|
Otakufag posted:Bring toothpaste™ back Be careful; they'll bring it back and make it the die even thicker. Maybe they could wedge a layer of plastic between the die and the lid.
|
#
?
Oct 20, 2018 03:53
|
|
|
|
| # ? Jun 2, 2024 19:00 |
|
|
Anandtech has updated their review Notice: When we initially posted this page, we ran numbers with an ASRock Z370 board. We have since discovered that the voltage applied by the board was super high, beyond normal expectations. We have since re-run the numbers using the MSI MPG Z390 Gaming Edge AC motherboard, which does not have this issue.   They're now in line with the GN review for stock power consumption with an AVX load. So it's actually a pretty efficient chip in the end. Better than my expectations anyway. Does that bug affect their 8700k numbers too, which is why they're so high compared to the 8086k? I think it's more the 8086k numbers are low though. Winks fucked around with this message at 06:02 on Oct 20, 2018 |
|
#
?
Oct 20, 2018 05:58
|
|
|
I was just screwing around with LLC on my ASrock Z370 but with with an 8700K. LLC levels are kind of wacky compared to temps/vid/vcore. I bet that is one of the issues they are having.
|
|
#
?
Oct 20, 2018 06:03
|
|
|
They also posted their OC results (note that their default settings are from before they discovered the auto voltage issue).
|
|
#
?
Oct 20, 2018 07:00
|
|
|
Ok so the 9700K is 69° at 141W gets praised over the moon, while the 9900K at 70°/144W is "one warm chip" Video game journalism right there.
|
|
#
?
Oct 20, 2018 07:07
|
|
|
sauer kraut posted:Ok so the 9700K is 69° at 141W gets praised over the moon, while the 9900K at 70°/144W is "one warm chip" The i7 is clocked 300mhz higher in those given examples? The i7 stays cooler and uses less power at 5.3GHz than the i9 uses at 5.0Ghz. It's not like their assessment comes out of left field. Another way to think about it: Say I have two 8-core CPUs of same architecture generation. Say both of them are running at 5ghz, and both of them are using the exact same cooler, motherboard, power supply, etc. One of them is at 65C under load, and the other is at 87C under load. Would you find it acceptable to consider the latter one "one warm chip," relative to the first? EDIT: Anandtech isn't a gaming website, btw. They have been one of the few reliable PC hardware resources for the past two decades. il serpente cosmico fucked around with this message at 07:59 on Oct 20, 2018 |
|
#
?
Oct 20, 2018 07:44
|
|
|
il serpente cosmico posted:The i7 is clocked 300mhz higher in those given examples? The i7 uses stays cooler and uses less power at 5.3GHz than the i9 uses at 5.0Ghz. It's not like their assessment comes out of left field Yeah, I'd be fine with 5.2-5.3 stable on a 9700K for the next five years, since I really doubt Intel's going to foster in another "every enthusiast is able to get 7+ years of comfortable use out of their CPU" period ever again. There'll be some ~evolution~ in the next five years that'll either sunset current chip architecture...of course, that's kind of like fusion power - everyone thought 10Ghz was coming in five years in the days of Pentium 4 3.2Ghz chips, then Sandy/Ivy Bridge buyers probably thought they'd get ~3-5 years out of their builds before something irresistibly better came along, and so on, and so on...
|
|
#
?
Oct 20, 2018 07:52
|
|
|
Collect your thoughts, buddy! What's your point? Isn't it starting to seem like we're hitting real technological plateau as far as serial computation goes?
|
|
#
?
Oct 20, 2018 07:57
|
|
|
Delidding a Ryzen or Sandy Bridge CPU at room temperature will destroy the die leaving you with half a chip on the IHS and the rest on the substrate. I wonder why this CFL refresh is so different in that regard. People are delidding them just like the old TIM version, just with much more force.
|
|
#
?
Oct 20, 2018 07:58
|
|
|
il serpente cosmico posted:The i7 is clocked 300mhz higher in those given examples? The i7 stays cooler and uses less power at 5.3GHz than the i9 uses at 5.0Ghz. It's not like their assessment comes out of left field. It kind of does because they seem to have forgotten about hyperthreading. It's been easier to clock the 8600k higher than the 8700k, but the 8700k still generally wins out in the end. Usually overclocking comes with performance measurements not just stability, assuming they even tested stability competently, which may be a bold assumption given their issues. And it's not like you couldn't flip HT off on the 9900k and clock it higher for a while if that's what your heart truly desired...then turned it back on because it would be stupid to leave it off. il serpente cosmico posted:EDIT: Anandtech isn't a gaming website, btw. They have been one of the few reliable PC hardware resources for the past two decades. eames posted:Delidding a Ryzen or Sandy Bridge CPU at room temperature will destroy the die leaving you with half a chip on the IHS and the rest on the substrate. I wonder why this CFL refresh is so different in that regard. People are delidding them just like the old TIM version, just with much more force. Thickness of the solder and the die/PCB. Winks fucked around with this message at 08:02 on Oct 20, 2018 |
|
#
?
Oct 20, 2018 08:00
|
|
|
Yep, if 5.3 GHz! is the argument, they'd need to disable hyperthreading and recheck how high the i9 gets in that case. The extra performance from HT is not free.
|
|
#
?
Oct 20, 2018 08:03
|
|
|
They're not wrong, though. With all the threads loaded, the i9 is one warm chip. I wouldn't have thought this would be controversial, given what every other review has found, but here we are.
|
|
#
?
Oct 20, 2018 08:04
|
|
|
il serpente cosmico posted:They're not wrong, though. With all the threads loaded, the i9 is one warm chip. I wouldn't have thought this would be controversial, given what every other review has found, but here we are. Yes if you run 200W through a small CPU, you'll have a warm chip. The argument you want to make is that Intel needed to clamp down on their board partners, Nvidia style, to not do that dumb poo poo.
|
|
#
?
Oct 20, 2018 08:12
|
|
|
sauer kraut posted:Ok so the 9700K is 69° at 141W gets praised over the moon, while the 9900K at 70°/144W is "one warm chip" That's Ian Cutress for you. After Anand left the site to become an actual engineer (more power to him), Anandtech's desktop/enthusiast CPU reviews were handed over to Cutress, and his ability to write coherent sentences is poor. He's still following the review template and methodology Anand put in place, so the reviews are not entirely awful, but you're better off just skimming them for the numbers if you have a negative reaction to mangled English which doesn't mean what the writer intended it to. (Probably. I'm not sure Cutress isn't actually as dumb as some of the things he writes; iirc he's a cryptocurrency enthusiast.) Beyond the language issues, the excellent deep dives into CPU microarchitecture are gone because Cutress does not have the background to do them.
|
|
#
?
Oct 20, 2018 08:32
|
|
|
BobHoward posted:excellent deep dives into CPU microarchitecture If this is what you're looking for then look up anything Jon Stokes ever wrote for Ars Technica. https://arstechnica.com/author/hannibal/ He hasn't been with them for a long time, but his work is second to none. He wrote a book called Inside the Machine that stops around the Core 2 era: https://www.amazon.com/Inside-Machi...27%3AJon+Stokes
|
|
#
?
Oct 20, 2018 08:38
|
|
|
Some other good sources would be Kanter and DeMone who do/did some drat good CPU architecture write ups over at Real World Tech too but DeMone pretty much disappeared after some sort of drama years ago and Kanter isn't as publicly active as he used to be now. Its from 2004 so its dated and some of the conclusions/predictions are off but its amazing how much of the information in this piece by DeMone has held up over the years and its still a real interesting read that is relatively layman friendly.
|
|
#
?
Oct 20, 2018 09:01
|
|
|
sauer kraut posted:Yes if you run 200W through a small CPU, you'll have a warm chip. I think we're just arguing over semantics at this point. Anandtech's initial power tests at stock settings were indeed skewed by their motherboard over-volting, but I wouldn't blame Intel's board partner's for everyone's results. Gigabyte's internal testing results yielded a power draw of 245W at 1.3v at 5 GHz. They hit a temperature of 90C on water. Tom's hardware had similar results: 1.33v, 5.0Ghz, 250 watts, and 85-95C on an AIO water cooler (they hit 205 watts at load at default settings). HardOCP: 1.27v, 5.15ghz, 230 watts, and 85-95C on a custom water-cooler. The 9900k draws a lot of power, and consequently puts out a lot of heat. If you are taking umbrage with the fact that the 9700k vs 9900k isn't a fair comparison because hyper-threading creates additional work, then I will concede that. But the 9900k is still one warm chip. der8auer demonstrates that changes to the physical properties of the die itself are a substantial part of the issue, so it isn't simply a matter of additional work drawing additional power and creating additional heat: https://www.youtube.com/watch?v=r5Doo-zgyQs. On top of all that, tomshw.de's review demonstrated that runaway leakage is also part of the problem. As the chip goes over 80C, it draws more power to do the same amount of work, which raises the temperature, which draws more power. il serpente cosmico fucked around with this message at 09:12 on Oct 20, 2018 |
|
#
?
Oct 20, 2018 09:02
|
|
|
Fixating on GHz is not helpful. Just like with the new GPU boosts, people need to look at what fps/encoding time/whatever is achieved with a certain power limit. Advanced users can tweak a voltage offset on top of that if they don't mind going through a few crashes. And I absolutely blame the partners, imagine buying a 2080 that shipped with a 220% power target by default. I already posted this in the parts picking thread, here is a 8700K/9900K direct comparison at 160W with a crazy water setup, done by German Igor. The solder reduced temps by ~18° in this case. 
sauer kraut fucked around with this message at 11:15 on Oct 20, 2018 |
|
#
?
Oct 20, 2018 11:12
|
|
|
I wonder just how many of the 9900K reviews are OCing the chip with the Fisher-Price one-touch superovervolt EZ edition
|
|
#
?
Oct 20, 2018 12:56
|
|
|
sauer kraut posted:Ok so the 9700K is 69° at 141W gets praised over the moon, while the 9900K at 70°/144W is "one warm chip" I suppose under normal loads it was closer to 78c-80c. 5-5.3GHz on an 8 core CPU with current processes is going to throw off serious heat. That's just physics. Thermals are a poor reason to slam the CPU because it doesn't risk degradation or impede its performance. The revised power draw numbers are reasonable too unless you're comparing them vs zen2 next year because you're from the future. If you care about power draw or thermals, clock it at a number lower than 5, scale voltage appropriately, and it will do great. Cost and cost:performance are reasonable attacks on it. I haven't looked at the benchmarks much yet because I'm trying real hard to hold out for zen2 and I suspect they are "the same as the 8700k, except when the 2 extra cores come into play". Which isn't bad. At the same time, it's just another coffee/sky lake CPU. Khorne fucked around with this message at 15:04 on Oct 20, 2018 |
|
#
?
Oct 20, 2018 14:46
|
|
|
These CPUs are pretty drat solid and the power concerns seem to be completely overblown. Techreport has a few charts that show the 9900 is actually more efficient than 9700 because it does significantly more work in return for the extra power to feed HT.  https://techreport.com/review/34192/intel-core-i9-9900k-cpu-reviewed/13 Also lol 70C° under heavy load isn't worth worrying at all wtf.
|
|
#
?
Oct 20, 2018 15:35
|
|
|
mobby_6kl posted:These CPUs are pretty drat solid and the power concerns seem to be completely overblown. Techreport has a few charts that show the 9900 is actually more efficient than 9700 because it does significantly more work in return for the extra power to feed HT. There is heavy load and then there is AVX load. I don't think those tests have any AVX in them
|
|
#
?
Oct 20, 2018 15:54
|
|
|
B&H has a rather...loose estimate as to when they'll get more 9900K stock in:
|
|
#
?
Oct 20, 2018 23:35
|
|
|
BIG HEADLINE posted:B&H has a rather...loose estimate as to when they'll get more 9900K stock in: That’s when the world will finally be ready for the dodecahedron.
|
|
#
?
Oct 21, 2018 00:53
|
|
|
According to this thread on Reddit, Newegg took 13,690 orders and got 72 units for launch day. US stock, according to Fry's supply manager, was 500 units.
|
|
#
?
Oct 21, 2018 00:58
|
|
|
Jesus Christ, Intel...
|
|
#
?
Oct 21, 2018 01:10
|
|
|
Can you imagine what the initial yield's going to be on the new Core-Xs?
|
|
#
?
Oct 21, 2018 01:20
|
|
|
Meanwhile, in the EU, pretty much all of the Intel CPU's from the i5-8400 and up are still either unobtanium or come with a >100€ premium or both. It's been like this for almost a month now. The i5-8400 used to be 180€. Currently it's 240€, but it's not actually available at that price - estimated delivery early November. If you want it now, it's 280€, and at that point you might as well get an i5-8600K, which is available, at 290€ - used to be around 220€. It's the same for the 8th gen i7's - the i7-8700 is 400€ (used to be well under 300€) and the i7-8700K is €440 (used to be €340). You can get a i5-9600K, but it's 350€, more than what the i7-8700K cost a month ago, and the 8700K is a significantly better CPU. The 9700K is 500€ and out of stock for the foreseeable future, the 9900K 700€ and likewise. 700€ is quite close to the launch price for the Threadripper 2920X.
|
|
#
?
Oct 21, 2018 01:26
|
|
|
Yeah, the 9700K is probably going to be rarer than the 9900K since Intel's probably going to be pretty singularly focused on getting those 9900K preorders satisfied. I know I'm now thinking about getting a Z390 board and pairing it with an 8700K and just ~riding it out~. BIG HEADLINE fucked around with this message at 01:33 on Oct 21, 2018 |
|
#
?
Oct 21, 2018 01:30
|
|
|
Can you overclock fewer than all cores? Does it even make sense?
|
|
#
?
Oct 21, 2018 01:58
|
|
|
I wouldn't be surprised if there is a "coffee canyon" revision that irons out the thermal problems. There was a "CFL-R refresh" after CFL-R on some of the roadmaps. Like Devil's Canyon. You don't need to tape out again, just tweak the packaging, maybe bump clocks a little.
|
|
#
?
Oct 21, 2018 02:00
|
|
|
Wooper posted:Can you overclock fewer than all cores? Does it even make sense? Yes and yes. Intel CPU's with stock settings will run higher frequencies the fewer cores are under load, and you can use that kind of setting for overclocking as well. You could for example configure your CPU to run 5 GHz if all cores are under load but 5.1 if two or fewer cores are loaded, and so on.
|
|
#
?
Oct 21, 2018 02:56
|
|
|
If that’s all the 9900Ks they are shipping (and I don’t believe that) then there are going to be a lot of lonely Z390 boards waiting for a dance partner
|
|
#
?
Oct 21, 2018 03:02
|
|
|
Hed posted:If that’s all the 9900Ks they are shipping (and I don’t believe that) then there are going to be a lot of lonely Z390 boards waiting for a dance partner Micro Center (which traditionally tends to get a fair bit of stock from Intel's distributors) has in the last few hours gone from SOLD OUT on their respective store listings for the 9th gens to NOT CARRIED. The lucky pricks who managed to score what few didn't go to reviewers are going to have one hell of a windfall on eBay, where people who were selling their "guaranteed preorder" slots are probably now hoping their bidders aren't keeping up with the news. I do wonder just how many 9800X chips will be put out initially, or if Intel's over a barrel because they've got too much poo poo coming out at once. BIG HEADLINE fucked around with this message at 03:06 on Oct 21, 2018 |
|
#
?
Oct 21, 2018 03:04
|
|
|
Kinda feel bad selling the 8086k I won seeing this entire fiasco unfold... should probably have kept it. Oh well, guess i'll wait till they have a revision of the 9900k or the next generation Zen to upgrade my 5960x
|
|
#
?
Oct 21, 2018 03:10
|
|
|
BIG HEADLINE posted:Yeah, the 9700K is probably going to be rarer than the 9900K since Intel's probably going to be pretty singularly focused on getting those 9900K preorders satisfied. Paul MaudDib posted:There was a "CFL-R refresh" after CFL-R on some of the roadmaps. You're probably thinking of the non-K ninth generation chips that are coming out end of Q1 or start of Q2 next year. Winks fucked around with this message at 03:22 on Oct 21, 2018 |
|
#
?
Oct 21, 2018 03:19
|
|
|
No, this is what I saw. https://mobile.twitter.com/witeken/status/1042043796660387840 Grain of salt and all. Paul MaudDib fucked around with this message at 03:31 on Oct 21, 2018 |
|
#
?
Oct 21, 2018 03:28
|
|
|
The R in CFL-R already stands for refresh right?
|
|
#
?
Oct 21, 2018 06:30
|
|
|
.
sincx fucked around with this message at 05:55 on Mar 23, 2021 |
|
#
?
Oct 21, 2018 07:29
|
|
|
|
| # ? Jun 2, 2024 19:00 |
|
|
Intel lack of 10nm is background noise versus the complete lack of a single mainstream uarch improvement for 5 years since 6700K.
|
|
#
?
Oct 21, 2018 07:47
|
|