https://www.techpowerup.com/256842/...that-even-intel-is-impressed-by-amds-progress
There are some good nuggets in there.
There are some good nuggets in there.
ICX was added back to the roadmap which takes advantage of 10nm and our new Sunny Cove uARCH to improve our Perf/ W
This comment was very interesting. We all speak of Intel hiring like crazy, but I never saw this mentioned "hemorrhaging talent".
"
JohnThe "secret sauce" section of this article (personally) makes me cringe. There isn't one piece of evidence that makes me hopeful for Intel in the short term when it comes to competing vs. AMD - no numerics at all, except for how big our former SSG [editor: Software and Services Group] is. Having more SW engineers than AMD? Not convincing. Technology portfolio? Not convincing, esp. given that AMD has much more Graphics/Gaming technology expertise AI/DL expertise? It's pretty clear that we are following in this regard as well. AMD is on fire, esp. given it's size. They continually squeak out new products with less people, and have leap-frogged our HW several times in the past. They have a competitive product with FAR more cores/threads -- which is a huge marketing ploy, and has been what our customers have been asking for. There's no excuse for why we didn't do the same, and now we're in a race to catch up at a time where we are hemorrhaging talent, while also feeling the consequences of the past 5-8 years of mgmt decisions in the way we manage our CPU teams, and the priorities of that mgmt.
Interesting... Wonder what quantities that is going to be available in.
AMD CPUs looked completely hopeless 5 years ago, I don't think anyone from the outside could've predicted this, so I'm not surprised at Intel not really being "prepared" for it,
At least in ST performance and power efficiency. AMD obviously thought they need to out-core Intel in the race to as many cores as soon as possible. Now with Zen 2 launching soon AMD is approaching 16 cores on desktop, 64 cores on HEDT and 128 cores on dual socket servers. Meanwhile Intel is still stuck in 2016 and even their 10nm plans very likely were focusing on monolithic dies, meaning they wouldn't have kept up in the cores race. Makes me wonder if Zen 3/4 again adds more cores.This all goes down to Intel’s 10nm problem. If they have released 10nm on schedule, Intel would have been in good position to counter.
At least in ST performance and power efficiency. AMD obviously thought they need to out-core Intel in the race to as many cores as soon as possible. Now with Zen 2 launching soon AMD is approaching 16 cores on desktop, 64 cores on HEDT and 128 cores on dual socket servers. Meanwhile Intel is still stuck in 2016 and even their 10nm plans very likely were focusing on monolithic dies, meaning they wouldn't have kept up in the cores race. Makes me wonder if Zen 3/4 again adds more cores.
From the memo:
- Intel 9th Gen Core processors are likely to lead AMD's Ryzen-based products on lightly threaded productivity benchmarks as well as many gaming benchmarks. For multi-threaded workloads, such as heavy content creation workloads, AMD's Matisse is expected to lead.
I think the author will likely be disappointed to find Ryzen 3000 beating or matching Intel even in lightly threaded or gaming benchmarks. Just my speculation, of course.
From the memo:
Q. What accounts for AMD's competitive resurgence? Did TSMC turn AMD into our biggest competitor, or is it AMD's focus on higher-end desktop and server parts?
- From 2006 to 2017, AMD had positive net income only three of the twelve years. I'm not sure we can point to a single thing that turned AMD around. But I do think it's was absolutely rooted in the strategic changes AMD initiated in 2015/2016 that narrowed and simplified their focus. AMD shifted to focus on higher margin or premium segments, specifically high-end client, datacenter, graphics for gaming. And they continued their investment in their semi-custom and console business.
- Rather than going after lower-margin, low-end products, they refocused on how to win higher-margin business. AMD added much-needed clarity since they were previously distracted by markets that didn't align with their strengths. They simplified their investments and roadmap and started leveraging best-in-class foundries. Most importantly, they executed to that strategy. Having a clear focus and direction helps enable great execution.
- I also believe AMD's comeback was a result of being very product-centric. A top priority for AMD was building great products - high-performance compute and graphics solutions - from definition to development to delivery.
This part is spot on, IMO, and a very flattering picture of AMD's choices and, particularly, of management. It basically says Lisa Su is doing a great job without saying her name directly.
Zen 3 is 7nm+ so I don't think they will be able to increase core count on that one.
There's talk that the number of chiplets might increase.
Intel Internal Memo said:Intel 9th Gen Core processors are likely to lead AMD's Ryzen-based products on lightly threaded productivity benchmarks as well as many gaming benchmarks.
Intel Internal Memo said:Intel continues to work with press on using real applications for evaluating performance, to produce pieces such as this one from PCPerspective.
Intel Internal Memo said:Xeon is still expected to have cache and memory latency advantages. For this reason, Intel still expects Xeon to be competitive on applications that require fast response times and are sensitive to memory latencies like database, analytics, web serving, and so on.
Intel Internal Memo said:Intel is a premium brand. At times, and on some workloads, we might dip below on performance, like the second half of this year. At other times, and on other workloads, we are 3x or more the performance. Our pricing will continue to reflect the value we deliver to our customers.
Intel Internal Memo said:Especially for enterprise customers, acquisition cost is just one part of the total cost of ownership. Customers using an alternative solution may need additional validation, optimization, debugging, and certifications - all normal cost adders when introducing a new solution in an IT environment.
Intel Internal Memo said:While it has been a number of years since we've faced a similar competitive environment (in the early 2000s with 1 GHz barrier, integrated memory controller, 64-bit, and so on) Intel has risen to every situation and almost always emerged better and stronger.
Of course vastly superior Conroe had nothing to do with Intel dominating the market. It was only due to cheating, (or sour grapes from the AMD camp).That may be wishful thinking on Intel's part. All indications so far are that the Ryzen 3000 series is going to strongly contend with Intel's best current offerings on single-thread performance. Most likely they'll trade blows, with Intel having a slight advantage in some ST workloads and AMD a slight advantage in others. But the pathological cases where AMD was behind by double-digit percentages should mostly be gone.
That sounds pretty bad for PCPerspective. Are they acting as an Intel mouthpiece?
How important is this actually in most real-world applications? Does Intel have a point or are they just blowing smoke?
Translation: Don't expect any price cuts, even if AMD's processors offer far better performance at the same price point. Instead, Intel is going to try to find a handful of outlier examples where their CPUs do better, and justify pricing with that.
Yet it always seems to be Intel products with the exclusive new security flaws (Meltdown, etc.) - which often damage performance to fix.
They dug out of that hole last time via a combination of blatantly anti-competitive tactics and poor follow-up execution by AMD. This time, I think, such tactics would receive more scrutiny and faster smackdown, and they can't count on AMD dropping the ball, certainly not nearly as badly as they did with Bulldozer.
In the first place, where was Conroe mentioned ? Second, Conroe was over 10 years ago, it was great then, but we are talking about TODAY ! Give us a break. And sour grapes ? Over what ? The stuff Intel pulled that cost them over a billion dollars for cheating ? This entire thread is about the fact that Intel acknowledges that AMD is very competitive . So why are you in denial ?Of course vastly superior Conroe had nothing to do with Intel dominating the market. It was only due to cheating, (or sour grapes from the AMD camp).
I'd argue that package area constraints are less important than power scaling where adding CCX chiplets is concerned.I guess it might be possible if IO chiplet moves to 7nm, and they might be able to squeeze in 12 chiplets.
I think Intel treats her name like Bloody Mary's name, They're afraid she will appear when her name is said 3 times and she will bring with her 64c Epyc2 CPUs.
No it's not about that.This entire thread is about the fact that Intel acknowledges that AMD is very competitive . So why are you in denial ?
Who's commenting, Intel employees or not? Should tell anyone everything they need to know.No it's not about that.
Steve Collins isn't even employed by intel as far as I can tell,it's a co-founder of Havok,so even if this article actually was shared on intel Circuit News it has nothing to do with what intel thinks about this.
This is basically a listing of how the laymen see the competition right now,if intel posted this it was to show it's employees how "the media" can shape opinions.
No it's not about that.
Steve Collins isn't even employed by intel as far as I can tell,it's a co-founder of Havok,so even if this article actually was shared on intel Circuit News it has nothing to do with what intel thinks about this.
This is basically a listing of how the laymen see the competition right now,if intel posted this it was to show it's employees how "the media" can shape opinions.