prevent any cooling system from being able to remove heat from the die as fast as it can be produced. I remember thinking it might be a good idea to buy a shim and go bare-die on my 9900K using Kryonaut instead of Conductonaut (since I still don't trust liquid metal), but then I started reading reviews and watching YouTube videos of people trying just that and getting shittier thermals as a result. Even though Intel's indium solder isn't the greatest in the world, it's better than any current paste, and it does seem like the IHS performs a ~kind~ of service by spreading out the heat across a larger surface area when transferring it to the cold plate. Bare-die cooling modern Intel CPUs only works better than stock when using liquid metal.
|
|
#
?
Mar 6, 2021 17:10
|
|
|
HUB has a video about the AT review up and they mention that they don’t expect performance to increase by launch date. Frankly can’t see why anyone wouldn’t just get the $350 10850k over the Rocket Lake parts if that’s true.
|
|
#
?
Mar 6, 2021 18:44
|
|
|
(Model) Number goes up
|
|
#
?
Mar 6, 2021 18:48
|
|
|
Cygni posted:HUB has a video about the AT review up and they mention that they don’t expect performance to increase by launch date. Frankly can’t see why anyone wouldn’t just get the $350 10850k over the Rocket Lake parts if that’s true. AMD doesn't have the capacity to cover the entire desktop market, let alone the bigger laptop and server markets. Intel is going to stop making 10th gen chips, so once old stock falls out of the distribution pipelines, it's going to be 11th gen or AMD. At least they're going to be cheap, for i5 and i7.
|
|
#
?
Mar 6, 2021 18:52
|
|
|
Intel is just doing their part to reduce demand on semiconductors and the associated shortage by putting out a crap CPU so everyone says “eh I’ll wait for the next one”. So smart!
|
|
#
?
Mar 6, 2021 19:01
|
|
|
priznat posted:Intel is just doing their part to reduce demand on semiconductors and the associated shortage by putting out a crap CPU so everyone says “eh I’ll wait for the next one”. So smart! Intel has managed to make me say that for the last 6 years.
|
|
#
?
Mar 6, 2021 20:02
|
|
|
|
| # ? Apr 27, 2024 03:54 |
|
|
gradenko_2000 posted:An FX-9590 could draw up to 370w when OC'd to 4.6Ghz all-core (at stock settings it could draw about 230w) Well, yeah. I should have been more clear on talking specifically about stock settings.
|
|
#
?
Mar 6, 2021 20:11
|
|