|
BangersInMyKnickers posted:I really hope the new Eypc gen has similar backwards compatibility so Dell doesn't drag their feet getting the new chips out the door. I have a looming project that could really benefit from doubling the core density Yeah it’s drop in compatible with Naples. (Epyc 1)
|
#
?
May 2, 2019 20:08
|
|
|
|
| # ? May 29, 2024 10:33 |
|
|
somebody over at overclock.net found independent Zen2 vendor leaks with matching specs if that 3850X is real... yikes. the site claims 16 cores, 4.3/5.1 Ghz, 125W
|
|
#
?
May 3, 2019 07:24
|
|
|
eames posted:somebody over at overclock.net found independent Zen2 vendor leaks with matching specs I really wouldn't get excited about any "leaks", we've had so many
|
|
#
?
May 3, 2019 07:54
|
|
|
Those are just the same “leaked” specs that wccftech posted back in December, so I’m pretty sure that’s where the numbers came from.
|
|
#
?
May 3, 2019 08:32
|
|
|
eames posted:somebody over at overclock.net found independent Zen2 vendor leaks with matching specs I'm much more interested in the popular options: the 3600X, with 8/16 cores at 4/4.8Ghz, would seem to be a HUGE upgrade over the 2600X at 6/12 cores 3.6/4.2Ghz, and if it's priced similarly, it's an incredible bargain and Intel should start getting really scared the 3700X with 12/24 cores and 4.2/5.0Ghz is a wet dream, especially if it can sustain that boost on, say, 6 cores (or even all core  ) and if these specs come out to be true I'm seriously considering getting it. Hopefully a good B450 (MSI Pro Carbon Gaming AC) motherboard will be able to handle it  Anything more than that and you're either losing clockspeed with the 3800, which isn't really ideal for us gaming people, or probably going to have to sell your firstborn son to afford the 3850X (plus I think 16 cores at 5.1Ghz will be pretty toasty). For games I bet the 3600X and 3700X would be the sweet spot depending on budget, that's of course if leaks prove to be true - but by now we had many, they all seem to point in the same direction and announcement/release date has to be near, so we can hope
|
|
#
?
May 3, 2019 08:35
|
|
|
Dude, forget all that. Look at the Ryzen 3 3300 with 6/12, 3.2GHz/4.0GHz. AMD is loving coming for Intel's balls.
|
|
#
?
May 3, 2019 09:04
|
|
|
SwissArmyDruid posted:Dude, forget all that. Look at the Ryzen 3 3300 with 6/12, 3.2GHz/4.0GHz. Yeah that's basically the 2600/2600X, only now could be priced around 100$ (so much for reselling my current 2600X if that's true, by the way  )If they pull this off they won this generation IMO, even though sheer staying power and market share will keep Intel in business no matter what (and probably thriving too - who cares if AMD turns out to have more cores, similar IPC and a hugely better price/performance ratio? Intel is bigger and better because it has the 1% top performing processor for 700$, also Pentium, also everybody else has it, also  )
|
|
#
?
May 3, 2019 09:12
|
|
|
Stickman posted:Those are just the same “leaked” specs that wccftech posted back in December, so I’m pretty sure that’s where the numbers came from. Oh wccftech  Even if Intel manages to keep the singlethread crown with the new R0 stepping, very few people would care if you can get a CPU with 10% lower singlethread performance but nearly double the core count at a similar price and power envelope. It does all sound a bit too good to be true but not entirely impossible considering Intel has been stuck on refined 14nm Skylakes for so, so long.
|
|
#
?
May 3, 2019 09:51
|
|
|
TorakFade posted:If they pull this off they won this generation IMO, even though sheer staying power and market share will keep Intel in business no matter what (and probably thriving too - who cares if AMD turns out to have more cores, similar IPC and a hugely better price/performance ratio? Intel is bigger and better because it has the 1% top performing processor for 700$, also Pentium, also everybody else has it, also Crying about mindshare when AMD is already outselling Intel 2:1 in the DIY market  And that's even with AMD being >20% behind Intel in both single-thread and multi-thread performance. If they can close up that gap to something a little more insignificant then they will start pulling some of the crowd who is more interested in performance than value. 1% behind and 50-100% more cores for the money is a completely different ballgame than 20% behind in single-threaded and a tie in multithreaded at equal pricing (8700K vs 2700X) or just 20% behind across the board at the high end (9900K). Nobody cares about margin-of-error stuff but the current gap is far beyond margin-of-error. Coffee Lake is actually a pretty solid response from Intel and that's why AMD has cut prices in response. If they were still trying to sell the 2700X at $530 like the 1800X was, there would be no reason to buy it over Coffee Lake. The 8700K would be a better performer at a cheaper price and the 9900K would be a much better performer at the same price. They priced the 2000 series where it needed to be to sell against the revised Intel lineups. But again, AMD is doing quite well in the mainstream DIY market, even if performance-focused enthusiasts are still chasing the faster processors. AMD has gone from "you would be outright stupid to buy that" to literally a majority of the DIY market in under two years. Turns out mindshare isn't really a thing, offer products with a compelling value or performance story and people will buy them. RTG should take notes. eames posted:Oh wccftech WCCF's numbers are a re-hash of an AdoredTV video that already has some major cracks in its credibility. All these parts were supposed to have launched at CES, and AMD has said Matisse won't have chiplet-based -G parts, although Adored is creatively re-interpreting their statement to mean that chiplet APUs have a different code-name. But at a minimum you can probably expect that AMD will follow the 2000-series pricing (eg 8-core for $300-350, 6-core for $200-250, etc), which effectively gives you an i9 at i7 pricing. I mean, they kinda have to, if all they offered was an i9 at i9 pricing then Intel already satisfied a lot of that market a year before they launched. Giving you an i9 at i5 pricing seems unnecessary. AMD wants decent margins too. But, who knows, all will (hopefully) be revealed in another month. Paul MaudDib fucked around with this message at 10:29 on May 3, 2019 |
|
#
?
May 3, 2019 09:57
|
|
|
TorakFade posted:3700X with 12/24 cores and 4.2/5.0Ghz is a wet dream, especially if it can sustain that boost on, say, 6 cores IMO the nice part about the a high boost clock is that it can be a good indicator (not a guarantee) of what you can expect a all core overclock could get to. Assuming you've got the cooling of course. Even if they end up using less power over all than Intel's 8 or 10 core parts at those speeds a 8+ core Zen2 will probably put out lots of heat with a all core overclock at ~5Ghz.
|
|
#
?
May 3, 2019 11:22
|
|
|
Paul MaudDib posted:Crying about mindshare when AMD is already outselling Intel 2:1 in the DIY market Do you have any sources on that? The only one I can find is this, https://www.extremetech.com/computing/281741-new-retail-data-shows-amd-outselling-intel-21 , and it says that it's only data for one single retailer in Germany. Hardly proof that AMD is outselling Intel 2:1 on the whole market. If that's what you're basing your sarcastic emoticons and bitterness on, well, better luck next time My firsthand experience, anecdotal of course but still, is that 90% of people I know and that are videogaming enthusiast have some sort of Intel CPU in their rig. And that's even for rigs built in 2019. From a few tech and videogaming forums I read, it seems that it is a common trend. Every "true" "gamer" has an Intel CPU and looks down on us poors with Ryzens, because we can't get to 120FPS in whatever new game came out yesterday, but only to 117FPS so their choice to spend 600$ on a 9900K is fully justified 
|
|
#
?
May 3, 2019 11:33
|
|
|
Pretty much the thing where the CPU really matters in gaming is doing CPU-encoded streaming and Intel 6C/6T are terrible versus the Ryzen 6C/12T.
|
|
#
?
May 3, 2019 11:57
|
|
|
Paul MaudDib posted:But at a minimum you can probably expect that AMD will follow the 2000-series pricing (eg 8-core for $300-350, 6-core for $200-250, etc), which effectively gives you an i9 at i7 pricing. I mean, they kinda have to, if all they offered was an i9 at i9 pricing then Intel already satisfied a lot of that market a year before they launched. Yeah, I'd expect the pricing structure to stay the same with the 12-16 core parts slotting above that. It's the non-x 8/16 part I'm most interested in, near 9900k performance at half the tdp 
Arzachel fucked around with this message at 12:03 on May 3, 2019 |
|
#
?
May 3, 2019 11:59
|
|
|
In my opinion the adored leak was a marketing/"product manager" type of spreadsheet from sometime in 2018. Those were the zen2 targets and not yet reality. AMD planned around Intel having 10nm+ online by that time. I expect prices to be bumped up a bit over it, too. The APUs are for the gen using zen2, which traditionally would be scheduled for early 2020. If intel's 10nm+ were out they'd have to scramble to release APUs. Intel has stagnated in the APU market due to their fab delays. Before then they were trading with AMD blow for blow. AMD knows 10nm APU isn't going to be a joke. Q3 is a reasonable timeline because Navi is ready to ship at that point. AMD also ran into hiccups somewhere. The prediction was it was with x570 chipset, but the reports from people with late engineering samples is zen2 isn't stable on x370/x470 motherboards yet (reported in late april). Whether this is an intentional strategy to avoid leaks or more work being required than they thought is unclear. Paul MaudDib posted:And that's even with AMD being >20% behind Intel in both single-thread and multi-thread performance. The majority of the IPC gap from zen+ is closed with doubling the width of the FPU. They also optimized branch prediction and the micro op architecture quite a bit. And are throwing a ton of cache in. They'll still be behind, but 9900k vs 2700x is a 15% single threaded clock gap at 5.1 vs 4.4. The ~5%-8% IPC+Latency combination is what causes most of the gaps larger than that. With AVX2 being the other cause in benchmarks. quote:If they can close up that gap to something a little more insignificant then they will start pulling some of the crowd who is more interested in performance than value. 1% behind and 50-100% more cores for the money is a completely different ballgame than 20% behind in single-threaded and a tie in multithreaded at equal pricing (8700K vs 2700X) or just 20% behind across the board at the high end (9900K). Nobody cares about margin-of-error stuff but the current gap is far beyond margin-of-error. quote:But at a minimum you can probably expect that AMD will follow the 2000-series pricing (eg 8-core for $300-350, 6-core for $200-250, etc), which effectively gives you an i9 at i7 pricing. I mean, they kinda have to, if all they offered was an i9 at i9 pricing then Intel already satisfied a lot of that market a year before they launched. 8/12 for R5/R7 pricing is really likely if they are still lagging in clocks. The leaks so far of actual samples are 4.5 all core with 4.7 stable, 4.8 unstable single threaded peak reported. However, that was a 4c 4+0 sample. So all core speed may drop a bit especially for chiplets with all 8 cores in use. The reality may be they don't even hit 4.7 single core, 4.5 all core. 6c R3 pricing sounds crazy until you realize the possible combinations. I suppose R3 could also be 2+2 at 4. Note, they may still call 6c R5, 8C R7. It's hard to say. This is in the realm of non-technical people. I feel more confident predicting core bumps for R5/R7. The possible layouts for 8c include 8+0 and 4+4, which means they can use defective samples that won't be great in higher lineups. Depending on how the IO die is, they'll either have to commit to one combination or can potentially freely use both. People are viewing using two chiplets as more expensive, but the reality is probably those are kind of junk and would get thrown at r5. This depends a lot on yields, too. 7nm is cheaper per-chiplet for AMD than zen and zen+ were. Even with the estimated 70% yields. People who think they can't do this due to price are living under a rock. They price everything aggressively except for chiplets that go into the highest end stuff. Their server strategy is to chase cloud vendors and similarly sized enterprises. That's where the margins are. Zen2 architecture allows them to have up to 12 cores without eating into those chiplets. 1950x has been on sale for as low as $450, 1700 $100 (out of stock for ages), 1700x $129.99 with $119.99 sales, 1600 everyday is price $50 right now, and R5 2600 is $119.99 right now with a game bundle you could resell for up to $70 when it first happened. Note, those prices include motherboard bundle at microcenter which is mildly dishonest and true at the same time. I'm predicting Intel still leads 5%-8% on many games due to clock speeds alone. Worst-case games may still be 12% or so. Provided the clocks I posted above are true. If they can't hit those, then it's even more likely they'll go for the core strategy because Intel will still lead by even more in things that aren't heavily multithreaded. I'm also ignoring new Intel stepping which could be another 2%-6%. Another trajectory is high initial prices and then huge slashes when intel gets its supply issues worked out in Q4. AMD isn't shy about charging Intel-tier prices when they have the better product, and their initial zen pricing was real high despite its shortcomings. So much speculation, so little credible information! I am excited for the actual announcements and real-world benchmarks of actual products. In the end I may end up buying Intel's new stepping and a second computer with zen2 to run as a home server/workstation. Khorne fucked around with this message at 19:42 on May 3, 2019 |
|
#
?
May 3, 2019 13:04
|
|
|
Khorne posted:8/12 for R5/R7 pricing is really likely if they are still lagging in clocks. The leaks so far of actual samples are 4.5 all core with 4.7 stable, 4.8 unstable single threaded peak reported. However, that was a 4c 4+0 sample. So all core speed may drop a bit especially for chiplets with all 8 cores in use. The reality may be they don't even hit 4.7 single core, 4.5 all core. Just my 2¢, but I think the single threaded delta between Ryzen 2000 and Ryzen 3000 is going to be large. I think 4.7Ghz all core is attainable actually, with golden samples hitting 4.8ghz, and with the estimated 13% increase in IPC that puts them head and shoulders over the most tuned 2000 series overclocks. The massively reduced prices of Ryzen 2000 seem to indicate this, they're fully clearing the channels as fast as reasonably possible because this stuff won't sell after, especially if price per core only slightly increases. 4.7-4.8Ghz and a 13% increase in IPC means they should be equal to if not beat Intel if latency shenanigans don't gently caress them. I think a 4.7Ghz all core clock is reasonable since the chip being tested was a 4+0 as you said, which kind of indicates a defective chip. In the Ryzen 1000 series, it wasn't uncommon for early R5 1600s and lower to only hit 3.8Ghz despite the real cap sitting around 4.0Ghz, but as yields improved this became more rare. It's possible there will be much higher variance on 7nm DUV, so there may be some real binning between skus now where the X part is legitimately better. Also I'll call that the low end remains the exact same: Athlons are still 2C/4T, R3s are 4C/4T, R5s are 4C/8T and 6C/12T, R7s are 8C/16T and 12C/24T. 16C/32T will occupy the new R9 branding and will release later because demand for perfect dies is going to be immense with EPYC2 releasing.
|
|
#
?
May 3, 2019 14:44
|
|
|
I sort of want to revise my post now. There's actually a decent chance 6c/8c will be single chiplet and 12 will be 6+6. It's very hard to know the approach without actual numbers on defects, and unfortunately I lack a crystal ball. I'm trying to be pessimistic about clock speeds and IPC increases, EmpyreanFlux! I'd rather plan for the worst and have easy choices once it gets announced. Khorne fucked around with this message at 19:25 on May 3, 2019 |
|
#
?
May 3, 2019 16:31
|
|
|
Can they do anything with mismatched core counts now? I thought they were limited to identical configurations 4+4 / 6+6 / 8+8
|
|
#
?
May 3, 2019 16:50
|
|
|
Khorne posted:I sort of want to revise my post now. There's actually a decent chance 6c/8c will be single chiplet and 12 will be 8+4. It's very hard to know the approach without actual numbers on defects, and unfortunately I lack a crystal ball. I get where you are coming from, and I still think 4.5Ghz is pretty decent with the IPC uplift and they've already hit that target. I just think if they can even achieve 4.8Ghz unstable single core on a defective 4+0 chip then I feel like it's within range for golden chips which usually hit similar clocks at 500mV less. "Wait for Computex" Risky Bisquick posted:Can they do anything with mismatched core counts now? I thought they were limited to identical configurations 4+4 / 6+6 / 8+8 I don't think so? I'm fairly positive they have to be mirrored to function properly, or even at all.
|
|
#
?
May 3, 2019 16:59
|
|
|
I'm personally being super pessimistic because more often than not with CPUs, that tends to be right. The Adored stuff isn't even worth looking at imo. The 7nm chiplet is still expensive to make, and is still only 8 cores. I fully expect that the mainstream core counts stay the same, with 8 being the high end and 6 being mainstream. Lisa Su hinted that 2 chiplet parts wont be available at launch, and I don't think 12 and 16 core parts will be cheap at all once they show up. 2950X is $849, and Intel's cheapest 16 core is $1,700. I don't really see a business reason for them to undercut the current pricing that much if the performance is there, as mega-core needing users are still a minority. I fully expect that if these parts are 1:1 performance competitive with Intel, they will be price matched to Intel too. Frankly, I'm also pessimistic about big IPC gains. The Zen2 vs 9900k comparison was in Cinebench... a program thats already very Zen friendly. It showed a 12% increase over the 2700x with the same thread count. 2700x sits around 4ghz all core boost in cinebench, if you assume there is no IPC increase, then Zen 2 needed to be clocked at ~4.5ghz to pull that off... which doesn't sound too crazy to me for a part with a die shrink. AMD did talk about some new architectural changes beyond the chiplet level: better branch prediction, better prefetcher, lots of cache improvements etc. If you combine a modest single digit IPC increase with a clock speed boost from the die shrink, I think youre in the window to what we saw. IPC is expensive and hard, clock speed is likely the cheaper path to that performance. Of course, it seems like AMD has more in the tank with the power budget, which I think they will need to go after the 9900k in games, but I'm just talkin about the Cinebench example we've seen. The best part of being pessimistic like this is if AMD blows it out of the water? Great! You get awesome new stuff to buy. But if they don't? Your predictions were right and you can tell yourself that you are a GENIUS. Cygni fucked around with this message at 17:09 on May 3, 2019 |
|
#
?
May 3, 2019 17:07
|
|
|
Chiplets that are 80mm^2 will be a lot less expensive to make than the 9900k at 177mm^2. That’s not too hard to figure out just using one of those chip-yield calculators
|
|
#
?
May 3, 2019 17:19
|
|
|
Lube banjo posted:Chiplets that are 80mm^2 will be a lot less expensive to make than the 9900k at 177mm^2. That’s not too hard to figure out just using one of those chip-yield calculators Defect rates and per-wafer foundry charges are much higher on 7nm. Cost per yielded transistor is going to be equal or higher, and that already takes into account the fact that you get more dies per wafer. You can already see that on Radeon VII... it's more than twice the cost of Vega 56 despite only having 6% more transistors and $150 of extra memory, and despite being nominally more than 30% smaller than Vega 10. A lot of that cost is 7nm just being really loving expensive. The gain from the chiplet strategy is that you have fewer transistors per chiplet since you've moved the IO to a separate die... although compared to a traditional monolithic die, that will be somewhat offset by the need for more Infinity Fabric PHYs (it is a straight gain compared to Zeppelin where the Infinity Fabric is already baked into the die though). e: which is not to say that Zen2 will be massively more expensive to produce, but once you factor in 7nm being equal to slightly more expensive per-transistor, chiplets are really an incremental price savings over Zeppelin, which was probably roughly as expensive to produce as the 9900K. They're both using 8C dies in a monolithic package. Intel just has been running massive margins... think about them charging $350 for a 4-core processor on a mature third-gen 14nm process, performance aside the 7700K was unquestionably a massive ripoff. Of course they also have to factor in the costs of developing their own fabs too but even still. The chiplets let you sidestep what would have otherwise been a huge price increase when you move to 16C, but they aren't themselves a huge cost reduction over previous 8C implementations. And there are some minor costs too, the IO die and interposer packaging isn't free either. Paul MaudDib fucked around with this message at 18:00 on May 3, 2019 |
|
#
?
May 3, 2019 17:29
|
|
|
With the space to run two chiplets, I suspect we'll see some creative layouts pretty early out of the gate to maximize returns. If you have a bunch of chiplets where only 4 cores can clock high consistently, why wouldn't you slap two of those together in to an 8c processor package and sell it as an enthusiast part with a higher price and TDP?
|
|
#
?
May 3, 2019 18:03
|
|
|
Are there really that many parts with great clocks but lots of broken cores? I was under the impression that both defects and bad clocks correlated pretty well.
Paul MaudDib fucked around with this message at 20:01 on May 3, 2019 |
|
#
?
May 3, 2019 18:06
|
|
|
Defects in the core and L1/2 cache do happen, its part of why Intel's Xeon binning is all over the map with their monolithic designs biting them. Rates are all proprietary, but the proof is in the absolute mess of package configurations they're forced to ship. I mean for christ's sake: https://ark.intel.com/content/www/us/en/ark/products/series/91287/intel-xeon-processor-e5-v4-family.html Their core density, base clocks, and turbo clocks are absolutely all over the map and it gets even worse if you drill in to some of their white papers for the matrix on which SKUs are able to do what turbo clocks on all/half/4/2/1 cores. This is all a result of working around manufacturing defects and working the available part to maximize core count and clock as best you can. AMD isn't accommodating defect rates at that scale because of the chiplet architecture, but I assure you it still happens. Even deactivating every other core on two marginal chiplets and slapping them in to the same package is going to buy you more thermal headroom to hit turbo clocks on because of the physical geometry of how they are laid out.
|
|
#
?
May 3, 2019 18:17
|
|
|
EmpyreanFlux posted:I don't think so? I'm fairly positive they have to be mirrored to function properly, or even at all. I thought it was just to ensure consistent performance within an SKU? Otherwise you'd end up with wildly different latencies depending on the physical configuration of the chips, or an inpeneratrable SKU matrix for all the possible configurations.
|
|
#
?
May 3, 2019 19:04
|
|
|
Risky Bisquick posted:Can they do anything with mismatched core counts now? I thought they were limited to identical configurations 4+4 / 6+6 / 8+8 
|
|
#
?
May 3, 2019 19:40
|
|
|
Would AMD switch to the new TSMC 6nm for their Zen 2 or just wait until their Zen 3 (Ryzen 4000?) with a different process? Oh god, I am so confused about everything. Is Zen 3 to Zen 2 as Ryzen 2000 series were to Ryzen 1000? TSMC saying a bunch of their customers already switched over to the 6nm process from their 7nm cause it was easy to do so makes me wonder if AMD would do the same, but not include it as part of their tick-tock cycle. But I also imagine that "easy to carry over" is a relative term and still require lots of resources to do so, especially so with a CPU. Also I expect single-core boost to be 4.7 GHz on their fastest chip. Reaching parity with intels current but nothing more if any. But God I sooo do hope I am wrong.
|
|
#
?
May 3, 2019 19:59
|
|
|
Zen2 is already in sampling and they're not changing process in the runup to production. I would expect them to do a Zen2+ gen refresh on the new process though since refactoring is minimal and it would give them some amount of performance bump while they're prepping zen3.
|
|
#
?
May 3, 2019 20:08
|
|
|
Just checked the press release. It says risk production for N6 starts Q1/2020.
|
|
#
?
May 3, 2019 20:42
|
|
|
Arzachel posted:You still need new masks so probably not unless they decide to do a refresh. I would expect Vermeer and Renoir to be fabbed on N7+ or N5 depending on availabilty since common tooling won't matter for new designs. Nthing the "FX series chips aren't worth free". Doesn't help that they were bad so AMD motherboards during that period were garbage-tier at best with all sorts of flakiness that same-priced intel boards didn't. I "won" one in a giveaway and regret not instantly flipping it on ebay. Or into a bin.
|
|
#
?
May 3, 2019 23:08
|
|
|
The A10 I rescued from an ewaste event is an excellent MadVR/emulator thing, and still played modern stuff alright when I threw a 1080 in it after my 4790k board died, that's my non-histrionic Bulldozer story
|
|
#
?
May 3, 2019 23:25
|
|
|
Setzer Gabbiani posted:The A10 I rescued from an ewaste event is an excellent MadVR/emulator thing, and still played modern stuff alright when I threw a 1080 in it after my 4790k board died, that's my non-histrionic Bulldozer story I don't like bulldozer chat. We have ryzen now and have for a few years! Khorne fucked around with this message at 23:32 on May 3, 2019 |
|
#
?
May 3, 2019 23:30
|
|
|
Khorne posted:Bulldozer is better than nothing and better than most things that came before it. The problem is what came at the same time. Of the "construction" cores only Excavator managed to consistently beat K10 (the previous AMD microarchitecture) on IPC. Bulldozer, Piledriver, and Steamroller were all literally worse than K10:  For reference, here is the chart for Intel CPUs over roughly the same time period:  (Images stolen from Reddit.)
|
|
#
?
May 3, 2019 23:39
|
|
|
Mr.Radar posted:Of the "construction" cores only Excavator managed to consistently beat K10 (the previous AMD microarchitecture) on IPC. Bulldozer, Piledriver, and Steamroller were all literally worse than K10:  K8 90nm has better IPC than Bulldozer.
|
|
#
?
May 4, 2019 03:45
|
|
|
Right ends of those charts really illustrate how little difference there is ghz for ghz. And also why a high ghz zen 2 part would be so devastating.
|
|
#
?
May 4, 2019 05:23
|
|
|
The difference is very small in cinebench, but in some other stuff (especially games) Intel still has quite a significant IPC lead. Even if Zen 2 doesn't seem to show much IPC gain where it's already close, AMD has probably focused on improving the areas where they lag behind since those are the obvious low hanging fruit. As long as the CPUs are powerful enough to last through the next console generation (and the high end SKUs probably are), it's very likely Zen 2 turns out to be a very good buy. Probably the biggest thing that will hold back sales is the fact that even Nvidia doesn't have appealing GPUs to sell right now, so a lot of people will probably delay upgrades until the full system looks more promising. K8.0 fucked around with this message at 05:40 on May 4, 2019 |
|
#
?
May 4, 2019 05:37
|
|
|
I'm pretty curious about the performance of AMD hardware for next gen consoles - I'm guessing 4K60 is out of the question?
|
|
#
?
May 4, 2019 06:45
|
|
|
shrike82 posted:I'm pretty curious about the performance of AMD hardware for next gen consoles - I'm guessing 4K60 is out of the question? Lol a 2080 ti can't guarantee 4k60.
|
|
#
?
May 4, 2019 06:47
|
|
|
consoles sometimes still struggle with 1080p30 at graphics quality people expect these days, 4k60 isn't going to happen until the mid/late 2020s, probably.
|
|
#
?
May 4, 2019 07:43
|
|
|
|
| # ? May 29, 2024 10:33 |
|
|
I'm thinking dynamic resolution scaling will come into play again in the next-gen consoles where the rendering resolution dynamically adjusts depending on the workload and target framerate with maybe 1080p60 being the floor (but still up-scaled to 4K) and native 4K60 being the ceiling with only games with simpler graphics hitting native 4K60. I just wish more PC games supported dynamic resolution scaling.
|
|
#
?
May 4, 2019 08:08
|
|
