|
SwissCM posted:What I want is a mobile 2c/4t Ryzen with as many Vega cores as possible crammed in alongside it for mobile gaming, something that could fit into a 3DS/GPDWin like formfactor. I have no sources whatever but I think 4c/4t is the smallest CPU you will see in the Zen2 lineup. Perhaps they will roll out the Athlon or whatever old name for smaller parts. That said, compared to the size of a GPU, 4c/4t 7nm Zen CPUs will be tiny. I also have no idea if they could make Zen or Vega/Navi that efficient.
|
# ? Dec 18, 2018 05:15 |
|
|
# ? May 15, 2024 01:49 |
|
.
sincx fucked around with this message at 05:50 on Mar 23, 2021 |
# ? Dec 18, 2018 05:24 |
|
Bring back the Thunderbird but name it Phoenix
|
# ? Dec 18, 2018 06:17 |
|
Wrong thread.
|
# ? Dec 18, 2018 06:55 |
|
ratbert90 posted:Bring back the Thunderbird but name it Phoenix They already did, just
|
# ? Dec 18, 2018 14:32 |
|
I don't know much about CPU architectures and whatnot. I'm looking at Ryzen for my next build but I'm just curious, why is there such a pronounced IPC difference between AMD and Intel CPUs? Is it just clockspeed basically? Also if I buy an AM4 platform now, will it actually be just drop in replacement later? Or will it need firmware updates from the motherboard manufacturers?
|
# ? Dec 18, 2018 19:00 |
|
The Gunslinger posted:I don't know much about CPU architectures and whatnot. I'm looking at Ryzen for my next build but I'm just curious, why is there such a pronounced IPC difference between AMD and Intel CPUs? Is it just clockspeed basically? IPC from what i know is just how much info you can do per hz/clockspeed and intel is going increasingly wide so they can have better IPC for their clock speed as they can't win the core wars.
|
# ? Dec 18, 2018 19:12 |
|
The Gunslinger posted:I don't know much about CPU architectures and whatnot. I'm looking at Ryzen for my next build but I'm just curious, why is there such a pronounced IPC difference between AMD and Intel CPUs? AMD essentially sat and did nothing of value towards improving their CPUs for a good 6 or 7 years, if anything their performance regressed slightly. Then they finally came out with CPUs that were big improvements on what they had back in 2010, but they still haven't had enough time with competency to catch up to Intel all the way.
|
# ? Dec 18, 2018 19:28 |
|
Also, Intel Quicksync is a monster when optimized for it.
|
# ? Dec 18, 2018 19:57 |
|
wargames posted:IPC from what i know is just how much info you can do per hz/clockspeed and intel is going increasingly wide so they can have better IPC for their clock speed as they can't win the core wars. Ah yes, longer and deeper pipelines per clock cycle, where have I heard this before... /s
|
# ? Dec 18, 2018 19:58 |
|
SwissArmyDruid posted:Ah yes, longer and deeper pipelines per clock cycle, where have I heard this before... I remember those days of the inferno of heat coming out of the Pentium 4 and CRT monitor. So much heat for a 2.4 GHz single cpu with hyperthreading. The temperature in the office was intolerable back then. I'm waiting to see how Zen 2 compares with the new intel cpus. Especially given that intel seems to be secretly adding an extra AVX pipeline to some of their cpus.
|
# ? Dec 18, 2018 23:08 |
|
AVX (especially the 512bit version) isn't used all that much though still and probably will never be. Developers just don't seem to be as interesting in pushing it into software everywhere as they were with SSE2 for instance. 256bit AVX does seem to be getting some more use but its still being slow rolled unfortunately quite a bit. Unless you've got something specifically that uses it then it doesn't seem to be all that much to be excited about right now as far as CPU features go. Something like increasing the L1 by 50%+ would matter far more for instance. Brings back memories of the first Intel MMX CPU's. The Gunslinger posted:why is there such a pronounced IPC difference between AMD and Intel CPUs? Is it just clockspeed basically? Now vs the Bulldozer derivatives there was a big difference, but that was because AMD hosed up. It takes years to design a new CPU so there wasn't much they could do about it for a long time. The Gunslinger posted:Or will it need firmware updates from the motherboard manufacturers?
|
# ? Dec 19, 2018 00:52 |
|
SwissCM posted:What I want is a mobile 2c/4t Ryzen with as many Vega cores as possible crammed in alongside it for mobile gaming, something that could fit into a 3DS/GPDWin like formfactor. I, too, want computers that aren't physically possible.
|
# ? Dec 19, 2018 01:08 |
|
PC LOAD LETTER posted:AVX (especially the 512bit version) isn't used all that much though still and probably will never be. Developers just don't seem to be as interesting in pushing it into software everywhere as they were with SSE2 for instance. 256bit AVX does seem to be getting some more use but its still being slow rolled unfortunately quite a bit. Unless you've got something specifically that uses it then it doesn't seem to be all that much to be excited about right now as far as CPU features go. I do have some software that uses AVX/AVX2 and might be able to compile for AVX-512. With Zen 2 doubling the pipeline for AVX2 buying intel doesn't provide much difference now. AMD having Zen 2 with a higher core count with high clock speeds is likely to win me over for my next workstation purchase. Just waiting to see what the official announcement is.
|
# ? Dec 19, 2018 01:11 |
|
I wonder how feasible it would be to move most AVX units to some kind of math co-processor socket. Ideally there would be some units on package for the average user, but if you have workloads that lean hard on AVX, you could trade a little bit of latency for not having the central processor thermal throttle.
|
# ? Dec 19, 2018 01:22 |
|
Bloody Antlers posted:I wonder how feasible it would be to move most AVX units to some kind of math co-processor socket. Ideally there would be some units on package for the average user, but if you have workloads that lean hard on AVX, you could trade a little bit of latency for not having the central processor thermal throttle. I want to see this just because it would be so 1988.
|
# ? Dec 19, 2018 01:41 |
|
Devian666 posted:I do have some software that uses AVX/AVX2 Bloody Antlers posted:I wonder how feasible it would be to move most AVX units to some kind of math co-processor socket. Latency would be too high to be practical for most things and at that point something like a FPGA or doing it on the GPU starts to look more interesting anyways. Putting a bunch of AVX FPU's on a separate die on the same package or in the interposer might be more practical. Dunno if anyone would buy it though. Look at how well Knights Landing did in the market place for instance. Seemed to be mostly a flop despite the hype.
|
# ? Dec 19, 2018 01:51 |
|
Bloody Antlers posted:I wonder how feasible it would be to move most AVX units to some kind of math co-processor socket. Ideally there would be some units on package for the average user, but if you have workloads that lean hard on AVX, you could trade a little bit of latency for not having the central processor thermal throttle. This is a dsp or Gpu basically and would suffer from the same issues
|
# ? Dec 19, 2018 02:00 |
|
Yeah if it were done it would probably be in the context of something like this where it could be a chiplet added onto a package with cores, I/O, GPU, FPGA, etc as part of a heterogeneous design.
|
# ? Dec 19, 2018 03:46 |
|
PC LOAD LETTER posted:Ah fair enough. AVX has to be super duper close to the memory system, specifically the L2/L3 caches for it to work well. AVX has a shitload of throughput, and requires equally impressive memory speeds to feed it. If you're gonna make something that's super good at AVX, it's gotta have a crapton of memory bandwidth, which isn't something you get off-package without a pretty hefty penalty. I could kinda see a dedicated AVX co-proc hung off the PCI-e 4 bus, but not really outside of a dedicated compute card, and I can't think of something that's so AVX heavy you'd want to use it over a general purpose GPU and associated frameworks and compilers. More likely I could see a FPGA card or a security co-proc, something that'll let Facebook's gateway servers handle more of the increasingly expensive HTTPS handshakes for less electricity.
|
# ? Dec 19, 2018 06:04 |
|
Bloody Antlers posted:I wonder how feasible it would be to move most AVX units to some kind of math co-processor socket. Ideally there would be some units on package for the average user, but if you have workloads that lean hard on AVX, you could trade a little bit of latency for not having the central processor thermal throttle. Hello friend I would like to introduce you to something called a graphics processing unit
|
# ? Dec 19, 2018 06:50 |
|
Dr. Fishopolis posted:I, too, want computers that aren't physically possible. The SoC in the GPD Win 2 (a Core M3) can be pushed to 15w TDP and though it runs pretty hot and noisy pushed that hard, GPD are fairly amateurish and the cooling solution could likely be done better. The Ryzen 3 2200U is 15w but what I'd rather is to sacrifice CPU frequency and core count for more Vega cores.
|
# ? Dec 19, 2018 07:00 |
|
Dr. Fishopolis posted:I, too, want computers that aren't physically possible. I too, want computers that aren't physically possible, but am under no illusions that we might not get there this generation or the next, but might the generation after. It's kind of how progress works. edit: SPEAKING OF. quote:Researchers from MIT and the University of Colorado have fabricated a 3-D transistor that’s less than half the size of today’s smallest commercial models. To do so, they developed a novel microfabrication technique that modifies semiconductor material atom by atom. http://news.mit.edu/2018/smallest-3-d-transistor-1207 The paper notes that this is NOT silicon, mind. This was done using InGaAs. SwissArmyDruid fucked around with this message at 08:33 on Dec 19, 2018 |
# ? Dec 19, 2018 07:22 |
|
SwissCM posted:The SoC in the GPD Win 2 (a Core M3) can be pushed to 15w TDP and though it runs pretty hot and noisy pushed that hard, GPD are fairly amateurish and the cooling solution could likely be done better. The Ryzen 3 2200U is 15w but what I'd rather is to sacrifice CPU frequency and core count for more Vega cores. The whole Ryzen mobile stack can be configured for 15w tdp.
|
# ? Dec 19, 2018 08:55 |
|
Arzachel posted:The whole Ryzen mobile stack can be configured for 15w tdp. In my theoretical magic computer I'd rather sacrifice cores for more frequency. Also a freesync display. Under $1000.
|
# ? Dec 19, 2018 09:24 |
|
Kazinsal posted:Hello friend I would like to introduce you to something called a graphics processing unit More like a Xeon Phi which had its uses but was not what I'd call an unmitigated success. It was a lot easier to code for than a GPU, though.
|
# ? Dec 19, 2018 10:24 |
|
My gut feel says that even in single core bound tasks the difference should be negligible if turbo works properly and everything that uses 3+ cores would run much better. I built a 2200g desktop for my dad, might test it and see if I'm wrong
|
# ? Dec 19, 2018 10:27 |
|
The Tegra X1 (arm + maxwell) has a TDP of 15w but likely uses less in many cases. As mentioned above, the entire mobile stack from AMD can be configured for 15w, though I have a hard time believing it does not drink deeper from the VRM given how dangerous these laptops are for one's reproductive health. Maybe it is possible, but at the moment a 20nm chip (fabed on a mobile focused process) from 2015 that likely costs way less seems more efficient for a mobile game do-dad than AMD's products that have their roots in all profitable data center designs.
|
# ? Dec 19, 2018 20:03 |
|
I replaced my 1700 and Asrock AB350 Pro4 with a 2700 and an MSI B450M Mortar. I've managed to overclock the CPU to 4GHz and my 32GB of 2666MHz memory is running at 2933MHz.
|
# ? Dec 24, 2018 10:05 |
|
Measly Twerp posted:I replaced my 1700 and Asrock AB350 Pro4 with a 2700 and an MSI B450M Mortar. I've managed to overclock the CPU to 4GHz and my 32GB of 2666MHz memory is running at 2933MHz. Is this without fiddling with voltage?
|
# ? Dec 25, 2018 18:21 |
|
EmpyreanFlux posted:Is this without fiddling with voltage? By selecting a preset provided by the MSI BIOS I'm now running at 3066MHz. When I purchased the motherboard I was expecting to be stuck at 2666MHz like I was previously, I'm more than happy.
|
# ? Dec 26, 2018 11:49 |
|
I'm super happy with my new 2700x setup so far. Running PE level 4 with a -0.075 offset and it will boost all cores at 4.25 GHz without any issues. I still have a ton of different things to try out too.
|
# ? Dec 27, 2018 06:49 |
|
Do you get a measurable improvement in performance compared to just letting the CPU do it's own thing? I wasn't able to.
|
# ? Dec 27, 2018 08:59 |
|
LRADIKAL posted:Do you get a measurable improvement in performance compared to just letting the CPU do it's own thing? I wasn't able to. I recently built a 2700X build for my younger brother, and other than updating the bios and installing the chipset drivers, I didn't mess with cpu settings, the attraction of the 2700X being it's pretty much max'd out out-of-the-box. (of course I corrected the ram speed) vv Hm, I'm pretty sure I was seeing 4050MHz all-core with the newest BIOS updates and the "AMD Balanced" or whatever it's called power profile from the newest chipset drivers, but I don't have the machine in front of me to confirm HalloKitty fucked around with this message at 19:02 on Dec 27, 2018 |
# ? Dec 27, 2018 10:22 |
|
LRADIKAL posted:Do you get a measurable improvement in performance compared to just letting the CPU do it's own thing? I wasn't able to. Yes. If I remember correctly, stock was right around 3.9 GHz all core max. PE 3 got me up to 4.1 all core, and PE 4 gets me to to 4.25 while both will still boost up to 4.34 on individual cores . Temps were pretty wild on PE 4 so I started using the negative voltage offset. I've seen people run PE3 with BCLK adjustments and positive voltage offsets up to 4.2-4.3 all core and I'll probably play around with that at some point. I'll eventually get around to seeing what I can do with my RAM as well. I was able to get to 3200 without issue with the DOCP profile.
|
# ? Dec 27, 2018 18:29 |
|
I mean as far as benchmarking performance, I was able to squeeze out slightly higher frequencies, but letting it do it's own thing was just as performant, but more cool and stable.
|
# ? Dec 27, 2018 19:38 |
|
LRADIKAL posted:I mean as far as benchmarking performance, I was able to squeeze out slightly higher frequencies, but letting it do it's own thing was just as performant, but more cool and stable. On my current settings I got a 1914 in Cinebench. The lowest I have saved is an 1825, but I can't remember what the settings for that were. I'm pretty sure I had some runs in the high 1700's stock but didn't get them saved.
|
# ? Dec 27, 2018 19:52 |
|
HalloKitty posted:I recently built a 2700X build for my younger brother, and other than updating the bios and installing the chipset drivers, I didn't mess with cpu settings, the attraction of the 2700X being it's pretty much max'd out out-of-the-box. (of course I corrected the ram speed) Don't use that power profile anymore. The Microsoft balanced one is the one to use.
|
# ? Dec 27, 2018 22:31 |
|
Stanley Pain posted:Don't use that power profile anymore. The Microsoft balanced one is the one to use. I just wonder why it installs by default with the chipset drivers then. It holds much higher idle speeds than the built in balanced profile, and I'm going to guess AMD probably has left it like that for a reason.
|
# ? Dec 28, 2018 01:49 |
|
|
# ? May 15, 2024 01:49 |
|
HalloKitty posted:I just wonder why it installs by default with the chipset drivers then. It holds much higher idle speeds than the built in balanced profile, and I'm going to guess AMD probably has left it like that for a reason. There's an ancient blog on this: https://community.amd.com/community/gaming/blog/2017/04/06/amd-ryzen-community-update-3
|
# ? Dec 28, 2018 02:12 |