|
e: nm!
Cygni fucked around with this message at 21:40 on Oct 19, 2018 |
# ? Oct 19, 2018 21:33 |
|
|
# ? May 31, 2024 02:19 |
|
BIG HEADLINE posted:The clincher for me is that it seems the 9900K *cannot* be used reliably with high-end air. Period, full stop. It supposedly auto-throttles to 4Ghz whenever it hits its thermal limit, unless you remove that limit in BIOS, at which point if it goes too high it simply crashes, which I'd have to worry about too much even with a Noctua D15. You could always figure out what wattage you cooler can handle for the CPU to stay at 95C (lol) and set your short and long term turbo power limit (PL1/PL2) to that number. It’ll do pretty granular throttling to hit that target (VID, multiplier in 1x steps, all per individual core). Bonus: you won’t have to worry about an AVX offset either. I do agree that the 9700k seems to make more sense though. Has any page tested 6C/12T vs 8C/8T at otherwise identical conditions (mainly frequency) yet? e: vvv 95W „complex task“ TDP with a 210W power limit eames fucked around with this message at 21:50 on Oct 19, 2018 |
# ? Oct 19, 2018 21:35 |
|
Pretty toasty https://twitter.com/IanCutress/status/1053357591580082176
|
# ? Oct 19, 2018 21:40 |
|
BIG HEADLINE posted:The clincher for me is that it seems the 9900K *cannot* be used reliably with high-end air. Period, full stop. I think you're overreacting a little bit. The Anandtech review we're talking about was on air (e: assuming TRUE copper is actually talking about the old Thermalright heatsink). Others were on air or 240/280 AIOs which are going to be equal to the highest end Noctua. It's borderline if not actually an HEDT sort of chip in consumer clothing and you should expect to cool it like one. Under other loads it does its job pretty reasonably too: Winks fucked around with this message at 21:51 on Oct 19, 2018 |
# ? Oct 19, 2018 21:41 |
|
BIG HEADLINE posted:And yes, the 2700X option is there, but sorry, even though the numbers are impressive, I like monolithic dies and Infinity Fabric reminds me way too much of HyperTransport from the bad old days and gives me a "duct tape and baling wire" feeling. And Hypertransport was awesome back then and IF is awesome now. Kinda power hungry for what it is but over all power consumption for AMD CPU's isn't bad at all with Zen/Zen+, especially vs these new top end 9xxx series Intel chips, so its a moot issue. Buy what you want and all but your reasoning here doesn't seem to pan out, just saying it seems more emotional that reality based.
|
# ? Oct 19, 2018 21:47 |
|
Oh boy good luck z370 owners with VRMs not specced for above 140w draw
|
# ? Oct 19, 2018 21:50 |
|
Anime Schoolgirl posted:
Kind of want to go back and look at Buildzoid tearing apart the Godlike to see if that VRM setup makes any more sense now that it's blown past even his most absurd ideas of the draw
|
# ? Oct 19, 2018 21:54 |
|
240 watts on a desktop consumer cpu yuck. Buy the 9700k instead
|
# ? Oct 19, 2018 21:54 |
|
Wonder how much power it uses at 4.3 turbo.
|
# ? Oct 19, 2018 22:01 |
|
Anime Schoolgirl posted:
I certainly feel better about recommending the Extreme4 to people in the builder's thread.
|
# ? Oct 19, 2018 22:02 |
|
Cygni posted:AT has some power charts using POV Ray, which you can extrapolate to thermals: Anandtech's power figures are a complete outlier and possibly a result of a BIOS fluke since Kitguru found auto-voltage on their Z370 Extreme4 was putting 1.51v into the core. Most other sites are finding 200-220W total system power in non-AVX tasks... so comparable to the 2700X +/- a bit for variation between different boards and stuff. https://www.computerbase.de/2018-10/intel-core-i9-9900k-i7-9700k-cpu-test/3/ https://www.pcper.com/reviews/Processors/Intel-Core-i9-9900K-Review-Competition-Renewed/Power-Consumption-Overclocking-and https://techreport.com/review/34192/intel-core-i9-9900k-cpu-reviewed/13 https://www.overclock3d.net/reviews/cpu_mainboard/intel_core_i9-9900k_and_asus_z390_strix-e_review/20 of course, those thermals are utterly atrocious regardless, doesn't seem like anyone got them much past 5 GHz so they are pretty much maxed out right out of the box. And the reason is... die thickness and a thick layer of solder, really? Der8auer says maybe the thick layer gives it more flex so it doesn't crack but the performance still seems pretty poor. I thought solder was going to help a lot more gently caress it Intel just give me a bare die already and I'll get a die guard, jesus christ (lol they probably will release a refresh or something, just like Haswell) edit: BIG HEADLINE posted:I certainly feel better about recommending the Extreme4 to people in the builder's thread. Bad news friendo (fixable, but still an oopsie) Paul MaudDib fucked around with this message at 22:46 on Oct 19, 2018 |
# ? Oct 19, 2018 22:06 |
|
We shouldn't have complained about the toothpaste
|
# ? Oct 19, 2018 22:10 |
|
Paul MaudDib posted:Bad news friendo (fixable, but still an oopsie) I've already steered people away from getting the 9900K, and it should still run the 9700K just fine.
|
# ? Oct 19, 2018 22:11 |
|
140w nominal power consumption probably isn't going to kill a high-end z370 board which are already specced for that anyway.
|
# ? Oct 19, 2018 22:12 |
|
Winks posted:We shouldn't have complained about the toothpaste
|
# ? Oct 19, 2018 22:13 |
|
BIG HEADLINE posted:I've already steered people away from getting the 9900K, and it should still run the 9700K just fine. at 1.51v default voltage I bet it runs really well for a week
|
# ? Oct 19, 2018 22:14 |
|
Anime Schoolgirl posted:140w nominal power consumption probably isn't going to kill a high-end z370 board which are already specced for that anyway. Will you ever render or compile? No? -> 9900K is
|
# ? Oct 19, 2018 22:14 |
|
I'm happy to be on a X99, I figured a 9900K would be drawing like 90 watts and get to 5.5GHz regularly. Once I have apps that can use 8 cores I'll go to a 5960x. I was going to do that anyways, but I'm not terribly out of date with Haswells like I thought I might be.
|
# ? Oct 19, 2018 22:25 |
|
Winks posted:Work/Watt is in line with everything else. Same amount of work in a shorter period of time means higher heat output during that shorter period They're dumb if you want real world numbers, but they are handy if used as a consistent benchmark across different CPUs. As far as wattage and work, eams already mentioned this, but reviewers have noted a power-use runaway effect as the tempertures climb. As he puts it: eames posted:
|
# ? Oct 19, 2018 22:26 |
|
eames posted:Intel never wanted to switch, the reviewers and consumers put a lot of pressure on them with the toothpaste memes and all. I’m not going to lie, I also wanted soldered CPUs back but so far it doesn’t look like it was worth it. They've got a lot out of the 14nm process, but it does look like they've hit a wall. Their 10nm shrink has been a fiasco and their inability to make it happen has cost them a lot of ground. Cannonlake was supposd to be a die shrink of Kaby Lake. I'm sure they would have loved to be putting out the 9900k as a 10nm part, which would have brought wattages down to a level more in line with their other high-end consumer products. It'll be an interesting time for CPUs in the next couple years. eames posted:
I haven't seen anyone do this yet but I am also really interested to see some benchmarks of the 8700k vs 9700k, both at 4.9ghz. Digital Foundry is usually good about doing these types of like-for-like tests, so fingers crossed. In most multi-threaded benchmarks, the stock 9700k beats the 8700k by a small margin, but the 8700k @ 4.9ghz leapfrogs the stock 9700k. il serpente cosmico fucked around with this message at 23:07 on Oct 19, 2018 |
# ? Oct 19, 2018 22:34 |
|
Intel CPU and Platform Discussion: Meltdowns ITT
|
# ? Oct 19, 2018 22:58 |
|
Lol, no poo poo, reactions like this are great popcorn material! Soldering didn't help, AMD has more cores! No performance per watt improvement!
|
# ? Oct 19, 2018 23:02 |
|
Worth running with aio if you are not overclocking? For 9900k?
|
# ? Oct 19, 2018 23:05 |
|
It depends on your case, if you have no space for a big tower a AIO is a good option, but they're louder if you can fit a big tower. Well, I had a H100I and a D14 to compare, newer 280MM coolers might be better.
|
# ? Oct 19, 2018 23:09 |
|
Ulio posted:Worth running with aio if you are not overclocking? For 9900k? I personally like the look of AIOs over the giant slab of metal. Performance is going to be about the same between a 280 AIO and something like a Noctua D15. It's too bad the Noctua refresh isn't on shelves yet which would have better performance and ditch the brown. LRADIKAL posted:No performance per watt improvement! Were people expecting improvements other than a handful of spectre hardware things? It's still just
|
# ? Oct 19, 2018 23:16 |
|
It seems like some people were expecting IPC or power consumption improvements. Probably it was just taken for granted at this point.
|
# ? Oct 19, 2018 23:20 |
|
LRADIKAL posted:Lol, no poo poo, reactions like this are great popcorn material! Soldering didn't help, AMD has more cores! No performance per watt improvement! Its pretty much the same core as before with a couple more CPU's and modestly higher boost clocks from a minor process improvement. It was pretty much guaranteed to be a power hungry, furnace-like, and expensive chip for a minor performance improvement vs previous Intel series CPU's. It was never reasonable to expect anything else. The only surprise to me is that Intel managed to mess up a soldered TIM which is exacerbating the heat issue just enough to make it irritating to cool even with top end air HSF's or a decent 240mm radiator AIO water cooler.
|
# ? Oct 19, 2018 23:23 |
|
Ulio posted:Worth running with aio if you are not overclocking? For 9900k? It depends on where you (or your mainboard) set the hard power limit. You could get stock to ~250W with torture loads by tinkering with just that, or choose a sane setting like I dunno 120W which is fine for any decent tower cooler. No easy answers, just make sure you get a mainboard that exposes that setting to the user. sauer kraut fucked around with this message at 23:35 on Oct 19, 2018 |
# ? Oct 19, 2018 23:24 |
|
LRADIKAL posted:It seems like some people were expecting IPC or power consumption improvements. Probably it was just taken for granted at this point. I didn’t see anyone expecting IPC or power consumption improvements, it’s skylake with 8 cores. Obviously we now know that soldered TIM was a requirement for these chips, I just wasn’t expecting temps be as high as they are with the 9900k being soldered. Though temps and power consumption are all over place based on different reviews.
|
# ? Oct 19, 2018 23:33 |
|
I only posted the thread title because the 9900K runs a bit too warm at full sprint, not because any posters were reacting too intensely
|
# ? Oct 19, 2018 23:34 |
|
Paul MaudDib posted:Anandtech's power figures are a complete outlier and possibly a result of a BIOS fluke since Kitguru found auto-voltage on their Z370 Extreme4 was putting 1.51v into the core. Most other sites are finding 200-220W total system power in non-AVX tasks... so comparable to the 2700X +/- a bit for variation between different boards and stuff. welp, sure seems like it. https://twitter.com/IanCutress/status/1053397551112695814 ive been pretty loyal to AT for decades cause they were always the only ones who actually understood the market, but im gettin shaken lately a bit. hope our boy Tebe over at GN does some more power numbers later.
|
# ? Oct 19, 2018 23:43 |
|
I was expecting the change to a soldiered TIM to make up for the additional power and heat the two additional cores were putting out. I also thought it'd overclock decently well, since delidded coffee lakes have much better thermals than stock ones.
|
# ? Oct 20, 2018 00:02 |
|
il serpente cosmico posted:I was expecting the change to a soldiered TIM to make up for the additional power and heat the two additional cores were putting out. I also thought it'd overclock decently well, since delidded coffee lakes have much better thermals than stock ones. It's more that they're putting the default boost clocks right at the remaining headroom they had
|
# ? Oct 20, 2018 00:32 |
|
Xerophyte posted:This is still about equal to the work gain from HT, even in ancient povray which has rubbish scaling. 25%ish watts for 25%ish ops. E: I am bad at charts, 50% ish watts is of course less fun.
|
# ? Oct 20, 2018 00:47 |
|
Anime Schoolgirl posted:It's more that they're putting the default boost clocks right at the remaining headroom they had Completely agree.
|
# ? Oct 20, 2018 01:07 |
|
Yep, it was pretty obvious why they soldered these: 210w through a 178mm2 strip of thermal paste is unrealistic for more than a fraction of a second. I highly doubt it was in response to consumers complaining about the old paste, in order to hit 4.7-5.0 turbo speeds for any useful duration or load they had to solder it or it would thermal throttle before it even came close. I'd say the 7700k pretty much hit the efficiency wall on 14nm at 4.5 GHz, they made it to the 8700k/8086k on binning but it looks like even that avenue is pretty much exhausted. On the other hand, my custom loop water cooler with a 420x38MM high density copper radiator is suddenly looking like a pretty good investment after all. Especially after reading the THG review where their 280mm AIO hit a 90C package temp. (My 7700k used to do that at stock, then I delidded it and swapped the paste for liquid metal, now 4.8 GHz AVX torture only draws it up to the low 70s and any realistic load typically stops in the mid 60s.)
|
# ? Oct 20, 2018 01:11 |
|
Indiana_Krom posted:Yep, it was pretty obvious why they soldered these: 210w through a 178mm2 strip of thermal paste is unrealistic for more than a fraction of a second. I highly doubt it was in response to consumers complaining about the old paste, in order to hit 4.7-5.0 turbo speeds for any useful duration or load they had to solder it or it would thermal throttle before it even came close. I'd say the 7700k pretty much hit the efficiency wall on 14nm at 4.5 GHz, they made it to the 8700k/8086k on binning but it looks like even that avenue is pretty much exhausted. idk it looks like they shot themselves in the foot by doubling the thickness of the die. It isn't clear that the thick die + soldier is actually a better solution than 8th gen's thin die with paste. We're going to see companies offering delidding / lapping /liquid metaling services this generation, rather than the previous few generations of simply delidding / liquid metaling. This whole things seems like a bit of a boondoggle.
|
# ? Oct 20, 2018 01:25 |
|
Bring toothpaste™ back
|
# ? Oct 20, 2018 03:23 |
|
FX-9900k Vishera Lake
|
# ? Oct 20, 2018 03:29 |
|
|
# ? May 31, 2024 02:19 |
|
Tokamak posted:FX-9900k Vishera Lake There's a joke in here about a molten salt reactor, but I can't quite make it work.
|
# ? Oct 20, 2018 03:42 |