|
.
sincx fucked around with this message at 05:55 on Mar 23, 2021 |
#
?
Oct 11, 2017 20:04
|
|
|
|
| # ? Apr 28, 2024 10:27 |
|
|
sincx posted:Don't underestimate all the small IPC improvements. I mean 30% is totally not trivial, it only really looks that way as it's 5+ years of billions of dollars profit to get that.
|
|
#
?
Oct 11, 2017 20:55
|
|
|
sincx posted:Don't underestimate all the small IPC improvements. 30% ipc when virtually nothing is cpu-limited these days is not really worth buying a new computer over
|
|
#
?
Oct 11, 2017 21:06
|
|
|
My 2500k gets 2.1 single core score in Cinebench 11.5, and an 8700k gets 2.27. Not too bad.  I wish they did more legacy benchmarks like 100+ modpack minecraft using late game full factory setup processing 5000 Laser Miners + 200 Mekanism's Elite Factories. Or late game OpenTTD with stations looking like this:  gently caress you, grudston!
|
|
#
?
Oct 11, 2017 21:08
|
|
|
.
sincx fucked around with this message at 05:55 on Mar 23, 2021 |
|
#
?
Oct 11, 2017 21:11
|
|
|
MaxxBot posted:From the linus vid It says value - productivity, but then it gives "MSRP" and "Total System". Presumably it's some coefficient involving price and productivity, but uh who the hell knows how they work. The chart does not even appear to be sorted by any known metric because as far as I know Threadripper 1950x is the most expensive setup on it. Khorne fucked around with this message at 04:38 on Mar 15, 2018 |
|
#
?
Oct 11, 2017 21:18
|
|
|
Khorne posted:Hello. I am a data scientist with an MS in Physics. What the gently caress is this chart showing? I have no idea but it has a meme in it so it’s obviously good.
|
|
#
?
Oct 11, 2017 21:21
|
|
|
Ak Gara posted:My 2500k gets 2.1 single core score in Cinebench 11.5, and an 8700k gets 2.27. Not too bad. You might like some of GN's tests, specifically the Civ 6 turn time one. https://www.gamersnexus.net/hwreviews/3076-intel-i7-8700k-review-vs-ryzen-streaming-gaming-overclocking/page-5 I wish he would run it on all the older CPUs too. Khorne posted:Hello. I am a data scientist with an MS in Physics. What the gently caress is this chart showing? The vid has a key that was cropped from the screenshot.  MSRP is chip only, total system includes all the platform costs. I don't know how they weight the scores although I think remember a forum post about it. https://www.youtube.com/watch?v=EJOnwF8mgXc&t=522s Cygni fucked around with this message at 21:32 on Oct 11, 2017 |
|
#
?
Oct 11, 2017 21:26
|
|
|
Khorne posted:Hello. I am a data scientist with an MS in Physics. What the gently caress is this chart showing? more like MS in Paint
|
|
#
?
Oct 11, 2017 22:43
|
|
|
Combat Pretzel posted:I'm wary of them used, because who knows how the previous owner massacred them.
|
|
#
?
Oct 11, 2017 22:56
|
|
|
Anime Schoolgirl posted:They're soldered, if they're busted it's because they were busted to begin with. Nobody who looks to resell is going to bother delidding solder because that's the realm of liquid helium. It's more a question of whether they've been run for years with excessive amounts of voltage. Electromigration is a thing and pushing your chips really hard will eventually hurt them. Some people ran into that with Sandy Bridge...
|
|
#
?
Oct 11, 2017 23:25
|
|
|
Volguus posted:Heh. With the latest and greatest CPU you may be able to cut that time in half. Compiling on tmpfs even more so (Linus got his kernel compiled in a bunch of seconds some time ago on tmpfs). Then again ... isn't 2 minutes enough? In my days it would take me 40min to 1 hour for the drat thing. I know, I know - but everything's built and I don't want to RMA everything. Thanks for the tip on tmpfs. Next month I'm going to score another 16GB of RAM and hopefully keep a 5GB tmpfs drive
|
|
#
?
Oct 12, 2017 04:12
|
|
|
Just finished putting together my 8600K build, and it feels good man. However, wtf, even mid-range boards now have gaudy color shifting LEDs over everything, and they're on even when the board is off? I thought it was ridiculous before.
|
|
#
?
Oct 12, 2017 05:23
|
|
|
.
sincx fucked around with this message at 05:55 on Mar 23, 2021 |
|
#
?
Oct 12, 2017 06:53
|
|
|
mobby_6kl posted:Still, after almost a decade ( Paul MaudDib posted:Eurogamer: Followup: TechPowerUp did the 8700K:  tl;dr: always use at least 2666, 3000 CAS14/3200 CAS15 is a few percent better, don't worry too much about going past that, or about timings or whatnot. It may make more of a difference in memory-bound games like ARMA or Witcher 3 in Novigrad. Paul MaudDib fucked around with this message at 17:25 on Oct 12, 2017 |
|
#
?
Oct 12, 2017 17:22
|
|
) with a Q6600, I just want to go full retard with a 8700 or one of the other 6-cores that can be comparably overclocked. RAM is still ridiculously expensive though and fast RAM even more so. Has anyone seen any benchmarks testing the 8700 with varying memory speeds?
|
My 8700k system just shipped, much earlier than expected. I hope this lasts as long as my 2500k did 
|
|
#
?
Oct 12, 2017 17:30
|
|
|
the only shop in my country with 8700K in stock is already sold four of them at $685 (incl. vat). launch day preorders keep getting pushed back to as late as end of november. paper launch is real, at least in the eu.
|
|
#
?
Oct 12, 2017 17:30
|
|
|
Paul MaudDib posted:It's more a question of whether they've been run for years with excessive amounts of voltage. Electromigration is a thing and pushing your chips really hard will eventually hurt them. Some people ran into that with Sandy Bridge...
|
|
#
?
Oct 12, 2017 18:22
|
|
|
So my Asus motherboard enabled BLCK overclocking to 4.1~ on my i5-8400 by default. I'm not really mad about this, I don't think its amazing but since I have the cooling capacity for it, I'll let it ride. However, I can see why it would be problematic for a lot of users.
|
|
#
?
Oct 12, 2017 22:12
|
|
|
Michael Jackson posted:more like MS in Paint
|
|
#
?
Oct 13, 2017 00:32
|
|
|
VulgarandStupid posted:So my Asus motherboard enabled BLCK overclocking to 4.1~ on my i5-8400 by default. I'm not really mad about this, I don't think its amazing but since I have the cooling capacity for it, I'll let it ride. However, I can see why it would be problematic for a lot of users. Interesting. Which board?
|
|
#
?
Oct 13, 2017 00:53
|
|
|
VulgarandStupid posted:So my Asus motherboard enabled BLCK overclocking to 4.1~ on my i5-8400 by default. I'm not really mad about this, I don't think its amazing but since I have the cooling capacity for it, I'll let it ride. However, I can see why it would be problematic for a lot of users. 
|
|
#
?
Oct 13, 2017 01:04
|
|
|
B-Mac posted:Interesting. Which board? Z370-I but from what I’ve been seeing on the YouTube circuit is it’s probably all of the Asus ones, although many other manufacturers might be guilty of it as well. Some don’t even know what they have on by default.
|
|
#
?
Oct 13, 2017 01:15
|
|
|
Combat Pretzel posted:What he says. If I'm going to drop a few hundred on an used part, I'd like for it to last a while. This is why I'm kinda eyeballing the Xeon 1660v3... unlocked, and it probably hasn't been overclocked. Not sure it's worth a $100 premium over a 5960X though.
|
|
#
?
Oct 13, 2017 01:20
|
|
|
VulgarandStupid posted:So my Asus motherboard enabled BLCK overclocking to 4.1~ on my i5-8400 by default. I'm not really mad about this, I don't think its amazing but since I have the cooling capacity for it, I'll let it ride. However, I can see why it would be problematic for a lot of users. Factory OCing the FSB by 1-2MHz to make boards look better on benches has been a favorite vendor trick since the Pentium II days at least.
|
|
#
?
Oct 13, 2017 04:24
|
|
|
Palladium posted:Factory OCing the FSB by 1-2MHz to make boards look better on benches has been a favorite vendor trick since the Pentium II days at least. Ha, wow, I never knew this was a thing. Seems kinda sketchy (imho) to have this set as default.
|
|
#
?
Oct 13, 2017 05:32
|
|
|
Has there been an IPC comparison with the last few generations yet? IIRC Skylake/Kaby lake had almost 0 IPC gain at all. I think they have to move to 10nm for actual gains, but I've been wrong before
|
|
#
?
Oct 13, 2017 12:09
|
|
|
Isn't kaby even very slightly slower per clock?
|
|
#
?
Oct 13, 2017 12:14
|
|
|
Skylake, Kaby Lake, and Coffee Lake all use the same basic pipeline so there should be no change clock-for-clock.
TheJeffers fucked around with this message at 12:59 on Oct 13, 2017 |
|
#
?
Oct 13, 2017 12:55
|
|
|
Don't expect the initial Intel 10nm chips to be any good if 14nm is of any indication; Think Broadwell, but without the IPC gains. Rumors on the Internets is Sapphire Rapid will finally be Intel's significant uarch revamp in awhile which won't be seen any earlier than 2019 . BTW reports are confirming multicore enhancement on non-K doesn't work anymore with CFL.
|
|
#
?
Oct 13, 2017 13:03
|
|
|
Palladium posted:BTW reports are confirming multicore enhancement on non-K doesn't work anymore with CFL. If by this you mean that non-K's can't bump their all-core max over Intel's listed all-core speed (vs bumping them generally up towards Intel's dual- or even single-core max), then CFL will be a hilarious stealth downgrade for a bunch of people.
|
|
#
?
Oct 13, 2017 13:08
|
|
|
DrDork posted:If by this you mean that non-K's can't bump their all-core max over Intel's listed all-core speed (vs bumping them generally up towards Intel's dual- or even single-core max), then CFL will be a hilarious stealth downgrade for a bunch of people. To clarify a bit more, one can no longer exceed the stock turbo for the respective # of cores on the non-K chips. In addition, BCLK OCing is also dead outright on CFL as the CPU will detect and refuse to work above 103MHz BCLK...Unless there is some mobo vendor exploit we don't know of yet.
|
|
#
?
Oct 13, 2017 13:18
|
|
|
Lube banjo posted:Has there been an IPC comparison with the last few generations yet? IIRC Skylake/Kaby lake had almost 0 IPC gain at all. I think they have to move to 10nm for actual gains, but I've been wrong before Skylake, Kaby and Coffee are all essentially the same
|
|
#
?
Oct 13, 2017 14:44
|
|
|
LLC on some Z370 Asus Strix boards is not working and "Auto" Vcore defaults to 1.4V  The amount of "auto OC" built into those motherboards is getting pretty ridiculous but how else are they going to top the benchmarks for differentiation from better and cheaper boards.
|
|
#
?
Oct 13, 2017 17:50
|
|
|
I've definitely had issues with the "Auto" vcore even with older chips, with my 6600k it was giving the chip waaay more voltage than necessary before I dialed in a manual voltage. EDIT: Yeah this issue was with an Asus mobo MaxxBot fucked around with this message at 18:44 on Oct 13, 2017 |
|
#
?
Oct 13, 2017 18:30
|
|
|
That's been a problem since like the Pentium 2 days. I remember always telling people to set a manual voltage to what the processor was rated to or later when offset was the way to go offset by +.005. It also might be specifically an Asus problem because I've been using Asus motherboards since the Pentium 3 days too and way back then it already was a problem. This generation seems to be the worst yet though. Edit: Asus does it on their videocards too, but to a lesser extent and it's usually more beneficial there since you don't get nearly the voltage control as with a motherboard. craig588 fucked around with this message at 18:37 on Oct 13, 2017 |
|
#
?
Oct 13, 2017 18:35
|
|
|
Increasing BCLK makes my eye twitch from how it messes up some PCIe devices when the mobo uses that as the pcie refclk.
|
|
#
?
Oct 13, 2017 18:41
|
|
|
Got my 8700k up and running, just running with XMP and all-core turbo for now but it's already flying. An x264 job that took 3min22sec on my 2500k@4.4 takes 1min7sec on here 
|
|
#
?
Oct 13, 2017 19:13
|
|
|
priznat posted:Increasing BCLK makes my eye twitch from how it messes up some PCIe devices when the mobo uses that as the pcie refclk. I may be mistaken, but I believe Intel has long put BCLK and refclks on separate domains specifically so overclockers can toy with it without messing up everything else. AMD hasn't yet implemented similar domains on Ryzen CPUs.
|
|
#
?
Oct 13, 2017 19:33
|
|
|
|
| # ? Apr 28, 2024 10:27 |
|
|
TheJeffers posted:I may be mistaken, but I believe Intel has long put BCLK and refclks on separate domains specifically so overclockers can toy with it without messing up everything else. AMD hasn't yet implemented similar domains on Ryzen CPUs. That’d be good, at my work we had issue with an asus z77 board that came with the bclk set higher (or the co-op did it and lied about it) and it caused a lot of grief. That’s a pretty old board though.
|
|
#
?
Oct 13, 2017 19:35
|
|