|
DrDork posted:To be fair, the GPU thread has been 90% NVidia talk and 10% "if you are looking for low-end, AMD is ok, otherwise don't bother" for years, and Navi/Polaris hasn't really changed that up too much--unlike Ryzen/TR have on the CPU side, so maybe that's not the best comparison. sort of, its also been AMD drivers aren't terrible, and mid teir AMD is an option but the best and most $$$$ option is nvidia.
|
#
?
Oct 4, 2019 15:51
|
|
|
|
| # ? May 27, 2024 05:09 |
|
|
A grand unified CPU thread would be neat. I think it's mostly the same posters in both threads anyways. As long as no one starts talking about Mill, I can't see any flame wars happening.
|
|
#
?
Oct 4, 2019 16:28
|
|
|
I'm cool with separate threads, we get enough crosstalk already with different convos goin simultaneously
|
|
#
?
Oct 4, 2019 16:37
|
|
|
I think having separate threads is fine and having a unified one would also be fine. Thank you for reading my opinion and have a nice day.
|
|
#
?
Oct 4, 2019 17:54
|
|
|
Separate threads are important because you can address a lot of platfrom specific things and troubleshooting and component guides. I think a really good addition would be a general CPU thread. Now that both RISC-V and POWER are open source and there's cool ARM Workstation and server boards getting released regularly, it could be a really informative addition.
|
|
#
?
Oct 4, 2019 18:00
|
|
|
We can consolidate them by having a third, catch-all thread.
|
|
#
?
Oct 4, 2019 18:37
|
|
|
Kazinsal posted:As long as no one starts talking about Mill, I can't see any flame wars happening. NewFatMike posted:Now that both RISC-V and POWER are open source and there's cool ARM Workstation and server boards getting released regularly, it could be a really informative addition.
|
|
#
?
Oct 4, 2019 18:47
|
|
|
NewFatMike posted:Separate threads are important because you can address a lot of platfrom specific things and troubleshooting and component guides. Don't forget, more threads = better than.
|
|
#
?
Oct 4, 2019 18:55
|
|
|
Whatever happened to Gandalf the cpu architect and the mill
|
|
#
?
Oct 4, 2019 19:28
|
|
|
I would go with a single thread for news and discussion and a different thread for personal projects, support or questions.
|
|
#
?
Oct 4, 2019 19:28
|
|
|
The Something Awful Forums > Discussion > Serious Hardware / Software Crap > CPU Architecture Megathread: RISC-Vailures
|
|
#
?
Oct 4, 2019 19:30
|
|
|
Mill so fucky that intel used the name for the most irrelevant knights product
|
|
#
?
Oct 4, 2019 20:42
|
|
|
Drakhoran posted:On the other hand Intel now has actual competition and you can get ECC support on AM4 boards. there is also the X470D4U/X470D4U-2T if you want something with IPMI for a server build. There's quite a few X399s with IPMI as well, I think both Asrock and Supermicro have them and probably others. TheFluff posted:The prices listed are wholesale - MSRP for purchases of batches of 1000 units. In the videocardz leak article though the other CPU's they're showing for comparison are shown with their retail prices. These new parts are almost certainly going to be more expensive retail. Tray prices usually end up being pretty much retail prices, and some retailers (Microcenter) even push notably beneath them.
|
|
#
?
Oct 4, 2019 22:09
|
|
|
Separate threads are better imo, because every now and then, there's going to be some tangent that goes on for five pages. I think it's great when it happens, but it's going to feel annoying if it happens in an omnibus thread.
|
|
#
?
Oct 4, 2019 22:46
|
|
|
We can always have a general CPU thread and when a release happens make one for that and close it down when it's been more than a few months or so after release.
|
|
#
?
Oct 4, 2019 22:56
|
|
|
NewFatMike posted:platfrom specific
|
|
#
?
Oct 4, 2019 22:59
|
|
|
I mean, any time it's been brought up in the AMD thread, I've been against it, but if you guys are down, what the heck, we can do a grand experiment, and if it all blows up, we can go back to our little corners.
|
|
#
?
Oct 4, 2019 23:23
|
|
|
im gonna miss the multithreaded meltdowns
|
|
#
?
Oct 4, 2019 23:44
|
|
|
eames posted:im gonna miss the multithreaded meltdowns
|
|
#
?
Oct 5, 2019 00:06
|
|
|
I like the idea of combining the threads and simultaneously pushing out tech support to focus on news and technology.
|
|
#
?
Oct 5, 2019 00:38
|
|
|
ufarn posted:no way to predict how the discussion will branch now 
|
|
#
?
Oct 5, 2019 02:01
|
|
|
What if we abstract this further and make the thread able to support discussion about other types of computation like FPGAs and DSPs and what about mechanical computers or analogue electronic computers and what about
|
|
#
?
Oct 5, 2019 02:40
|
|
|
As a reader for the majority of the time, I am quite happy with two threads. I have been in the Intel camp for a loooong time and switched to the new Ryzen recently and was glad I could keep myself updated in a single, dedicated thread respectively. Even now I hardly want to keep up with all things Intel, so I appreciate the two separate threads. I'd rather have a third general CPU architecture thread than merge the two.
|
|
#
?
Oct 5, 2019 03:27
|
|
|
mcbexx posted:I'd rather have a third general CPU architecture thread than merge the two.
|
|
#
?
Oct 5, 2019 11:25
|
|
|
VostokProgram posted:What if we abstract this further and make the thread able to support discussion about other types of computation like FPGAs and DSPs and what about mechanical computers or analogue electronic computers and what about I think there is probably enough discussion to support a thread about vector processing and its implications for computer graphics edit: ironically, speaking of which...  "it's really all about the shroud" -- raja koduri Paul MaudDib fucked around with this message at 13:12 on Oct 5, 2019 |
|
#
?
Oct 5, 2019 12:50
|
|
|
Raja Koduri and Jim Keller both have white beards because they are silicon wizards
|
|
#
?
Oct 6, 2019 05:43
|
|
|
Raja is a shroud elemental.
|
|
#
?
Oct 6, 2019 11:00
|
|
|
Upon that fan is etched thousands of tiny runes, and its rotation opens the portal to an elemental plane whereupon data is cast and data is returned. The shroud is necessary to keep portal from leaking.
|
|
#
?
Oct 6, 2019 11:44
|
|
|
Intel cut the prices of their 9th gen non-iGPU SKUs by $20-30. Not a huge deal but it's something I guess.  https://www.anandtech.com/show/14941/intel-announces-price-cut-for-9th-generation-f-and-kf-processors
|
|
#
?
Oct 7, 2019 14:52
|
|
|
Didn't we do the whole 'merge the threads' thing to the IT bitching threads like 2 years ago, and it was met with a resounding 'ehhhhhhh' and went back to 2 threads fairly quickly?
|
|
#
?
Oct 7, 2019 23:25
|
|
|
eames posted:Intel cut the prices of their 9th gen non-iGPU SKUs by $20-30. Not a huge deal but it's something I guess. That's what the F/KFs should have been in the first place. Too little too late at this point. Intel's product is fine as far as it goes, the 8700K and 9700K would be fine in the mid $200s (8700K is $250 at Microcenter), 9900K would be fine around $400 (putting it in competition with the 3800X) and the 9900KF in mid/high 300s. But pricing the 9900K head to head against 12C AMD processors for another year is a losing move. They can justify a small premium, not like a 50% premium anymore. I had hopes after the aggressive price cuts on HEDT. Maybe they're planning to drop the price points of the 10-series chips instead.
|
|
#
?
Oct 7, 2019 23:36
|
|
|
As much as we give intel poo poo about about 14nm++++++++++ it is fairly impressive they’ve been able to squeeze more cores and clock speed out at lower voltages. My 6700K required 1.38V to hit 4.6 ghz, my 9900K requires 1.216V for 4.7 ghz. Sample size of one of course but thats double the cores and threads at a higher clock at a much lower voltage.
|
|
#
?
Oct 7, 2019 23:40
|
|
|
B-Mac posted:As much as we give intel poo poo about about 14nm++++++++++ it is fairly impressive they’ve been able to squeeze more cores and clock speed out at lower voltages. My 6700K required 1.38V to hit 4.6 ghz, my 9900K requires 1.216V for 4.7 ghz. Sample size of one of course but thats double the cores and threads at a higher clock at a much lower voltage. I just commented on this in the GPU thread (someone was asking about overclocking) but even if you compare the 8600K and 9700K, 87% of 8600Ks getting 5 GHz at 1.40V while 90% of 9700Ks get 5 GHz at 1.35V. So, 2 more cores running and they still dropped 0.05V off the core, and a few more chips per batch meet the bin. Considering that Intel didn't even add a + that's kind of impressive there. Apart from power consumption being higher than 7nm (duh), 14nm has always been a super impressive node. It may be one of the highest-clocking nodes of all time, given that 10nm and 7nm are not looking so great. I wonder if delidded golden-sample 9900KS will be able to punch up to 5.2 or 5.3 at higher voltage... maybe 5.4 with HT turned off? Paul MaudDib fucked around with this message at 23:54 on Oct 7, 2019 |
|
#
?
Oct 7, 2019 23:45
|
|
|
I'm not saying you're necessarily wrong about 7nm in the long run, but Zen 2 is clocking right up there with Skylake while also having a significant IPC and core count advantage.
|
|
#
?
Oct 8, 2019 00:39
|
|
|
Stickman posted:I'm not saying you're necessarily wrong about 7nm in the long run, but Zen 2 is clocking right up there with Skylake while also having a significant IPC and core count advantage. I don't think Zen 2 has much, if any, IPC advantage over Skylake. If you control for cache sizes, Skylake IPC is basically identical to Coffee Lake R.
|
|
#
?
Oct 8, 2019 05:56
|
|
|
B-Mac posted:As much as we give intel poo poo about about 14nm++++++++++ it is fairly impressive they’ve been able to squeeze more cores and clock speed out at lower voltages. My 6700K required 1.38V to hit 4.6 ghz, my 9900K requires 1.216V for 4.7 ghz. Sample size of one of course but thats double the cores and threads at a higher clock at a much lower voltage. Z390 changed the way Vcore is measured, so 1.2V on Z390 is roughly equivalent to 1.3V on Z170-370. CPUs are often being tested on their respective platforms so that’s something to keep in mind. If Vcores were that much lower then 9th gen would run much cooler than 8th (not to mention the soldered IHS!) but that’s not the case. Intel’s optimizations are still impressive but I thought that’s worth mentioning. We’re now in the middle of the era that people predicted years ago, with 14nm being so refined that it takes years for smaller processes to catch up with it in frequency/single thread performance assuming an unlimited power envelope, unless they come with major architectural design changes (and security fixes, heh). eames fucked around with this message at 07:16 on Oct 8, 2019 |
|
#
?
Oct 8, 2019 07:04
|
|
|
eames posted:Z390 changed the way Vcore is measured, so 1.2V on Z390 is roughly equivalent to 1.3V on Z170-370. CPUs are often being tested on their respective platforms so that’s something to keep in mind. If Vcores were that much lower then 9th gen would run much cooler than 8th (not to mention the soldered IHS!) but that’s not the case. I’m running my 9900K on a z370 taichi. Do you have a link that explains how vcore measures have changed for z390? I’d be interested in reading it.
|
|
#
?
Oct 8, 2019 12:38
|
|
|
B-Mac posted:I’m running my 9900K on a z370 taichi. Do you have a link that explains how vcore measures have changed for z390? I’d be interested in reading it. This is for Asus boards but the effect was observed on most Z370/390 boards. Some speculated that Intel corrected the spec to make it appear like Vcore from 8th to 9th gen stayed the same when it was actually bumped up by >0.1V.  https://www.overclock.net/forum/27686004-post2664.html SIO (traditional) is what Z370 Maximus X Hero indicates, SIO (Maximus XI Hero) is what the Z390 shows. This is again right from elmor who worked in Asus motherboard R&D at the time. people have also found a way to read that "corrected" voltage on older boards: https://www.overclock.net/forum/6-intel-motherboards/1638955-z370-z390-vrm-discussion-thread-376.html#post27795758 To be fair it's hard to accidentally destroy a 9900K by overvolting because it simply becomes too hot under load (unless have a very advanced custom watercooling loop or something better). eames fucked around with this message at 13:54 on Oct 8, 2019 |
|
#
?
Oct 8, 2019 13:16
|
|
|
Methylethylaldehyde posted:Didn't we do the whole 'merge the threads' thing to the IT bitching threads like 2 years ago, and it was met with a resounding 'ehhhhhhh' and went back to 2 threads fairly quickly? It figures that the Intel thread would be in favor of reducing the number of threads and call it an improvement.
|
|
#
?
Oct 8, 2019 13:17
|
|
|
|
| # ? May 27, 2024 05:09 |
|
|
Dramicus posted:It figures that the Intel thread would be in favor of reducing the number of threads and call it an improvement.
|
|
#
?
Oct 8, 2019 13:41
|
|
