Cant wait for zen 3000 series the early engineer sample cpu from zen 2 runs at 4gig base 4.5gig boost.Remember that samples for zen1 run at base 3gig and boost 3.4 and we finally got 1800x good times!!
^ RUMOR land...rumors of such need to be taken with a MASSIVE load of salt, way too easy for them to screw around with numbers to make product look amazing when "not so good" or make look not so good when amazing, certainly could be a VERY early sample of first yield results etc etc etc.
would be nice one day (doubt ever happens) that so called "leaks" are 100% veted results from the makers of the product and the product alone with results that are truly unbiased to show them as they ARE not as the cherry pick the results to make them SEEM they might be (kind of like that whole marketing BS where they can say 100% beef which only means at least .000000000001% has to be 100% beef, the rest can be whatever else they want it to be, but, as long as even the smallest portion is 100% beef "it is legal to claim it)
I think that is the problem with marketing in general (just like media these days) loads of BS just to make the sale instead of saying and showing like it is (warts and all) ^.^
Awesome the Dr Su will be the host for CES event for 2019, at least SOME of the world is paying attention to AMD and the massive turn around they have done in 2016/2017, shame their stock price has not jumped anywhere close to as high as it should have unfortunately.
stupid analysts and shortsellers/pumpers not help anything
When you're in the powerful position (like Intel) the "leaks" are mainly BS that's meant to give people "on the fence" some reason to wait it out and not immediately order something good from the competition. It's just to buy time and rain on the other guy's parade.
When you're the "underdog" (like AMD) leaks have to be close to reality because you can't afford any flops or disappointments.
Either way they're like sex tapes, mostly leaked intentionally to get attention, not really accidentally.
Considering Vega 7nm should have shipped by then, I really hope not. Although I'm not all that confident about Navi in 2019.
Maybe a 12nm Vega with GDDR? Lop the Ryzens off the Zhongshan Subor console chip and you have a Vega24 with 256 bit GDDR5 RAM. Maybe even double the Vega cores. Of course, that would largely assume that AMD needs GoFlo wafers, and nobody's leaked the details of what AMD owes GoFlo after they threw in the towel.
Vega 7nm is aimed at servers/compute farms. I highly doubt AMD is going to go another year without an architectural change. I think we'll see 7nm Navi and then a 7nm Ryzen/Navi APU for the PS5 holiday season 2019
Not Navi, Vega 20 at 7nm, strictly for the professional market. I don't think Navi will be released before Q2, maybe even Q3 2019. Bear in mind though that Navi will still be limited by the GCN design, i.e. it will have a maximum of 4096 shader cores. So its real advantage will be the higher power efficiency and thus higher clocks due to the 7nm node, a possible switch to PCIe 4.0, and perhaps HBM3 memory, unless it is crazy expensive.
Arcturus, on the other hand, will AMD's first post-GCN architecture. Now that dual die consumer Navi was canned it's possible that Arcturus will be released in 2020. It is more realistic it will slip into 2021 though.
Can someone please ask Ms. Su why she is harming AMD shareholders, and circumventing a U.S. government ban, by creating a JV that transfers sensitive U.S.-engineered chip technology to China?
I'm not kidding please read the links below.
• In August 2015, the U.S. barred Intel (and implicitly AMD) from selling high-end server chips to a few (but not all) Chinese customers due to national security concerns. Intel complied. https://www.pcworld.com/article/2908692/us-blocks-... • A few months later, China went to Lisa Su to get basically the same chip from AMD via a complicated joint venture that complied with the letter of the law -- because the chip would not be sold by a U.S. company but would rather be sold by the Chinese JV (so no need to ask the U.S. for permission to export it to Chinese customers). https://www.forbes.com/sites/patrickmoorhead/2016/... • A month ago, China started producing these Epyc server chip clones. As this article explains, they have a different name but are EXACTLY Epyc chips. https://www.tomshardware.com/news/china-zen-x86-pr... • The U.S. government is due to publish a report on Chinese theft of U.S. semiconductor IP that could name and shame AMD. "ASSESSMENT OF THE U.S. INTEGRATED CIRCUIT DESIGN AND MANUFACTURING INDUSTRY" https://www.reddit.com/r/AMD_Stock/comments/87z5h9... • Based on the timing, it's clear the ONLY reason this JV was set up was to circumvent the government's ban. • These chips will be in the hands of those very customers the U.S. government said couldn't have them. • They will be taking market share from Epyc and only paying AMD a tiny royalty, instead of the large margins it would have made on Epyc.
" paying AMD a tiny royalty, instead of the large margins it would have made on Epyc."
Well that's just like consoles, AMD gets huge volume but tiny margins. They seem OK with that risk vs reward. In regards to the security issue, AMD is an international company, comprised of ATI which was Canadian, backed by middle east money Mubadala Development Company (which even last year bought 8% of AMD and kept them afloat last 5 years) and dev teams in India and USA. They're keeping the world in harmony.
Lol well the US Government should have stepped in long ago and made sure Amd had a fair share of the server market long ago. So now Amd has to do it to survive
It's a global company, with global clients and global shareholders. It doesn't have allegiance to any patriots of any particular single nation. I hope national patriot-politics stays out AnandTech, it's one of the few sane places left so far in regards to this.
AMD headquarters is in California and likely has logic in CPU and system which has security system which United States limits to foreign companies. If AMD has violated any of these laws with China, this is major issue and needs to be rectified immediately. This is not a political issue - but violation of trade laws This is far worst than any issue related to CPU cloning and such.
Companies that were actually violating agreements like ZTE just came back from the grave, these things don't mean anything anymore, just cut Trump a check and move on.
This is not related to who is President at the current time and if the President allows our security to be purchase by others than that President should not be allow to be off. But if President stops the China abuse of the law than he is doing his job.
The Chinese chips almost certainly do not contain the technology the U.S. government was concerned about sharing (otherwise the circumvention would have been an immediate issue since the U.S. government frequently enforces by 'spirit of the law' standards in such matters when the letter of the law is abused).
The Chinese CPU includes the base core IP and instruction set support. The ARM coprocessor is also likely included considering it is largely open knowledge, anyway, though it is likely that AMD has not shared their own sensitive IP regarding their PSP implementation - the Chinese would want their own implementation anyway to prevent U.S. spying.
Encryption aka FTS on Intel CPUs/Intel chipsets in laptops? AMD doesn't have it, hence it is not sensitive to have the Chinese getting it. EPYC doesn't have some instruction sets as well. But yeah, anything strategically sensitive would've been banned so far, which means AMD has no chance of getting any of its EPYC chips into federal contracts. It may explain why their stocks were battered on Wall Street recently - regardless Intel weaknesses, AMD has no chance of getting a financially meaningful market share outside of consumer electronics.
I think the blame her is a little bias. The board of directors are the body that is responsible for your accusations. I don't think Dr Su gives a shit about the shareholders. She is an engineer. Her focus is on technology, not profits.
"ask Ms. Su why she is . . . circumventing a U.S. government ban"
Because that's what capitalists are supposed to do. The government throws up roadblocks, and the role of the entrepreneur is to save the market from the government.
We should all be thankful that AMD was able to circumvent the US government here. The unfortunate fact is that Intel wasn't able to.
They are just catching up to INTEL who created their own Chinese JV with Spreadtrum starting TWO YEARS EARLIER in 2014 to license AMD64 IP from Atom Airmont and customize in order to create AMD64 chips like http://spreadtrum.com/en/SC9861G-IA.html .
You need to educate you PATRIOT behind into these issues and find out what company sold itself and American interests and licensed not only its own IP but also AMD64 IP.
She deserves the keynote. She's really smart, comes across really well in interviews, and AMD has executed nearly perfectly under her watch. They've managed to become very competitive with Intel, and not totally throw in the towel with Nvidia, despite being a tiny fraction of their size (much smaller, if you go by market cap as of say, the beginning of this year).
And she's an engineer-turned-exec, rather than an exec who happens to be leading a chip company.
I'm hoping we get some good news about Navi, since Vega wasn't really that competitive with (in terms of performance/watt) with Pascal, let alone Turing (plus, the exclusive use of HBM2 has meant prices have stayed astronomical, even as the mining craze has allowed prices to moderate for other GPUs).
I wonder if Nvidia is really going to release a 7nm chip so close on the heels of the 20-series. They're going to be using TSMC's 7nm (not 7nm+), so that points to a next year release. People that buy the 2080 RTX would be piiiiiissed
I think we are 2 years away from a launch. I don't see Nvidia trying anything since they have the enthusiasm market. AMD is looking to disrupt mainstream and impose their architecture as the main game development architecture. Performances are going to come with it.
That's my question, too. My thinking is that's why you should skip the 20xx cards if you've got a 10xx one (unless, say, you have a 1050 and now have enough money for say, a 2080 Ti or something) - performance is better, but its not RADICALLY better, and not that many games will take advantage of ray tracing in the next 2-ish years (and ones that do will probably have it "retrofitted" on to existing game art). It's the games that are just STARTING development now where they'll be thinking of ray tracing from the outset that'll see the real advantage, and by that time, to get max settings, you'll need one of Nvidia's 21xx (or whatever) series built on 7nm.
Of course, if you can look at the $1500 cost of a top-of-the-line GPU as money you might find in your couch, then go ahead.
Why do I get the feeling they're gonna pull a similar stunt to nVidia, and just release something marginally faster, while raising the price a whole lot?
AMD has neither market nor mindshare dominance, so I find this unlikely. There's every reason to expect Zen 2 to be a major step up (10-15% IPC improvement) given that it's the architecture's first major refresh. Plus, it's coupled with a new process node.
Navi is the big question mark for me. Can AMD's graphics division regain competitiveness with Nvidia?
At 7nm I think Navi will be competitive performance/watt with the 10 series Nvidia cards. Not sure if Arcturus will be a shrink or new architecture, but hopefully that will be competitive with their 30 series cards
In their last published roadmap, next gen (Arcturus?) i slated for 2020 on 7nm+, and they have previously announced that Navi (GCN 6) is the last GCN-architecture.
Actually, since about 30% of Nvidia's chip is targeted at ray tracing that few game use, AMD has an incredible opportunity to pack their GPU to surpass Nvidia, both in performance and power usage, for all games that don't use ray tracing. Plus: AMD prices.
Honnestly, Navi was a major concern for Lisa after Koduri Vega. I really doubt we should be worried about it at all. It cannot be worst than what Vega 64 was.
I think we are going to be surprised. I will buy one for my mini-itx system.
Vega 64 isn't *that* bad. It fell about 30% short of performance in gaming workloads than was intended mainly due to some features which AMD, for whatever reason, didn't get working in the Gaming driver or production silicon. The whole "Vega 10 is terrible" thing is tiring. Gaming performance is acceptable for most things and Vega 10 has other applications too, where it competes more favourably with the bigger Pascal chips.
Yes, why do you feel that? Given that AMD has delivered big time at least on CPUs, vastly increasing bang for the buck and essentially rebooting the x86 market. Even on GPUs where the Vega product didn't meet our hopes (whether or not those were due to AMD statements) they weren't raising prices.
CPUs aren't really comparable to GPUs. I've thought about upgrading my i2500k sandybridge from 2011, in 7 years we've only got a 25% improvement for Ryzen:
Compare that to how GPUs have sped up from 2011 to today, its something like 2.5-3X.
Sure GPUs have different bottlenecks and aren't single threaded but that just re-iterates the point that they can't be compared. AMD CPU performance should be compared independently of their GPUs. Intel was stagnant for 5 years allowing AMD to catchup, that's not the case for GPUs.
And you know what else? I assume 25% improvement is in single thread application. And most of single thread application is games. Until now, games still bottlenecked by GPU, not the CPU.
I think we hit a wall where single threaded performance is already good... Even Intel can't go pass 5Ghz.
You're complaining about single core performance in CPUs and comparing that to the massively multicore improvements in GPUs? Dude, those concepts are from 2 different planets. Here's what you should be comparing:
The multicore improvements in GPUs (3-4x) vs. the *multicore* improvements between the i7-2600K and TR 2990WX. 612 vs 5000 in Cinebench. It's something like 8x.
OK, maybe that's cheating. How about 2600K vs 9900K (leaked benchmarks if we can trust them). 612 vs. 2166. It's something like 3.5x, same as the GPUs.
Performance marches ahead apace for both CPUs and GPUs, thanks to a bit of pushing from AMD.
I am so excited about this. I have budgeted to buy the top-of-the-line 3000 series chip for AM4 which I intend to drop in my X470 motherboard. (based on pricing of the 2700X).
2019 should be a great year for AMD CPU's as they will, for the first time in a long time be on an even playing-field with Intel in per core performance and an advantage in process technology. I am even hopeful for the GPU side of things, too, as Nvidia is seemingly happy to use 12nm technology for Turing, I think we will see a 7nm RX card be offered against those, potentially competing favourably in Perf/watt as well as the traditional perf/dollar area. The former being an area where I really want to see RTG make headway in.
I read four layers of EUV lithography for the 7nm process somewhere: When you look at the huge list of challenges EUV has been adding, that is quite simply awsome.
We’ve updated our terms. By continuing to use the site and/or by logging into your account, you agree to the Site’s updated Terms of Use and Privacy Policy.
Facebook
Twitter
RSS
50 Comments
Back to Article
DazFG - Wednesday, October 3, 2018 - link
"The keynote will take place on January 8, 2018." - do you mean 2019?Ian Cutress - Wednesday, October 3, 2018 - link
That's the one.FreckledTrout - Wednesday, October 3, 2018 - link
88 mph here I come.Samus - Wednesday, October 3, 2018 - link
Intel did the 2018 keynote, and my God was it awkward (it was days after the Spectre/Meltdown vulnerabilities rocked the industry.)I look forward to seeing Dr Su on stage, I've never seen her speak publicly!
siberian3 - Wednesday, October 3, 2018 - link
Cant wait for zen 3000 series the early engineer sample cpu from zen 2 runs at 4gig base 4.5gig boost.Remember that samples for zen1 run at base 3gig and boost 3.4 and we finally got 1800x good times!!Dragonstongue - Wednesday, October 3, 2018 - link
^ RUMOR land...rumors of such need to be taken with a MASSIVE load of salt, way too easy for them to screw around with numbers to make product look amazing when "not so good" or make look not so good when amazing, certainly could be a VERY early sample of first yield results etc etc etc.would be nice one day (doubt ever happens) that so called "leaks" are 100% veted results from the makers of the product and the product alone with results that are truly unbiased to show them as they ARE not as the cherry pick the results to make them SEEM they might be (kind of like that whole marketing BS where they can say 100% beef which only means at least .000000000001% has to be 100% beef, the rest can be whatever else they want it to be, but, as long as even the smallest portion is 100% beef "it is legal to claim it)
I think that is the problem with marketing in general (just like media these days) loads of BS just to make the sale instead of saying and showing like it is (warts and all) ^.^
Awesome the Dr Su will be the host for CES event for 2019, at least SOME of the world is paying attention to AMD and the massive turn around they have done in 2016/2017, shame their stock price has not jumped anywhere close to as high as it should have unfortunately.
stupid analysts and shortsellers/pumpers not help anything
close - Wednesday, October 3, 2018 - link
When you're in the powerful position (like Intel) the "leaks" are mainly BS that's meant to give people "on the fence" some reason to wait it out and not immediately order something good from the competition. It's just to buy time and rain on the other guy's parade.When you're the "underdog" (like AMD) leaks have to be close to reality because you can't afford any flops or disappointments.
Either way they're like sex tapes, mostly leaked intentionally to get attention, not really accidentally.
NikosD - Thursday, October 4, 2018 - link
Best comments from "close" as always at Anandtech's siteNICOXIS - Wednesday, October 3, 2018 - link
Hopefully this will be about Navi rather than Vega 7nmwumpus - Wednesday, October 3, 2018 - link
Considering Vega 7nm should have shipped by then, I really hope not. Although I'm not all that confident about Navi in 2019.Maybe a 12nm Vega with GDDR? Lop the Ryzens off the Zhongshan Subor console chip and you have a Vega24 with 256 bit GDDR5 RAM. Maybe even double the Vega cores. Of course, that would largely assume that AMD needs GoFlo wafers, and nobody's leaked the details of what AMD owes GoFlo after they threw in the towel.
ishould - Wednesday, October 3, 2018 - link
Vega 7nm is aimed at servers/compute farms. I highly doubt AMD is going to go another year without an architectural change. I think we'll see 7nm Navi and then a 7nm Ryzen/Navi APU for the PS5 holiday season 2019Santoval - Wednesday, October 3, 2018 - link
Not Navi, Vega 20 at 7nm, strictly for the professional market. I don't think Navi will be released before Q2, maybe even Q3 2019. Bear in mind though that Navi will still be limited by the GCN design, i.e. it will have a maximum of 4096 shader cores. So its real advantage will be the higher power efficiency and thus higher clocks due to the 7nm node, a possible switch to PCIe 4.0, and perhaps HBM3 memory, unless it is crazy expensive.Arcturus, on the other hand, will AMD's first post-GCN architecture. Now that dual die consumer Navi was canned it's possible that Arcturus will be released in 2020. It is more realistic it will slip into 2021 though.
AmericanPatriot - Wednesday, October 3, 2018 - link
Can someone please ask Ms. Su why she is harming AMD shareholders, and circumventing a U.S. government ban, by creating a JV that transfers sensitive U.S.-engineered chip technology to China?I'm not kidding please read the links below.
• In August 2015, the U.S. barred Intel (and implicitly AMD) from selling high-end server chips to a few (but not all) Chinese customers due to national security concerns. Intel complied. https://www.pcworld.com/article/2908692/us-blocks-...
• A few months later, China went to Lisa Su to get basically the same chip from AMD via a complicated joint venture that complied with the letter of the law -- because the chip would not be sold by a U.S. company but would rather be sold by the Chinese JV (so no need to ask the U.S. for permission to export it to Chinese customers). https://www.forbes.com/sites/patrickmoorhead/2016/...
• A month ago, China started producing these Epyc server chip clones. As this article explains, they have a different name but are EXACTLY Epyc chips. https://www.tomshardware.com/news/china-zen-x86-pr...
• The U.S. government is due to publish a report on Chinese theft of U.S. semiconductor IP that could name and shame AMD. "ASSESSMENT OF THE U.S. INTEGRATED CIRCUIT DESIGN AND MANUFACTURING INDUSTRY" https://www.reddit.com/r/AMD_Stock/comments/87z5h9...
• Based on the timing, it's clear the ONLY reason this JV was set up was to circumvent the government's ban.
• These chips will be in the hands of those very customers the U.S. government said couldn't have them.
• They will be taking market share from Epyc and only paying AMD a tiny royalty, instead of the large margins it would have made on Epyc.
webdoctors - Wednesday, October 3, 2018 - link
" paying AMD a tiny royalty, instead of the large margins it would have made on Epyc."Well that's just like consoles, AMD gets huge volume but tiny margins. They seem OK with that risk vs reward. In regards to the security issue, AMD is an international company, comprised of ATI which was Canadian, backed by middle east money Mubadala Development Company (which even last year bought 8% of AMD and kept them afloat last 5 years) and dev teams in India and USA. They're keeping the world in harmony.
FreckledTrout - Wednesday, October 3, 2018 - link
Lol this person posted this plus a ton more under a different name and the mods removed it. Hopefully they can keep removing this nonsense.iwod - Wednesday, October 3, 2018 - link
Trolling? Or you will need to reread your first link.Jleppard - Wednesday, October 3, 2018 - link
Lol well the US Government should have stepped in long ago and made sure Amd had a fair share of the server market long ago. So now Amd has to do it to surviveJleppard - Wednesday, October 3, 2018 - link
Also you have no idea of what amd is getting from thier IP.halcyon - Wednesday, October 3, 2018 - link
It's a global company, with global clients and global shareholders. It doesn't have allegiance to any patriots of any particular single nation.I hope national patriot-politics stays out AnandTech, it's one of the few sane places left so far in regards to this.
HStewart - Wednesday, October 3, 2018 - link
AMD headquarters is in California and likely has logic in CPU and system which has security system which United States limits to foreign companies. If AMD has violated any of these laws with China, this is major issue and needs to be rectified immediately. This is not a political issue - but violation of trade laws This is far worst than any issue related to CPU cloning and such.webdoctors - Wednesday, October 3, 2018 - link
Companies that were actually violating agreements like ZTE just came back from the grave, these things don't mean anything anymore, just cut Trump a check and move on.HStewart - Wednesday, October 3, 2018 - link
This is not related to who is President at the current time and if the President allows our security to be purchase by others than that President should not be allow to be off. But if President stops the China abuse of the law than he is doing his job.looncraz - Wednesday, October 3, 2018 - link
The Chinese chips almost certainly do not contain the technology the U.S. government was concerned about sharing (otherwise the circumvention would have been an immediate issue since the U.S. government frequently enforces by 'spirit of the law' standards in such matters when the letter of the law is abused).The Chinese CPU includes the base core IP and instruction set support. The ARM coprocessor is also likely included considering it is largely open knowledge, anyway, though it is likely that AMD has not shared their own sensitive IP regarding their PSP implementation - the Chinese would want their own implementation anyway to prevent U.S. spying.
HStewart - Wednesday, October 3, 2018 - link
If CPU has anything do with encryption and such then it likely US to be concerned - also almost anything with telecommunications.Ananke - Wednesday, October 3, 2018 - link
Encryption aka FTS on Intel CPUs/Intel chipsets in laptops? AMD doesn't have it, hence it is not sensitive to have the Chinese getting it. EPYC doesn't have some instruction sets as well. But yeah, anything strategically sensitive would've been banned so far, which means AMD has no chance of getting any of its EPYC chips into federal contracts. It may explain why their stocks were battered on Wall Street recently - regardless Intel weaknesses, AMD has no chance of getting a financially meaningful market share outside of consumer electronics.Samus - Wednesday, October 3, 2018 - link
I think the blame her is a little bias. The board of directors are the body that is responsible for your accusations. I don't think Dr Su gives a shit about the shareholders. She is an engineer. Her focus is on technology, not profits.oRAirwolf - Wednesday, October 3, 2018 - link
Apparently the Russian troll farms have started invading Anandtech.PeachNCream - Thursday, October 4, 2018 - link
Yup and because people post before they think, it works disappointingly well on them.Mikewind Dale - Friday, October 5, 2018 - link
"ask Ms. Su why she is . . . circumventing a U.S. government ban"Because that's what capitalists are supposed to do. The government throws up roadblocks, and the role of the entrepreneur is to save the market from the government.
We should all be thankful that AMD was able to circumvent the US government here. The unfortunate fact is that Intel wasn't able to.
sgeocla - Wednesday, October 10, 2018 - link
They are just catching up to INTEL who created their own Chinese JV with Spreadtrum starting TWO YEARS EARLIER in 2014 to license AMD64 IP from Atom Airmont and customize in order to create AMD64 chips like http://spreadtrum.com/en/SC9861G-IA.html .You need to educate you PATRIOT behind into these issues and find out what company sold itself and American interests and licensed not only its own IP but also AMD64 IP.
Atari2600 - Wednesday, October 3, 2018 - link
Hopefully a sign that AMD is feeling very bullish about Zen2's performance (and ETA).Batmeat - Wednesday, October 3, 2018 - link
5th paragraph down -"The company has already stated that it us testing its 7 nm Rome CPUs in the lab."
sing_electric - Wednesday, October 3, 2018 - link
She deserves the keynote. She's really smart, comes across really well in interviews, and AMD has executed nearly perfectly under her watch. They've managed to become very competitive with Intel, and not totally throw in the towel with Nvidia, despite being a tiny fraction of their size (much smaller, if you go by market cap as of say, the beginning of this year).And she's an engineer-turned-exec, rather than an exec who happens to be leading a chip company.
I'm hoping we get some good news about Navi, since Vega wasn't really that competitive with (in terms of performance/watt) with Pascal, let alone Turing (plus, the exclusive use of HBM2 has meant prices have stayed astronomical, even as the mining craze has allowed prices to moderate for other GPUs).
ishould - Wednesday, October 3, 2018 - link
I wonder if Nvidia is really going to release a 7nm chip so close on the heels of the 20-series. They're going to be using TSMC's 7nm (not 7nm+), so that points to a next year release. People that buy the 2080 RTX would be piiiiiissedeva02langley - Wednesday, October 3, 2018 - link
I think we are 2 years away from a launch. I don't see Nvidia trying anything since they have the enthusiasm market. AMD is looking to disrupt mainstream and impose their architecture as the main game development architecture. Performances are going to come with it.sing_electric - Friday, October 5, 2018 - link
That's my question, too. My thinking is that's why you should skip the 20xx cards if you've got a 10xx one (unless, say, you have a 1050 and now have enough money for say, a 2080 Ti or something) - performance is better, but its not RADICALLY better, and not that many games will take advantage of ray tracing in the next 2-ish years (and ones that do will probably have it "retrofitted" on to existing game art). It's the games that are just STARTING development now where they'll be thinking of ray tracing from the outset that'll see the real advantage, and by that time, to get max settings, you'll need one of Nvidia's 21xx (or whatever) series built on 7nm.Of course, if you can look at the $1500 cost of a top-of-the-line GPU as money you might find in your couch, then go ahead.
Kamus - Wednesday, October 3, 2018 - link
Why do I get the feeling they're gonna pull a similar stunt to nVidia, and just release something marginally faster, while raising the price a whole lot?Stochastic - Wednesday, October 3, 2018 - link
AMD has neither market nor mindshare dominance, so I find this unlikely. There's every reason to expect Zen 2 to be a major step up (10-15% IPC improvement) given that it's the architecture's first major refresh. Plus, it's coupled with a new process node.Navi is the big question mark for me. Can AMD's graphics division regain competitiveness with Nvidia?
ishould - Wednesday, October 3, 2018 - link
At 7nm I think Navi will be competitive performance/watt with the 10 series Nvidia cards. Not sure if Arcturus will be a shrink or new architecture, but hopefully that will be competitive with their 30 series cardsZoolook - Wednesday, October 3, 2018 - link
In their last published roadmap, next gen (Arcturus?) i slated for 2020 on 7nm+, and they have previously announced that Navi (GCN 6) is the last GCN-architecture.Da W - Wednesday, October 3, 2018 - link
Actually, since about 30% of Nvidia's chip is targeted at ray tracing that few game use, AMD has an incredible opportunity to pack their GPU to surpass Nvidia, both in performance and power usage, for all games that don't use ray tracing. Plus: AMD prices.eva02langley - Wednesday, October 3, 2018 - link
Honnestly, Navi was a major concern for Lisa after Koduri Vega. I really doubt we should be worried about it at all. It cannot be worst than what Vega 64 was.I think we are going to be surprised. I will buy one for my mini-itx system.
AshlayW - Saturday, October 6, 2018 - link
Vega 64 isn't *that* bad. It fell about 30% short of performance in gaming workloads than was intended mainly due to some features which AMD, for whatever reason, didn't get working in the Gaming driver or production silicon. The whole "Vega 10 is terrible" thing is tiring. Gaming performance is acceptable for most things and Vega 10 has other applications too, where it competes more favourably with the bigger Pascal chips.Arbie - Wednesday, October 3, 2018 - link
Yes, why do you feel that? Given that AMD has delivered big time at least on CPUs, vastly increasing bang for the buck and essentially rebooting the x86 market. Even on GPUs where the Vega product didn't meet our hopes (whether or not those were due to AMD statements) they weren't raising prices.webdoctors - Wednesday, October 3, 2018 - link
CPUs aren't really comparable to GPUs. I've thought about upgrading my i2500k sandybridge from 2011, in 7 years we've only got a 25% improvement for Ryzen:https://us.hardware.info/reviews/7898/14/intel-san...
Compare that to how GPUs have sped up from 2011 to today, its something like 2.5-3X.
Sure GPUs have different bottlenecks and aren't single threaded but that just re-iterates the point that they can't be compared. AMD CPU performance should be compared independently of their GPUs. Intel was stagnant for 5 years allowing AMD to catchup, that's not the case for GPUs.
Fritzkier - Thursday, October 4, 2018 - link
And you know what else? I assume 25% improvement is in single thread application. And most of single thread application is games. Until now, games still bottlenecked by GPU, not the CPU.I think we hit a wall where single threaded performance is already good... Even Intel can't go pass 5Ghz.
Lew Zealand - Friday, October 5, 2018 - link
You're complaining about single core performance in CPUs and comparing that to the massively multicore improvements in GPUs? Dude, those concepts are from 2 different planets. Here's what you should be comparing:The multicore improvements in GPUs (3-4x) vs. the *multicore* improvements between the i7-2600K and TR 2990WX. 612 vs 5000 in Cinebench. It's something like 8x.
OK, maybe that's cheating. How about 2600K vs 9900K (leaked benchmarks if we can trust them). 612 vs. 2166. It's something like 3.5x, same as the GPUs.
Performance marches ahead apace for both CPUs and GPUs, thanks to a bit of pushing from AMD.
AshlayW - Saturday, October 6, 2018 - link
I am so excited about this. I have budgeted to buy the top-of-the-line 3000 series chip for AM4 which I intend to drop in my X470 motherboard. (based on pricing of the 2700X).2019 should be a great year for AMD CPU's as they will, for the first time in a long time be on an even playing-field with Intel in per core performance and an advantage in process technology. I am even hopeful for the GPU side of things, too, as Nvidia is seemingly happy to use 12nm technology for Turing, I think we will see a 7nm RX card be offered against those, potentially competing favourably in Perf/watt as well as the traditional perf/dollar area. The former being an area where I really want to see RTG make headway in.
abufrejoval - Saturday, October 6, 2018 - link
I read four layers of EUV lithography for the 7nm process somewhere: When you look at the huge list of challenges EUV has been adding, that is quite simply awsome.https://en.wikipedia.org/wiki/Extreme_ultraviolet_...
I wonder if they can completely eliminate multi-masking with those four layers and how the fab times compare relative to the last non-EUV nodes.