|
SwissArmyDruid posted:I'm begging you, AMD, please don't put Vega into anything anymore. I want that Surface Go 2 with Navi in it, at minimum. I had no idea how badly I want that exact thing until you said it.
|
# ? Jun 26, 2019 19:54 |
|
|
# ? Apr 24, 2024 09:59 |
|
Got a lil surface go to complement my surfacebook 2 (when I need the whole laptop and need to notetake still) and a little extra oomph would be a nice add to the QOL of the device (and, ya know stop loving selling a eMMC storage tier).
|
# ? Jun 26, 2019 22:09 |
|
mdxi posted:Darn those a-cussed knaves in the press, with their unfair price comparisons of the chips Intel trots out to win benchmarks, to the chips AMD trots out to win benchmarks! And remember that Zeppelin is still a monolithic die, so it's not like AMD had an advantage in manufacturing on the consumer platform. CCX architecture actually reduces yields (there are no 4+3 or 4+2 chips, so a dead core on one CCX "kills" one on the other CCX). And AMD still drastically undercut Intel. Just think about the margins that Intel used to be getting with a quad-core as their $350 flagship part. It was a good performer for its core count but holy poo poo was Intel taking people to the cleaners with the 6000/7000 series pricing. Broadwell-E prices actually went up over Haswell-E. 18 months later and you're getting an 8-core (9700K) for the same (theoretical) price and AMD is pushing them down again despite being on a newer, more expensive node. If Intel had just put 6 cores on the mainstream platform in the 6000 or even 7000 series then Ryzen would have spun its wheels badly. Those launch reviews with the poo poo-tier BIOS and 2400 memory with loose timings, going against an 8700K, would have been absolute murder. Intel got greedy and gave AMD an opening and now AMD is running the table on them. Paul MaudDib fucked around with this message at 23:34 on Jun 26, 2019 |
# ? Jun 26, 2019 23:15 |
|
Intel got super greedy. I know 90% of the board still runs a 2500K but honestly, that shouldn't be possible to get away with. That CPU is 8 years old now, it came out in 2011. Imagine going back ANOTHER 8 years and getting a mid-range CPU from then (2003). Maybe a 2.4GHZ single-core Pentium 4? Then trying to still use that against a i5 2500K in 2011 to play the top games then (Battlefield 3, Deus Ex HR, Skyrim, Crysis 2). Not going to happen yet my 2500K can play anything released in 2019 even if frame times aren't as smooth as I might like.
|
# ? Jun 27, 2019 00:32 |
|
Alpha Mayo posted:Imagine going back ANOTHER 8 years and getting a mid-range CPU from then (2003). Maybe a 2.4GHZ single-core Pentium 4? Then trying to still use that against a i5 2500K in 2011 to play the top games then (Battlefield 3, Deus Ex HR, Skyrim, Crysis 2). Not going to happen yet my 2500K can play anything released in 2019 even if frame times aren't as smooth as I might like. Oh yes those bastards got greedy but the multicore "revolution" started 2 years after your quoted 2003 which is a long time in tech by itself and comparing single cores to real quad cores isn't remotely apples to apples either. Your statement also ignores all of the supporting technology around CPUs that make them run better like the bandwidth improvements from DDR2/3/4 and PCI-E for graphics
|
# ? Jun 27, 2019 03:18 |
|
I have a 3570k and I'm finally getting around to upgrading. This year is the first year where I've had issues with new releases, and I have actually had to question how well a title will run. That being said, how well does AMD stuff overclock usually? Outside of the crazy overclocking community. I attribute my ability to use my 3570k for 7 years because I overclocked it a solid amount.
|
# ? Jun 27, 2019 03:35 |
|
Zotix posted:I have a 3570k and I'm finally getting around to upgrading. This year is the first year where I've had issues with new releases, and I have actually had to question how well a title will run. Lately AMD has been pushing their chips nearly to the wall in terms of clocks at stock voltage, to the point where their "PBO" dynamic auto-OC can actually get you better performance in lightly-threaded tasks than manual all-core OCs. The Ryzen 3000 series is on a new process so you'll have to wait until reviews are out to see how they overclock but definitely don't expect Sandy/Ivy Bridge-level OCs (in terms of gains or ease).
|
# ? Jun 27, 2019 03:48 |
|
Zotix posted:I have a 3570k and I'm finally getting around to upgrading. This year is the first year where I've had issues with new releases, and I have actually had to question how well a title will run. Most of the overclocking for AMD is with memory. You can overclock the CPU a little bit, usually by feeding the auto overclock stuff (pbo/xfr) a bit more headroom. You can also do classical overclocking stuff, but it only comes out ahead in some workloads. Usually all-core. The lower end zen2 SKUs likely have overclock headroom. The higher end ones that can hit 4.6-4.7 single core likely won't. Khorne fucked around with this message at 05:08 on Jun 27, 2019 |
# ? Jun 27, 2019 05:05 |
|
Khorne posted:Modern CPUs don't overclock that well because the companies aren't leaving 30%+ performance gains on the table for you to overclock with. Why did Intel leave so much headroom on their stock clocks until recently? Just to make the extra $50 from gamers willing to jump to the K SKU?
|
# ? Jun 27, 2019 05:13 |
|
Alpha Mayo posted:Intel got super greedy. I know 90% of the board still runs a 2500K but honestly, that shouldn't be possible to get away with. That CPU is 8 years old now, it came out in 2011. How does “Intel is super greedy” square with “Intel gave a lot of people unprecedented value by not topping themselves in any significant manner”?
|
# ? Jun 27, 2019 05:20 |
|
Craptacular! posted:How does “Intel is super greedy” square with “Intel gave a lot of people unprecedented value by not topping themselves in any significant manner”? Personality I'd prefer having to buy a new CPU sooner over complete technological stagnation.
|
# ? Jun 27, 2019 05:43 |
|
BeastOfExmoor posted:Why did Intel leave so much headroom on their stock clocks until recently? Just to make the extra $50 from gamers willing to jump to the K SKU? can't have chips failing the bin if you set the bar incredibly low /taps forehead tbh I think the desktop enthusiast market has been an afterthought for a long time for Intel. A lot of stuff makes more sense if you view them as primarily a server/mobile/OEM driven company. They have a couple various bins with modest clocks and modest cooling requirements because that's what the OEMs will pay for, then they throw a generic overclocking SKU at the enthusiast market and say "go to town". And it's not like AMD gave any real competition up until recently, so they could get away with it. Even still they could be doing a better job of this. They still only have one clockspeed bin. Why didn't they keep going with the 8086K strategy into the 9-series? Bin out the best parts and make people pay another 20% for them. If they're looking for ways to keep their mainstream parts reasonable without completely tanking their margins, that's an obvious approach. Paul MaudDib fucked around with this message at 06:29 on Jun 27, 2019 |
# ? Jun 27, 2019 05:53 |
|
Paul MaudDib posted:can't have chips failing the bin if you set the bar incredibly low /taps forehead Intel:
|
# ? Jun 27, 2019 06:07 |
|
But seriously, in Intel's eyes, what brings in more money, the OEM market or the enthusiast market?
|
# ? Jun 27, 2019 06:10 |
|
In a similar boat. I want to hit 5ghz on all core, and I'll have a decent amount of water radiator space to throw at whatever I end up with. But not going to go LN2 or anything too crazy. Going to be interesting to see how much headroom they actually have.
|
# ? Jun 27, 2019 06:12 |
|
Another question, in my past week of looking at components and searching around for information, it seems like delidding CPU's is a thing now. I guess it was big for intel 8700k, and people had a lot of success with delidding their CPU's and having reduced temps. I haven't seen anyone talk about this on the AMD side for previous Ryzen releases. Is this something that really only has traction on the Intel side of things?
|
# ? Jun 27, 2019 06:18 |
|
Zotix posted:Another question, in my past week of looking at components and searching around for information, it seems like delidding CPU's is a thing now. I guess it was big for intel 8700k, and people had a lot of success with delidding their CPU's and having reduced temps. I haven't seen anyone talk about this on the AMD side for previous Ryzen releases. Is this something that really only has traction on the Intel side of things? Two reasons: 1) The IHS is soldered to the chip rather than pasted like the 8700k, so it's much more difficult to delid and the thermal interface is better anyway, so there's less reason to. 2) There's just not a lot of overclocking headroom on air/aios past what you can get with the stock soldering, so there's really not much performance benefit to delidding. Zen 2 is soldered as well, but there might still be some benefit to delidding if the power curve is less sharp than Zen or Zen+. The solder will still make delidding dangerous to try yourself! Intel's 9th-gen chips like the 9900k and 9700k are also soldered, so there's less incentive to delid than with the 8700k.
|
# ? Jun 27, 2019 06:26 |
|
MaxxBot posted:Personality I'd prefer having to buy a new CPU sooner over complete technological stagnation. I wish we didn’t have to wait until Threadripper to see more cores on a high end part, but for gaming the lack of pressing need to buy any new CPUs for years and years is great.
|
# ? Jun 27, 2019 06:28 |
|
Stickman posted:Two reasons: Thank you.
|
# ? Jun 27, 2019 06:36 |
|
Paul MaudDib posted:Even still they could be doing a better job of this. They still only have one clockspeed bin. Why didn't they keep going with the 8086K strategy into the 9-series? Bin out the best parts and make people pay another 20% for them. If they're looking for ways to keep their mainstream parts reasonable without completely tanking their margins, that's an obvious approach. The 9900KS is surely going to be just that, Intel playing silicon lottery.
|
# ? Jun 27, 2019 06:38 |
|
iospace posted:But seriously, in Intel's eyes, what brings in more money, the OEM market or the enthusiast market? The only reasons the "enthusiast market" isn't a complete afterthought for everyone are that: (1) They (we?) are a super-vocal super-minority of early adopters. Like a particularly neckbeardy and overweight (speaking for myself) sort of Instagram influencers. (2) In some cases, those enthusiasts either are the ones making purchasing decisions for other lines of business (datacenters/IT type stuff), or they might one day become those people. So it's worth a little marketing effort to make them like you. Videos, slogans, and tshirts are cheap as all hell compared to fabbing silicon. In terms of the addressable market, enthusiasts are noise compared to the signal of corporate budgets (including, in AMD's case, semicustom work).
|
# ? Jun 27, 2019 06:38 |
|
Zotix posted:Is this something that really only has traction on the Intel side of things? Its not necessary for most of the AMD Zen/+/2's desktop chips because they use a soldered IHS which conducts heat much better to the HSF and makes delidding mostly pointless. You can maybe reduce temps a few degrees C if you do direct die cooling vs a soldered IHS which isn't much and delidding a soldered IHS is a bigger risk and pain in the rear end too. So pretty much no one does it. It only became a thing with Intel's chips because their TIM-to-IHS interface caused the temps to skyrocket when overclocked so you'd easily end up being heat limited before you became limited by power or the quality of the die even with good cooling. AMD did use a cheaper and more generic 'toothpaste' TIM on their APU's but I think they're going to start soldering the IHS on some of those with the 3xxx series APU's. Overclocking the CPU/iGPU on those didn't tend to go far in giving you more performance anyways and they were very much budget oriented too so pretty much no one was going to invest in really good cooling to bother either which is why you don't hear anyone really complain to much about them. Generally overclocking the memory as much as you could was where you could get the most performance out of the APU's. The iGPU's are always bandwidth starved with DDR4 2400/2666.
|
# ? Jun 27, 2019 06:40 |
|
The 3600 geekbench scores have jumped up another 100+ points, this is still at 4.3GHz and almost cracking 6000, the memory performance here is really good. https://browser.geekbench.com/v4/cpu/13678723 MaxxBot fucked around with this message at 07:51 on Jun 27, 2019 |
# ? Jun 27, 2019 07:43 |
|
MaxxBot posted:The 3600 geekbench scores have jumped up another 100+ points, this is still at 4.3GHz and almost cracking 6000, the memory performance here is insane. I also plan on just driving to microcenter on release day and hoping for the best. So I may not even get zen2 on release day.
|
# ? Jun 27, 2019 07:46 |
|
This dude should have a lot of RAM benchmarks out hopefully on release day. https://twitter.com/1usmus/status/1143095451325009920
|
# ? Jun 27, 2019 07:58 |
|
MaxxBot posted:This dude should have a lot of RAM benchmarks out hopefully on release day.
|
# ? Jun 27, 2019 08:02 |
|
MaxxBot posted:The 3600 geekbench scores have jumped up another 100+ points, this is still at 4.3GHz and almost cracking 6000, the memory performance here is really good. Do I need to double the reported 1800MHz ram speed to get the actual one? I.e. this is running at 3600MHZ?
|
# ? Jun 27, 2019 08:15 |
|
Yeah it's 3600MHz
|
# ? Jun 27, 2019 08:18 |
|
What would be the best 32GB RAM configuration for a 3900X? 2x16 or 4x8? Or is it still too early to tell?
|
# ? Jun 27, 2019 08:45 |
|
No way to say for sure yet but Buildzoid has said that generally 4x 8GB DIMMs should be better than 2x 16GB DIMMs as far as overclocking goes in general. How much better though? No one knows yet. At least publicly. That 1usmus' twitter linked a little ways up would be the place to watch if you want to find out come launch day which is 7/7. That guy is pretty much the messiah of AMD memory overclocking right now.
|
# ? Jun 27, 2019 09:45 |
|
PC LOAD LETTER posted:No way to say for sure yet but Buildzoid has said that generally 4x 8GB DIMMs should be better than 2x 16GB DIMMs as far as overclocking goes in general. Ah, I think unless you are on a T-Topology motherboard then 2 is always better than 4 modules for overclocking. The reason is, desktop CPUs only have 2 memory controllers, and when it has to be split between two modules there is increased latency to one of the modules due to being physically further away. Best memory overclocking is done on DTX/ITX boards with only 2 DDR slots
|
# ? Jun 27, 2019 11:52 |
|
Lube banjo posted:Ah, I think unless you are on a T-Topology motherboard then 2 is always better than 4 modules for overclocking. But pretty much all the 16GB sticks right now are dual rank and they tend to OC poorly vs the 8GB single rank DIMMs in general on either Intel or AMD. Zen2 is supposed to be much better about handling DR DIMMs than Zen/+ was but according to guys like Buildzoid it still probably won't tolerate them as well as SR DIMMs when overclocking. Also AMD is supposed to be tuning all the AGESA's for daisy chain and not T topology layouts too for Zen2. Lube banjo posted:Best memory overclocking is done on DTX/ITX boards with only 2 DDR slots
|
# ? Jun 27, 2019 12:44 |
|
Also the superior slots are the further slots, the problems come from signals reflecting off what are effectively unterminated extensions if you occupy the near slots.
|
# ? Jun 27, 2019 12:51 |
|
If the "leaked" benchmarks for the R5 3600 are real...
|
# ? Jun 27, 2019 14:25 |
|
BeastOfExmoor posted:Why did Intel leave so much headroom on their stock clocks until recently? Just to make the extra $50 from gamers willing to jump to the K SKU? I suspect it was because they didn't know what the silicon was actually capable of long-term or if there was degradation from running them that hard. Intel is a cautious company, and shipping a bunch of chips only to find out that they cooked themselves and you have entire datacenters worth of RMA claims on your hands would be catastrophic.
|
# ? Jun 27, 2019 15:05 |
|
spasticColon posted:If the "leaked" benchmarks for the R5 3600 are real... As a small form factor person (65 W maximum), this is highly intriguing to me.
|
# ? Jun 27, 2019 15:09 |
|
Lambert posted:As a small form factor person (65 W maximum), this is highly intriguing to me. E: nm. The 3600 non X is 65w. Because that makes sense.
|
# ? Jun 27, 2019 15:14 |
|
BeastOfExmoor posted:Why did Intel leave so much headroom on their stock clocks until recently? Just to make the extra $50 from gamers willing to jump to the K SKU? Lack of reason to do so. If they set the bar really low (1 core 3.7GHz on a 2500K... are there any 2500Ks that even fail to run at 4.2GHz at stock voltage?) then their yields shoot up, and therefore so does profit. If you look at the competition from around that time, everything makes total sense. Why bin chips more rigorously when AMD was not really much of a threat? Easier to simply make more profit and sell overclocking as a benefit. AMD's binning and pushing their chips really hard these days, and it shows. They need to do it, and it's winning them customers. Intel's also having to do it too, to remain competitive. Overclocking headroom just isn't what it once was. ilkhan posted:Iirc the 3600 is the 95W 6 core. The 3700 is the 65w 8 core. Binning HalloKitty fucked around with this message at 16:43 on Jun 27, 2019 |
# ? Jun 27, 2019 16:40 |
|
You also have to keep in mind that 8 years ago Turbo Boost was still considered sorta radical and you'd see doofuses arguing to disable it, never mind something like PBO. CPU dynamic power and clock management was just starting to get good. Good coolers and remotely decent case fans were just becoming mainstream. People would post on forums afraid to OC a 2500k at stock voltage because they thought it would literally explode like an OG Athlon or something would if you let the heatsink fall off it. Intel was far from alone in being conservative. Nowadays we recognize that it's fine to just push poo poo to the moon as long as the voltage and thermals are fine, but back then it was far from mainstream to think that way.
|
# ? Jun 27, 2019 17:01 |
|
|
# ? Apr 24, 2024 09:59 |
|
Scott Herkelman and Robert Hallock on PCWorld podcast stream: https://www.youtube.com/watch?v=OY8qvK5XRgA They're talking both Ryzen 3000 and Radeon 5700.
|
# ? Jun 27, 2019 19:27 |