"Intel in jeopardy" from insiders. Much much worse than what they tell.

bigbluefe

Gawd
Joined
Aug 23, 2014
Messages
743
Many phones are 3d stacked with ram on top of the cpu. It sucks ass.
With sillicon level stuff it'll suck just as bad for high power envelope stuff (OC). Don't get too excited.


If 5% makes a difference you're kidding yourself.
Intel has always been significantly faster than AMD in emulators. It's more than 5%, and to get the lowest latency you actually need pretty beefy CPUs.
 

techguymaxc

Limp Gawd
Joined
Jul 6, 2016
Messages
139
Intel has always been significantly faster than AMD in emulators. It's more than 5%, and to get the lowest latency you actually need pretty beefy CPUs.
Seems like Zen2 has largely caught up.

Anandtech tests CPUs with Dolphin 5.0.
https://www.anandtech.com/bench/product/2550?vs=2558Ryzen 9 3700x: 259 seconds
Core i9 9900k: 251 seconds
Core i9 9900ks: 245 seconds


The difference there is quite small at just 3-5.4%, and really just comes down to clockspeed.

But don't listen to me, I'm an Intel shill! ;)
 

bigbluefe

Gawd
Joined
Aug 23, 2014
Messages
743
Seems like Zen2 has largely caught up.

Anandtech tests CPUs with Dolphin 5.0.
https://www.anandtech.com/bench/product/2550?vs=2558Ryzen 9 3700x: 259 seconds
Core i9 9900k: 251 seconds
Core i9 9900ks: 245 seconds


The difference there is quite small at just 3-5.4%, and really just comes down to clockspeed.

But don't listen to me, I'm an Intel shill! ;)
There's a bigger difference in MAME than Dolphin (I think Dolphin actually uses multiple cores).
 

Tapper

n00b
Joined
Jan 31, 2020
Messages
21
When I was building my PC. I was looking around at CPUs and because I am skint AMD was a know brainer. Intel have bin taking the piss for years on the desktop with price. My 2600 is a grate CPU. I went from a core I920 to the r5 2600 and it's nite and day. I can build OpenWrt in about 3 mins! Not including the toolchane. Sorry if my spelling is shit. O and I saved £300 on my build by going with AMD.
 

techguymaxc

Limp Gawd
Joined
Jul 6, 2016
Messages
139
There's a bigger difference in MAME than Dolphin (I think Dolphin actually uses multiple cores).
https://dolphin-emu.org/docs/faq/"With four cores, Dolphin has two cores for the main emulation threads, a third core for other tasks, and another core for the operating system and background tasks to run without taking resources from the emulator."

4 threads, then. That puts most CPUs in the 3rd tier of turbo bins, so for e.g. a stock 9900k it should operate at 4.8GHz when running Dolphin.
 
Joined
Jun 20, 2004
Messages
778
Well they are going desktop gpus but who knows if they are going to get into game in gpu market but for cpus may be awhile.
 

singe_101

2[H]4U
Joined
Dec 17, 2005
Messages
2,110
It doesn't make sense to have an 8600k or 9600k with a 2080 Ti but they aren't CPU-limited in 1440p? Is that like the 480 or 580 where it was burning hot and much less powerful than the stock 2500K (running cool)?
 

dpoverlord

[H]ard|Gawd
Joined
Nov 18, 2004
Messages
1,724
This is very fascinating.

On one end AMD is nearing Nvidia, on the other its on par with Intel.

Meanwhile Nvidia is moving ahead with 7nm
Meanwhile Intel made some bad decisions and laid down the wrong cards

The difference I feel is that Intel has (or so I think) one of the largest cash reserve funds of any company. They should be able to recover. They are like the juggernaut that cannot move fast anymore. However, when they focus their direction and pick up speed its VERY hard to stop them.

So it seems AMD's stock will continue to rise as it has from the teens to 30's then back to 19 then back to $54. People who don't know or read what we do will hedge bets.

In reality AMD has always been the underdog since 3dnow, so it's in a way really nice to see them actually catching a break.

Lets stay tuned to the next episode. I really like the fact that now I have a reason to buy AMD again when I have an upgrade itch.

:)

Thoughts?
 

vick1000

[H]ard|Gawd
Joined
Sep 15, 2007
Messages
1,974
Just because AMD has improved it's DIY, and small custom builder sales greatly, does not mean they have even scratched the surface of Intel's market share of over all chip sales. When you take into account the enterpise market, AMD isn't even a flash in the pan. Intel isn't going anywhere, and is even expanding it's market segment with new hardware.
 

IdiotInCharge

[H]F Junkie
Joined
Jun 13, 2003
Messages
13,268
Just because AMD has improved it's DIY, and small custom builder sales greatly, does not mean they have even scratched the surface of Intel's market share of over all chip sales. When you take into account the enterpise market, AMD isn't even a flash in the pan. Intel isn't going anywhere, and is even expanding it's market segment with new hardware.
To add support, it can be put in simpler terms than that -- Intel still produces the vast majority of CPUs sold, and while AMDs CPUs are absolutely desirable, they have no chance of supplying the market fully in the near term.

If Intel shut their doors today, we'd just not be able to buy new CPUs.
 

techguymaxc

Limp Gawd
Joined
Jul 6, 2016
Messages
139
To add support, it can be put in simpler terms than that -- Intel still produces the vast majority of CPUs sold, and while AMDs CPUs are absolutely desirable, they have no chance of supplying the market fully in the near term.

If Intel shut their doors today, we'd just not be able to buy new CPUs.
In addition to supply issues (for both vendors), there is a very real lack of demand for AMD CPUs outside of desktop market segments.
 

jardows

[H]ard|Gawd
Joined
Jun 10, 2015
Messages
1,797
The difference I feel is that Intel has (or so I think) one of the largest cash reserve funds of any company. They should be able to recover. They are like the juggernaut that cannot move fast anymore. However, when they focus their direction and pick up speed its VERY hard to stop them.
I remember in the late 1990's a discussion being made regarding Novell's cash reserve, that it was so large they could cease all operations and pay all their employee for 20+ years. Less than 20 years later, they were wholly purchased and dissolved as a company. Not saying this will happen to Intel, but a company in stagnation can make some serious missteps and go quickly down.
 

IdiotInCharge

[H]F Junkie
Joined
Jun 13, 2003
Messages
13,268
In addition to supply issues (for both vendors), there is a very real lack of demand for AMD CPUs outside of desktop market segments.
AMD still has to prove that they can get power usage (battery life) of their mobile platforms competitive, and outside of the largest cloud vendors, issues stemming from virtualization and containerization in mixed CPU environments will certainly give many enterprises pause despite the clear on-paper benefits.

But at least in the enterprise, they're making as many CPUs as they can and selling them all. Still a drop in the bucket for that market as a whole, but that's simply how that market does business. AMD is going to have to put in the time with their partners and with the hardware and software vendors that their customers use to build confidence, and we do see them doing that today.
 

IdiotInCharge

[H]F Junkie
Joined
Jun 13, 2003
Messages
13,268
I remember in the late 1990's a discussion being made regarding Novell's cash reserve, that it was so large they could cease all operations and pay all their employee for 20+ years. Less than 20 years later, they were wholly purchased and dissolved as a company. Not saying this will happen to Intel, but a company in stagnation can make some serious missteps and go quickly down.
There're definitely lessons to be learned from Novell among many others -- though it's hard to draw real parallels. If Intel stays behind in fabrication technologies, which would be a departure from their established history up to this point, they'd still 'get by' for at least a decade by producing the parts they're producing today. So perhaps similar to Novell.

The biggest issue as noted above would be for their customers to find alternative sources to replace the sheer volume of products that Intel produces. AMD could backport Zen 2 to pre-7nm (not unlike Intel is doing with Ice Lake...) and requisition all available fab space and still fall short.
 

DooKey

[H]F Junkie
Joined
Apr 25, 2001
Messages
8,301
I remember in the late 1990's a discussion being made regarding Novell's cash reserve, that it was so large they could cease all operations and pay all their employee for 20+ years. Less than 20 years later, they were wholly purchased and dissolved as a company. Not saying this will happen to Intel, but a company in stagnation can make some serious missteps and go quickly down.
This doesn't really fit because Novell wasn't diversified like Intel is. Intel is taking a hit with DIY desktop sales, but they are still raking in the cash with datacenter and mobile sales. Not to mention Optane, SSD, etc. Apples to oranges comparison. Intel has a process problem for the moment. That will be solved one way or the other and life will go on for them.
 

jardows

[H]ard|Gawd
Joined
Jun 10, 2015
Messages
1,797
This doesn't really fit because Novell wasn't diversified like Intel is. Intel is taking a hit with DIY desktop sales, but they are still raking in the cash with datacenter and mobile sales. Not to mention Optane, SSD, etc. Apples to oranges comparison. Intel has a process problem for the moment. That will be solved one way or the other and life will go on for them.
It is simply a warning that a large cash reserve doesn't mean a whole lot if the company can't get its act together. Intel has given itself a lot of black eyes the last few years, CPU hardware security issues only being one (big) of them. If their CPU business starts to flounder, you can be assured that other competitors will start to seriously challenge Intel in other spaces. How Intel reacts to that will be crucial to the survival of the company as we know it.
 

DooKey

[H]F Junkie
Joined
Apr 25, 2001
Messages
8,301
It is simply a warning that a large cash reserve doesn't mean a whole lot if the company can't get its act together. Intel has given itself a lot of black eyes the last few years, CPU hardware security issues only being one (big) of them. If their CPU business starts to flounder, you can be assured that other competitors will start to seriously challenge Intel in other spaces. How Intel reacts to that will be crucial to the survival of the company as we know it.
What makes you think they can't get their act together? They have new architectures that fix the known security issues and simply have to get their process back on target. Heck, they could license better process tech if they needed to. AMD had much worse problems up until 2 years ago and they didn't die. Intel will be fine.
 

jardows

[H]ard|Gawd
Joined
Jun 10, 2015
Messages
1,797
What makes you think they can't get their act together? They have new architectures that fix the known security issues and simply have to get their process back on target. Heck, they could license better process tech if they needed to. AMD had much worse problems up until 2 years ago and they didn't die. Intel will be fine.
I'm not saying they can't. But corporate success usually depends more on management than technical ability. Intel has shown little taste for change in corporate culture, which is why I believe they will not implode or go away any time soon, but at the same time will not reverse the slide of the last couple of years any time soon. If they were prone to change, they may try something "drastic" that could end up destroying the company. Since they are not prone to change, they will not change course, which right now appears to be in the wrong direction.
 
Joined
May 4, 2018
Messages
42
How many generations do cpu companies think ahead... I am willing to bet more than 3. Intel might be watching to see where AMD lands with its 4000 series. Based off of that data it just might skip a planned refresh or entire planed cpu generation to gain back some ground and stay ahead. I bet you roadmaps change all the time due to competition. It is just we never hear about it because of it being top secret.
 

IdiotInCharge

[H]F Junkie
Joined
Jun 13, 2003
Messages
13,268
How many generations do cpu companies think ahead... I am willing to bet more than 3. Intel might be watching to see where AMD lands with its 4000 series. Based off of that data it just might skip a planned refresh or entire planed cpu generation to gain back some ground and stay ahead. I bet you roadmaps change all the time due to competition. It is just we never hear about it because of it being top secret.
Intel has had an architecture finalized that is faster than Zen 2 for at least three years. They just haven't been able to make it.
 

dpoverlord

[H]ard|Gawd
Joined
Nov 18, 2004
Messages
1,724
Intel has had an architecture finalized that is faster than Zen 2 for at least three years. They just haven't been able to make it.
We'll know when a factory is upgraded or a new one built.

No reason due then to not have accomplished this though.

I truly feel motherboards are really lagging behind. Why are there still usb 2.0 ports lol
 
Joined
May 4, 2018
Messages
42
Having usb 2.0 is probably due to the fact that not every device works perfect on usb 3.0. As long as there are 2.0 devices in the wild, I don't think they will completely phase them out.
 

IdiotInCharge

[H]F Junkie
Joined
Jun 13, 2003
Messages
13,268
Why are there still usb 2.0 ports lol
Does your mouse or keyboard need more? USB headset? I get the desire for external connectivity, but at least for desktop boards, there's plenty of stuff to plug in that couldn't make use of > USB 2.0 if it tried :).
 

lopoetve

Imhotep
Joined
Oct 11, 2001
Messages
29,107
This doesn't really fit because Novell wasn't diversified like Intel is. Intel is taking a hit with DIY desktop sales, but they are still raking in the cash with datacenter and mobile sales. Not to mention Optane, SSD, etc. Apples to oranges comparison. Intel has a process problem for the moment. That will be solved one way or the other and life will go on for them.
This.
Funny how some enterprise users say AMD migration is a little tricky but they are still doing it or at least giving it a crack, whilst others don't seem to want to do any of that work and act like ostritches.
AMD of 2020 isn't AMD of 2004 anymore and software support is improving across the board. But it still seems to be an emotion-based decision making process for some.
Depreciation cycles are a pain. It's not that easy.
This is true. Though I must say, surely planned maintenance windows allow for an organization to transition their VMware infrastructure to AMD from Intel.
You going to replace the entire datacenter at once? Without migration compatibility, AMD instances are siloed - that's why we're seeing it crop up in smaller and greenfield environments that do not have existing infrastructure, rather than existing datacenters where it would be the odd man out and isolated. It'll gradually expand if they remain competitive, although licensing changes are making it less cost effective than it used to be.

It's a damned good CPU, but it's not going to replace Intel in a year, or even any time soon. There's a lot that goes into that.
 

SmokeRngs

[H]ard|DCer of the Month - April 2008
Joined
Aug 9, 2001
Messages
16,060
Having usb 2.0 is probably due to the fact that not every device works perfect on usb 3.0. As long as there are 2.0 devices in the wild, I don't think they will completely phase them out.
Does your mouse or keyboard need more? USB headset? I get the desire for external connectivity, but at least for desktop boards, there's plenty of stuff to plug in that couldn't make use of > USB 2.0 if it tried :).
We have winners. USB 2.0 ports are going to be on motherboards for quite a while yet just the way USB 1 ports stayed on basically every motherboard until USB 3 was widely adopted. It's about compatibility, usability and cost. I know for quite a while that there were a number of USB 1 devices which wouldn't work or at least work properly on USB 2 ports. That was one of the biggest reasons it hung around for so long. The exact same thing will be seen with USB 2. USB 2 will not likely be phased out until USB 3 becomes the new standard and that only after USB 4 has been out for a while.
 

techguymaxc

Limp Gawd
Joined
Jul 6, 2016
Messages
139
You going to replace the entire datacenter at once? Without migration compatibility, AMD instances are siloed - that's why we're seeing it crop up in smaller and greenfield environments that do not have existing infrastructure, rather than existing datacenters where it would be the odd man out and isolated. It'll gradually expand if they remain competitive, although licensing changes are making it less cost effective than it used to be.

It's a damned good CPU, but it's not going to replace Intel in a year, or even any time soon. There's a lot that goes into that.
You probably should've read the rest of my posts, as I'm definitely not suggesting that cross-vendor VM migration is as simple as flipping a switch. I only said that it can be planned for.
 

lopoetve

Imhotep
Joined
Oct 11, 2001
Messages
29,107
You probably should've read the rest of my posts, as I'm definitely not suggesting that cross-vendor VM migration is as simple as flipping a switch. I only said that it can be planned for.
it’s still a silo then; can’t move in or out of that silo, unless you take an outage. Unless you’re replacing everything, or a major portion of it, why bother? It’s not worth the headache to most people. Eventually the depreciation cycles will catch up, but it takes time.
 

jeremyshaw

[H]F Junkie
Joined
Aug 26, 2009
Messages
12,182
We have winners. USB 2.0 ports are going to be on motherboards for quite a while yet just the way USB 1 ports stayed on basically every motherboard until USB 3 was widely adopted. It's about compatibility, usability and cost. I know for quite a while that there were a number of USB 1 devices which wouldn't work or at least work properly on USB 2 ports. That was one of the biggest reasons it hung around for so long. The exact same thing will be seen with USB 2. USB 2 will not likely be phased out until USB 3 becomes the new standard and that only after USB 4 has been out for a while.
I think the bigger problem is USB 2.0 only needs power, 2 data lines, and ground. USB 3.0 needs all that, 4 more for the 2 high speed differential pairs, and one more for a signal ground. I'd venture those high speed differential pairs also come with stricter routing requirements.

Either way, my mouse and KB don't need any more data than USB 2.0 can provide. Same with basic controllers (fan/pump/LED controllers) and my UPS. A lot of devices that don't need more than 2.0 for the foreseeable future.
 

Jandor

Gawd
Joined
Dec 30, 2018
Messages
527
News still worse for Intel. Ice Lake is coming out and it's a joke compared to AMD Renoir. Maybe replacing low end CPU on laptops. Intel earnings will melt down with this. Even Apple is now looking into AMD CPU stuff even if they will put Ice lake CPUs in their new Macbooks. Intel showed a GPU that's kind of a basic 25W discreet GPU with lower specs than competition. They compared it to AMD Renoir APU saying it's quite faster but not at the level of a discrete GPU. And they're preparing a big 500W TDP GPU with 4 chiplets inside and HBM2 and Foveros multi-chip technology. It's like they used a lot of expensive tech to get at the level of AMD and Nvidia. And the GPU may be made at TSMC or Samsung on 7nm. I'm not saying this is going to be a failure but the problem is not how good it is but how effective the architecture is on the GPU side.
 

5150Joker

2[H]4U
Joined
Aug 1, 2005
Messages
3,589
Intel has several business vulnerabilities right now in consumer and eventually enterprise. NVIDIA is boning them and AMD with their datacenter GPUS and Ryzen is improving quickly with each iteration. Intel discrete gpu won't come close to matching nVidia but I'm sure intel will supplement that deficiency with financial kickbacks to some companies.

The problem I see for Intel is that all it will take is a few top 5 companies to take a chance on AMD before a shit storm begins. Say tomorrow Amazon announces that in 2022 all server instances moving forward will be built using AMD Ryzen, that would be the beginning of Intels decline. You can be sure others would follow Amazon. If intel doesn't sort it's shit out in the next 2-3 years I can see this happening.

Also ARM is finally getting serious about server solutions and if they can keep their promise of huge gains every generation, they will eventually become a viable alternative: https://www.electronicdesign.com/industrial-automation/article/21807632/arm-introduces-chip-designs-dedicated-to-data-centers

https://www.anandtech.com/show/13959/arm-announces-neoverse-n1-platform
Then there's China, I think they are more likely to pick a design based on AMD than intel since AMD has shown a willingness to partner with Chinese companies and share their cpu designs with them and the Chinese are very much interested in producing indigenous cpus.

Things are a lot more complicated for Intel now than they were a decade or two ago where they had a guaranteed rebound. Many companies that were once giants of their industry have fallen into oblivion, Intel is no different in this case. Do I think they will fade away? No, but I don't think things are as rosy for them as some here would like you to believe just because they're some IT admin at some mid tier company. I'm not in IT nor would I want to be but I do make investments and this isn't hard to decipher even for the layman.

But it's not all bad news for Intel, one thing Intel has going for it is vertical integration of it's entire portfolio once it sorts out 10/7nm and introduces some halfway decent GPUs for the data center. Will it be in time and enough? Who knows, things aren't looking great right now that's for sure. The thing I don't like about NVIDIA/AMD business model is that they're entirely at TSMC's mercy for supplies and in case of AMD, scaling up both CPU/GPU volume production to serve a market as large as Intel does could be a problem if TSMC decides to raise prices on them or prioritizes allocation to Apple and other volume customers. Sure Samsung is kind of a competitor to TSMC but they haven't proven themselves yet for large CPU/GPUs at 7nm so as of now, TSMC has a monopoly and can ask for whatever price they want and AMD/NVIDIA have to play ball.

Look at the rising cost of production as these processes get smaller:

nano3.png

source: https://semiengineering.com/big-trouble-at-3nm/

NVIDIA can't keep raising prices forever to maintain margins, something will give and if MCM doesn't take off for GPUs in a meaningful way, their business will be in disarray. Same goes for AMD to a smaller extent, they will face increasing costs and TSMC sure as shit won't eat those costs for them. Intel can streamline their production costs because they own their foundries so they can maintain healthier margins in the long run vs AMD/NVIDIA and others who depend on outside foundries.
 
Last edited:

nEo717

Limp Gawd
Joined
Jun 2, 2017
Messages
344
Amazon has already been opening the doors up to AMD:

https://aws.amazon.com/ec2/amd/
Amazon Cloud Servers:

https://www.bloomberg.com/news/articles/2018-11-06/amd-says-amazon-s-cloud-unit-is-a-partner-for-epyc-server-chips
Amazon may not be the best customer long term for either Intel or AMD with them wanting to move to Graviton2 chips.
-------

Intel has several business vulnerabilities right now in consumer and eventually enterprise. NVIDIA is boning them and AMD with their datacenter GPUS and Ryzen is improving quickly with each iteration. Intel discrete gpu won't come close to matching nVidia but I'm sure intel will supplement that deficiency with financial kickbacks to some companies.

The problem I see for Intel is that all it will take is a few top 5 companies to take a chance on AMD before a shit storm begins. Say tomorrow Amazon announces that in 2022 all server instances moving forward will be built using AMD Ryzen, that would be the beginning of Intels decline. You can be sure others would follow Amazon. If intel doesn't sort it's shit out in the next 2-3 years I can see this happening.

Also ARM is finally getting serious about server solutions and if they can keep their promise of huge gains every generation, they will eventually become a viable alternative: https://www.electronicdesign.com/industrial-automation/article/21807632/arm-introduces-chip-designs-dedicated-to-data-centers

https://www.anandtech.com/show/13959/arm-announces-neoverse-n1-platform
Then there's China, I think they are more likely to pick a design based on AMD than intel since AMD has shown a willingness to partner with Chinese companies and share their cpu designs with them and the Chinese are very much interested in producing indigenous cpus.

Things are a lot more complicated for Intel now than they were a decade or two ago where they had a guaranteed rebound. Many companies that were once giants of their industry have fallen into oblivion, Intel is no different in this case. Do I think they will fade away? No, but I don't think things are as rosy for them as some here would like you to believe just because they're some IT admin at some mid tier company. I'm not in IT nor would I want to be but I do make investments and this isn't hard to decipher even for the layman.

But it's not all bad news for Intel, one thing Intel has going for it is vertical integration of it's entire portfolio once it sorts out 10/7nm and introduces some halfway decent GPUs for the data center. Will it be in time and enough? Who knows, things aren't looking great right now that's for sure. The thing I don't like about NVIDIA/AMD business model is that they're entirely at TSMC's mercy for supplies and in case of AMD, scaling up both CPU/GPU volume production to serve a market as large as Intel does could be a problem if TSMC decides to raise prices on them or prioritizes allocation to Apple and other volume customers. Sure Samsung is kind of a competitor to TSMC but they haven't proven themselves yet for large CPU/GPUs at 7nm so as of now, TSMC has a monopoly and can ask for whatever price they want and AMD/NVIDIA have to play ball.

Look at the rising cost of production as these processes get smaller:

View attachment 224902
source: https://semiengineering.com/big-trouble-at-3nm/

NVIDIA can't keep raising prices forever to maintain margins, something will give and if MCM doesn't take off for GPUs in a meaningful way, their business will be in disarray. Same goes for AMD to a smaller extent, they will face increasing costs and TSMC sure as shit won't eat those costs for them. Intel can streamline their production costs because they own their foundries so they can maintain healthier margins in the long run vs AMD/NVIDIA and others who depend on outside foundries.
 

techguymaxc

Limp Gawd
Joined
Jul 6, 2016
Messages
139
it’s still a silo then; can’t move in or out of that silo, unless you take an outage. Unless you’re replacing everything, or a major portion of it, why bother? It’s not worth the headache to most people. Eventually the depreciation cycles will catch up, but it takes time.
Whether or not AMD Epyc-powered hosts are best reserved for an isolated environment, or as the platform of all compute hosts for an entire environment depends on the needs of the customer. Some environments do not have cross-datacenter or cross-cluster failover, but only n-1 redundancy within a cluster. If this has been deemed appropriate for your environment, then replacing say all the hosts in a given cluster, but not another cluster, may be perfectly acceptable. Some environments dedicate clusters to specific workloads and never see cross-cluster migration outside of development and testing.
 

5150Joker

2[H]4U
Joined
Aug 1, 2005
Messages
3,589
Amazon has already been opening the doors up to AMD:

https://aws.amazon.com/ec2/amd/
Amazon Cloud Servers:

https://www.bloomberg.com/news/articles/2018-11-06/amd-says-amazon-s-cloud-unit-is-a-partner-for-epyc-server-chips
Amazon may not be the best customer long term for either Intel or AMD with them wanting to move to Graviton2 chips.
-------
Yup I mentioned the ARM neoverse in my post. Seems Amazons Graviton 2 does really well in AWS instances vs intel xeon. With MS also making their own ARM silicon, AMD and Intel should be worried.
https://www.zdnet.com/article/aws-graviton2-what-it-means-for-arm-in-the-data-center-cloud-enterprise-aws/
 
Last edited:

lopoetve

Imhotep
Joined
Oct 11, 2001
Messages
29,107
Whether or not AMD Epyc-powered hosts are best reserved for an isolated environment, or as the platform of all compute hosts for an entire environment depends on the needs of the customer. Some environments do not have cross-datacenter or cross-cluster failover, but only n-1 redundancy within a cluster. If this has been deemed appropriate for your environment, then replacing say all the hosts in a given cluster, but not another cluster, may be perfectly acceptable. Some environments dedicate clusters to specific workloads and never see cross-cluster migration outside of development and testing.
That's what I was saying. Existing environments that are growing will not suddenly switch to AMD - greenfield and new projects may, if it's an option (Depends on the use case), or when the depreciation cycle brings something up for refresh. Also lots of things where AMD isn't an option (appliances, etc) - customers stick with what works, and the choice of CPUs is a minor triviality to most enterprises compared to the rest of the solution (assuming the enterprise isn't stuck in the weeds). Intel isn't going anywhere, regardless of how good an individual part may be. Business is slowing, sure - but they ain't going anywhere.
 

techguymaxc

Limp Gawd
Joined
Jul 6, 2016
Messages
139
That's what I was saying. Existing environments that are growing will not suddenly switch to AMD - greenfield and new projects may, if it's an option (Depends on the use case), or when the depreciation cycle brings something up for refresh. Also lots of things where AMD isn't an option (appliances, etc) - customers stick with what works, and the choice of CPUs is a minor triviality to most enterprises compared to the rest of the solution (assuming the enterprise isn't stuck in the weeds). Intel isn't going anywhere, regardless of how good an individual part may be. Business is slowing, sure - but they ain't going anywhere.
I'm not making the argument that "Intel is in trouble", I've actually been arguing against that premise from the beginning so I don't think we're really in disagreement here. What I am saying is that AMD Epyc CPUs offer some long-needed competition in the datacenter market, and are worthy of consideration.
 

lopoetve

Imhotep
Joined
Oct 11, 2001
Messages
29,107
I'm not making the argument that "Intel is in trouble", I've actually been arguing against that premise from the beginning so I don't think we're really in disagreement here. What I am saying is that AMD Epyc CPUs offer some long-needed competition in the datacenter market, and are worthy of consideration.
Well then dammit, we’re in violent agreement. Fine! Be that way!
 

IdiotInCharge

[H]F Junkie
Joined
Jun 13, 2003
Messages
13,268
Say tomorrow Amazon announces that in 2022 all server instances moving forward will be built using AMD Ryzen, that would be the beginning of Intels decline. You can be sure others would follow Amazon.
They could all announce it -- but then they wouldn't be able to follow through. TSMC couldn't supply them.

TSMC supplying several orders of magnitude more AMD CPUs isn't going to happen in two or three years either, and it's hard to see a business case for them to try: Intel hasn't stopped shipping CPUs and Intel isn't likely to continue to fail to advance their fabs, meaning that the demand for AMD CPUs from TSMCs fabs is likely to flatline and even decrease in the future.

Yes, Intel could continue to fail -- and that'd be bad for everyone. Everything that uses advanced processes, from every advanced fab, will see supply decrease and prices increase. You thought phones were expensive now? Etc.
 

5150Joker

2[H]4U
Joined
Aug 1, 2005
Messages
3,589
They could all announce it -- but then they wouldn't be able to follow through. TSMC couldn't supply them.

TSMC supplying several orders of magnitude more AMD CPUs isn't going to happen in two or three years either, and it's hard to see a business case for them to try: Intel hasn't stopped shipping CPUs and Intel isn't likely to continue to fail to advance their fabs, meaning that the demand for AMD CPUs from TSMCs fabs is likely to flatline and even decrease in the future.

Yes, Intel could continue to fail -- and that'd be bad for everyone. Everything that uses advanced processes, from every advanced fab, will see supply decrease and prices increase. You thought phones were expensive now? Etc.
That's one of the key points i mentioned in my post, AMD's reliance on TSMC and why it would have trouble filling in the magnitude of orders Intel does reliably. TSMC has so many clients wanting their latest fabs that AMD wouldn't likely get first priority and even if it did, TSMC probably wouldn't have enough capacity for them w/out building more fabs that AMD couldn't afford to help subsidize. The bigger problem for x86 is ARM though, Amazon w/Graviton 2, MS w/their own ARM Neoverse chips and even Apple eventually going full ARM and abandoning x86. I can see medium scale enterprise businesses switching to cloud for their needs as it's cheaper that way and Azure/AWS in 5-10 years will likely run their own proprietary ARM chips thus cutting out Intel/AMD in a big way.
 

IdiotInCharge

[H]F Junkie
Joined
Jun 13, 2003
Messages
13,268
I'm still... very much not sold on ARM.

Mostly because it's been over-promising and under-delivering for a very long time, and it's taken Intel to get to the point of needing a come-to-Jesus moment over their own fabs for ARM to appear close to competitive.

ARM scales lower, x86 scales higher.

Now, for some / many server applications, that's actually fine. Transactions per watt becomes the game that cloud providers play, and in the enterprise arena, much of the work being done simply isn't heavy compute on CPUs as much as it is shuttling data around.

The problem I see is the 'backlash' from The Great Cloud Migration, which is hybrid cloud. Cloud costs more than what the provider charges in terms of control and latency, and there will always be a benefit to local hosting to consider; and once the workloads have all been enumerated and containerized, that balancing act will be fairly straightforward -- so long as the architectures are the same where they matter.
 
Top