|
Shipon posted:If only AMD would get rid of its PGA packaging and move to LGA like they should, lol at the idea of ripping the CPU out of the socket with your heatsink. this is like, actually impossible to do if one uses the most bare thought or care though, who on earth avoids a platform on the basis of "I'm an idiot and will simply yank on a thing instead of twisting it first"
|
# ? Mar 9, 2021 19:17 |
|
|
# ? Apr 27, 2024 14:48 |
|
Ihmemies posted:Probably just wiser to stick with what you have and buy a DDR5 platform next year. Supposing whatever you have is not too bad, like my 8700K. 3470 lol. I know the general advice is to buy when needed, but this is how I got here. It's still just fine. A bit choppy editing 4k drone footage but that's very occasional. And without a new GPU it's not even much of a bottleneck in games.
|
# ? Mar 9, 2021 19:41 |
|
I need another sanity check, especially after the dismal 11th gen preview. Is an i9 10850K for $320 a good idea from an 8086K?
|
# ? Mar 9, 2021 20:30 |
|
That's a good price for an 10850k to be sure, but I don't think you would feel much if any significant an upgrade from an 8086k unless you're doing exclusively heavily multi-threaded work. For gaming you'd likely not be able to tell and difference at all.
|
# ? Mar 9, 2021 20:36 |
|
Also you’re gonna need a new motherboard.
|
# ? Mar 9, 2021 20:45 |
|
I paid 300€ for 6 gigs of ddr3 in 2008. 16 gigs of ddr4 was 300€ too in 2017. So if I can get 32 gigs of ddr5 for 300€ in 2022 that is ok. If the price is 500-600€ well then I would have to think about it.
|
# ? Mar 9, 2021 20:54 |
|
Ihmemies posted:I paid 300€ for 6 gigs of ddr3 in 2008. 16 gigs of ddr4 was 300€ too in 2017. So if I can get 32 gigs of ddr5 for 300€ in 2022 that is ok. If the price is 500-600€ well then I would have to think about it. Ha, I I won a Phenom II 965 Black on release day at QuakeCon 2009, and after seeing DDR3 prices at the time I decided to take advantage of the fact that it could work with DDR2 or DDR3, and bought 8GB of DDR2-800 for under $80 that same weekend. Turns out I dodged a bullet, because that same CPU could barely drive 4 sticks of DDR3 and had to run DDR3 at a super low memory clock.
|
# ? Mar 9, 2021 21:18 |
|
Some dude over at Computerbase forum feels that an oc 11990K will beat the 5950x by 20% in gaming in a best case scenario. Another claims his 11700K is 0-10% faster than his 10900K with recent BIOS and a Z590 board and it would match a 5900x. We’ll see.
|
# ? Mar 10, 2021 00:44 |
|
Cygni posted:I probably already would have if the 5950X was in stock tbh Each week there have been several times that I could buy a 5950X with minimal effort. I could probably grab you one within a few days if you really want one.
|
# ? Mar 10, 2021 01:32 |
|
Mr.PayDay posted:Some dude over at Computerbase forum feels that an oc 11990K will beat the 5950x by 20% in gaming in a best case scenario. If that were true Intel would have showed that, not this.
|
# ? Mar 10, 2021 01:39 |
|
SourKraut posted:Each week there have been several times that I could buy a 5950X with minimal effort. I could probably grab you one within a few days if you really want one. how dare you call my bluff but for real im mostly just griping cause im a bored manchild that wants a shiny new toy, i dont really need anything new to run Loop Hero.
|
# ? Mar 10, 2021 01:47 |
|
Ugly In The Morning posted:Also you’re gonna need a new motherboard. Gwaihir posted:That's a good price for an 10850k to be sure, but I don't think you would feel much if any significant an upgrade from an 8086k unless you're doing exclusively heavily multi-threaded work. For gaming you'd likely not be able to tell and difference at all. It's strictly for gaming, and occasional streaming. I'm aware it wouldn't be a big upgrade, but if eBay is anything to go by I can flip the CPU for ~$290 and the MB for ~$130
|
# ? Mar 10, 2021 01:47 |
|
Eletriarnation posted:There are advantages to PGA, too. While it's a pain to unbend their pins, it's at least feasible - good luck unfucking an LGA which has taken any damage at all. When LGA came out everyone was scared shitless of bricking a mobo but I have never actually done it, including at least one time moving a mobo around without the little plastic covering in the CPU holder and another time where one of the screws holding the retention mechanism came all the way out so the CPU wasn't in properly.
|
# ? Mar 10, 2021 04:24 |
|
LGA means you don't get to know the fear of straightening out a pin with a mechanical pencil
|
# ? Mar 10, 2021 05:17 |
|
back during the Athlon days I got used to buying open box motherboards off newegg. Never had a problem with it. The first time I tried that with an Intel board when I jumped back in during the Conroe days had a shitload of bent pins and I couldn't get it to post.
|
# ? Mar 10, 2021 06:09 |
|
Gwaihir posted:this is like, actually impossible to do if one uses the most bare thought or care though, who on earth avoids a platform on the basis of "I'm an idiot and will simply yank on a thing instead of twisting it first" twisting a heatsink off sounds like a good way to accidentally scratch a trace on the board
|
# ? Mar 10, 2021 06:45 |
|
Shipon posted:twisting a heatsink off sounds like a good way to accidentally scratch a trace on the board The board that has an entire CPU and socket between it and the cooler? It's just a little twist to break the paste, you're not supposed to rock it back and forth.
|
# ? Mar 10, 2021 08:23 |
|
Shipon posted:twisting a heatsink off sounds like a good way to accidentally scratch a trace on the board They're talking about unscrewing / unclipping the heatsink first, so paste is the only thing retaining the heatsink.
|
# ? Mar 10, 2021 08:48 |
|
You're also supposed to warm up the CPU a little first, so the paste isn't rock solid. Honestly, these are all tips that anyone should do, people just got sloppy because LGA sockets took over.
|
# ? Mar 10, 2021 09:01 |
|
MaxxBot posted:If that were true Intel would have showed that, not this. If that's optimistic, that's underwhelming to say the least; since the power numbers seem to be way up on the 5900X, along with the fact it has 50% fewer cores, thus will eat dirt in everything else. This isn't looking good for ol' Chipzilla. Their only hope is pricing. If it's cheap and widely available, then they'll sell a bunch
|
# ? Mar 10, 2021 09:40 |
|
widely available is almost certainly going to be true - Intel is remaining competitive right now just because they have far more capacity to produce their 14nm parts, as compared to AMDs CPUs, GPUs, and console SOCs all being tied up with TSMC they could probably even price their six-core i5's as much they do the current Comet Lake i5, or even slightly higher, and they'd still sell well just because only the Ryzen 3600 is competing in that space, and that part is itself getting squeezed out since it's also on 7nm and then of course the quad-cores and dual-cores aren't even Rocket Lake, they're just Comet Lake refreshes, since Intel knows AMD isn't even pretending to have anything to compete with i3's and Pentiums that said, if the Rocket Lake i7 isn't anywhere close to the current i7-10700 in pricing, and if the Rocket Lake i9 isn't straight-up cheaper than the i9-10850K, I can't imagine that it's going to be compelling at all
|
# ? Mar 10, 2021 09:58 |
|
Endymion FRS MK1 posted:It's strictly for gaming, and occasional streaming. I'm aware it wouldn't be a big upgrade, but if eBay is anything to go by I can flip the CPU for ~$290 and the MB for ~$130
|
# ? Mar 10, 2021 12:35 |
|
I saw this tweet this morning: https://twitter.com/9550pro/status/1369442891198763011 And specifically this section: Am I reading this right that Intel is locking memory speeds such that only the i9-11900K can run 3200 MHz, and everything else has to run slower?
|
# ? Mar 11, 2021 05:48 |
|
gradenko_2000 posted:I saw this tweet this morning: You can manually set it to 1:1 memory ratio, it's just set by default to 1:2 at 3200 JEDEC speeds on the 11700K and below. A weird choice, but nothing you can't work around. At 2933 it runs at 1:1, or when using XMP It's actually why the Anandtech review had such awful performance and memory latency, it was set to 3200 1:2 which is markedly worse than just 2933 1:1. I'm waiting on a review with a decent kit of ram properly configured , it should make a significant difference in a lot of benchmarks
|
# ? Mar 11, 2021 06:02 |
|
BurritoJustice posted:You can manually set it to 1:1 memory ratio what is a memory ratio? Is that like how you want the Ryzen Infinity clock to match your memory clock?
|
# ? Mar 11, 2021 06:05 |
|
gradenko_2000 posted:Am I reading this right that Intel is locking memory speeds such that only the i9-11900K can run 3200 MHz, and everything else has to run slower? The footnote text says all SKUs support DDR4-3200, but only the i9 supports it in "Gear 1" while the rest have to drop back to "Gear 2". I googled and found some Intel 10th gen docs indicating that Gear-1 and Gear-2 refer to the clock speed ratio between the memory controller (MC) and DRAM. Gear 1 means the MC is 1:1 with memory, Gear 2 means it's 1:2, half speed. The Intel docs are less than perfectly clear, but I think this gear ratio only affects the state machine which generates DDR4 commands. If so, the performance penalty of Gear 2 should be much less dramatic than a factor of 2. The reason why has to do with the details of how the system tells DDR4 to do things. In an x86 system, DDR4 accesses should always be burst length 8 (BL8) since 8 transfers on a 64-bit datapath is 512 bits, or 64 bytes, the size of an x86 cache line. Each BL8 needs up to two commands sent from the MC to DRAM: the first to "open" a row, the second to perform a BL8 read or write within that row. I say "up to" because the row open command is optional; when you access the same row multiple times without closing it, you don't have to re-open it every time. (Don't want to, either, because it's a major performance win to avoid re-opening rows.) Commands are transferred at half the rate of data since they use only the rising edge of the clock. Four rising edges occur during the amount of time needed for one BL8 transfer, so the max command rate supported by DDR4 is twice as fast as required to keep the data pins 100% busy. So, if Gear 2 mode cuts the Intel MC's command generation rate to half of normal, that's still enough to saturate the DDR4 data pins with BL8 traffic. In practice there will be some penalties, but nothing like a factor of 2.
|
# ? Mar 11, 2021 07:45 |
|
In practise it seems like a crude way of making the 11900K seem much better compared to the 11700K out of the box, given that if you bench them with stock settings there will be more of a measurable difference than there ought to be I still maintain it was a bit of a whiff by Anandtech to not mention or test this in their review, their numbers basically aren't relevant to the actual end user. People have been happily running them at 1:1 at 3600+ with no additional effort
|
# ? Mar 11, 2021 11:43 |
|
Will the non -K processors allow that setting to be changed?
|
# ? Mar 11, 2021 14:26 |
|
I can't find if they had 1:1 or 1:2 for this review but hardwareluxx did a 11700k review with less strange results, still mediocre but better gaming performance than anandtech showed. https://twitter.com/3DCenter_org/status/1369929335226863628?s=20 https://www.hardwareluxx.de/index.p...vorab-test.html
|
# ? Mar 11, 2021 15:56 |
|
MaxxBot posted:I can't find if they had 1:1 or 1:2 for this review but hardwareluxx did a 11700k review with less strange results, still mediocre but better gaming performance than anandtech showed. I'm guessing there are some typos here - these results make even less sense!
|
# ? Mar 12, 2021 00:10 |
|
BurritoJustice posted:In practise it seems like a crude way of making the 11900K seem much better compared to the 11700K out of the box, given that if you bench them with stock settings there will be more of a measurable difference than there ought to be Anandtech uses jedec memory for all their testing. I agree that that makes low res gaming results not very relevant for people who would care about them but they're up front about it.
|
# ? Mar 12, 2021 10:58 |
|
forbidden dialectics posted:I'm guessing there are some typos here - these results make even less sense! how do you figure? looking at the games: * the 10th-gen 8-core is at 100% * the 11th-gen 8-core is at 106% * the AMD 8-core is at a dead heat with the 11th-gen 8-core, which is what we kinda sorta expected from Rocket Lake * the 10th-gen 10-core is at 107%, which is either statistical noise or simply a function of this chip clocking higher than the other CPUs looking at the apps: * the 10th-gen 8-core is at 100% * the 11th-gen 8-core is at 111% (and disregarding the old BIOS) * the AMD 8-core is at 120%, which is not surprising since Intel has been gunning for the gaming crown rather than the productivity crown * the 10th-gen 10-core is at 121%, which is not surprising since this has 2 cores more than the rest of the competition I mean it might still be disappointing for 11th gen to only be an incremental upgrade over 10th gen, despite already being a different design instead of warmed-over Skylake, but an incremental upgrade is already what Intel was generally promising, and isn't the outright regression that was seen with Anandtech
|
# ? Mar 12, 2021 11:45 |
|
Arzachel posted:Anandtech uses jedec memory for all their testing. I agree that that makes low res gaming results not very relevant for people who would care about them but they're up front about it. They could've used 2933 JEDEC however, as it's the max speed out of the box to use memory gear 1. It's much more impactful than the standard loss from 3200 JEDEC being like c20 or whatever
|
# ? Mar 12, 2021 13:07 |
|
This made me chuckle https://twitter.com/PianoREDACTED/status/1369864715523026944
|
# ? Mar 12, 2021 20:56 |
|
We need more comparisons between “top of the line new setup” and “top of the line setup from three-ish years ago”. I feel that a lot of people wanting to do new builds now are upgrading from like, an 8700K+1080Ti setup and are looking for a CPU to pair with a hypothetical 3080.
|
# ? Mar 12, 2021 21:04 |
|
gradenko_2000 posted:how do you figure? Ah, you're right, I just completely forgot that the 10900k has 10 cores!
|
# ? Mar 12, 2021 21:21 |
|
Kazinsal posted:We need more comparisons between “top of the line new setup” and “top of the line setup from three-ish years ago”. I feel that a lot of people wanting to do new builds now are upgrading from like, an 8700K+1080Ti setup and are looking for a CPU to pair with a hypothetical 3080. Yeah, ive been thinkin about this a lot lately. Also midrange never get mentions and you have to kinda look at the max end CPUs on the chart and guess where your midrange part would be. Like the mid tier parts regularly outsell the ballin parts by like 5-1, and you never see a i5 6500 or R5 1600 on charts.
|
# ? Mar 12, 2021 21:33 |
|
Cygni posted:Yeah, ive been thinkin about this a lot lately. https://www.anandtech.com/show/14043/upgrading-from-an-intel-core-i7-2600k-testing-sandy-bridge-in-2019 This was one of my favorite recent AnandTech articles -- probably another version but with... 6700Ks?
|
# ? Mar 12, 2021 21:37 |
|
Kazinsal posted:We need more comparisons between “top of the line new setup” and “top of the line setup from three-ish years ago”. I feel that a lot of people wanting to do new builds now are upgrading from like, an 8700K+1080Ti setup and are looking for a CPU to pair with a hypothetical 3080. https://www.anandtech.com/bench/CPU-2020/2758
|
# ? Mar 12, 2021 22:02 |
|
|
# ? Apr 27, 2024 14:48 |
|
Wow, 4790k -> 11700k is only a 25% improvement. E: And in some tests the 4790k is supposedly significantly faster. I don't trust those results. Ika fucked around with this message at 23:11 on Mar 12, 2021 |
# ? Mar 12, 2021 23:04 |