AMD takes a bite out of Intel's PC market share across Europe amid microprocessor shortages, rising Ryzen Intel is losing ground to AMD in every corner of the European PC industry serviced by the channel, according to official sales stats from distributors. Figures collated by Context for Western Europe, for the fourth quarter of 2019, reveal shipments of Intel processors for notebook, desktop and workstations fell 2.3 per cent …
> Gartner analyst Mikako Kitagawa said.. switch to AMD-based kit, but she said it is "time consuming for business users, especially large companies where a large number of PCs are deployed."
In what way is it time consuiming? It takes just as long to buy an AMD based PC as an Intel one, ditto installing Windows, apps, etc. Other than creating a new master image with the required drivers I don't see what is different and you would need to do that for a new range of Intel based PCs anyway.
Maybe if these analysts had actually worked in IT at some point they might understand the subject better and make fewer stupid remarks like the one above.
>> Gartner analyst Mikako Kitagawa said.. switch to AMD-based kit, but she said it is "time consuming for business users, especially large companies where a large number of PCs are deployed."
Yeah, about all those Intel specific vulnerabilities......
Perhaps it's a good reason to now switch to AMD as our servers and workstations performance has suddenly fallen off a cliff.
I used to think this "Intel runs more slowly when patched" was just a marginal thing until my laptop manufacturer finally produced a BIOS update with the microcode fixes in, and my program to solve Microsoft Wordament puzzles suddenly started taking over twice as long to run!
So yes, AMD for me next time around.
> Gartner analyst Mikako Kitagawa said.. switch to AMD-based kit, but she said it is "time consuming for business users, especially large companies where a large number of PCs are deployed."
This is so dumb! But then he reports the mindset of CIOs, so that should not be surprising. What's a CIO? Someone who is paid a fortune to do one job: Repeat all day long "You must install Windows and you must do so using a processor from Intel". Either Windows or Linux (and many others) are very oblivious to AMD/Intel. This is not like, once upon a time, having say SPARC workstations around. Now that's a whole other beast.
Recently I attended a talk given by John Cleese (of Monty Python fame), and one thing he said is that people who have a job and are actually knowledgeable about it (know what they are doing, know what they are talking about) are a surprisingly small minority, about 1 in 6 people. Well here's some additional evidence!
People will often make the case, “We can’t be that stupid, or we would have been evolutionarily wiped out as a species a long time ago.” I don’t agree. I find myself saying, “Well, no. Gee, all you need to do is be far enough along to be able to get three square meals or to solve the calorie problem long enough so that you can reproduce. And then, that’s it. You don’t need a lot of smarts. You don’t have to do tensor calculus. You don’t have to do quantum physics to be able to survive to the point where you can reproduce.” One could argue that evolution suggests we’re not idiots, but I would say, “Well, no. Evolution just makes sure we’re not blithering idiots. But, we could be idiots in a lot of different ways and still make it through the day.”
“Well, no. Gee, all you need to do is be far enough along to be able to get three square meals or to solve the calorie problem long enough so that you can reproduce. And then, that’s it. You don’t need a lot of smarts"
When we were all wandering the steppe looking for a meal, the rash, the stupid and the clumsy had a lot more chances to take themselves out.
Civilisation provides a safety net. This is a good thing in some will get to learn from a mistake or two.
But some never learn, and get to pass on rash, stupid and clumsy to another generation.
>What's a CIO? Someone who is paid a fortune to do one job: Repeat all day long "You must install Windows and you must do so using a processor from Intel".
Not really. He's the one paid a lot of money to make sure no-one messes up.
In this case, "is that weird but critical app we use from 2015 optimised for intel cpus and does it run badly on AMD?"
It isn't a question of "you must use intel" its a question of, "is it worth the cost savings to do all the required QA now and in the future to save a small proportion of the cost of a pc?" "How well does vmware run on AMD?" "Do I want to lose my job over that decision?"
The answer is usually, "I don't want to lose my job and I can do that by buying intel systems which no-one important will fault me for."
The laptop screen is probably a more expensive component. Do you use dual intel 10G nics in servers? Do you want to QA those in an AMD server? In all your AMD servers?
I'd love AMD to do even better, but I get why CIO's tell their staff not to bother spending time on switching.
I am glad it wasn't just me that read that and thought WTAF?
Even with just a new version of say a laptop, you often have to add drivers to an image-based solution. It makes not one jot if it's an AMD or Intel chipset.
And since we're talking in most cases here, of Windows systems, in corporations, it's even fucking simpler with Autopilot.
Dullards.
In the server space, Intel has some specific hardware locked down to their own systems.
However, the security risks and AMD adding new features (PCIe 5) etc means they are looking more and more worth the little downtime/testing/switch over costs.
I remember the argument being that if I ask say a Dell rep about a quote for say a specific config for an R630 server (which is intel based) then he'll be able to instantly give me a price. Simply because Dell sells a lot of R630 servers.
But when it comes to more obscure kit like, heaven forbid, AMD servers like a Dell R7425 then the poor rep would have to spend time checking pricing and maybe even ask colleagues.
Yeah, silly, I know, but that's the corporate world.
It takes just as long to buy an AMD based PC as an Intel one, ditto installing Windows, apps, etc. Other than creating a new master image with the required drivers I don't see what is different and you would need to do that for a new range of Intel based PCs anyway.
It's not more expensive to sell AMD PCs than Intel, but it is more expensive to have to sell BOTH, slightly different models.
For your home PC, you just put something together, and if it seems to work, you go. For big vendors, all their hardware gets extensively tested and certified as compatible with all the major software out there, and full compatibility in all the edge cases with all different possible (supported) combinations of hardware is tested as well.
Businesses that buy thousands of PCs from a vendor in a go don't expect to get something that works most of the time. If they want to stick a SCSI controller card next to high-end video card, it all has to work.
I remember my old PC Chips motherboard... If a case screw happened to make contact with the metal plating around the holes, the second IDE channel would freeze. Worked perfect out of the case on a bench for testing (as the seller told me when I returned it), but in a case, it would go wonky. Even with good, name brand gear, there's still all those edge cases where it doesn't always support every configuration somebody might try.
"For big vendors, all their hardware gets extensively tested and certified as compatible with all the major software out there"
How naive.
It's more like
"Big vendors deliberately limit the hardware combinations to minimize the chance of weird driver issues, and are *big enough to shout at the suppliers* until they fix most problems"
Exactly and in my past experience AMD driver support has been awful! Issues with deployment and software bugs are much more prevalent in AMD products. If however they have improved on these along with the improved CPU hardware then yes it will be more viable for companies to move to AMD products. Only in 2016 I tried rolling out some AMD servers and PCs for a firm and they were just awful in terms of product software, hardware and support.
Our lot "buys thousands of PCs" and they support a range of options from slim execs lappies, standard desktops, developer xeon workstations, windows on arm, macs and a lot of other kit that came with aquisitions.
They dont extensivly test PCs.
Same with servers.
Anything that supports amd64 :)
In what way is it time consuiming? It takes just as long to buy an AMD based PC as an Intel one, ditto installing Windows, apps, etc. Other than creating a new master image with the required drivers I don't see what is different and you would need to do that for a new range of Intel based PCs anyway.It's a little more complex than that at a large enterprise.
New hardware means new drivers as you pointed out, which is for all intents and purposes a new O/S from a SOE (Standard Operating Environment) point of view. You have to build a new SOE (master O/S image with drivers and other required software). As part of building that new SOE, all the apps used by the enterprise will have to be certified for the new SOE - will third-party vendors support that software on a non-intel PC - that's not as silly as it sounds in the Enterprise, paid support space. They also have to support the new SOE. There will be different support processes (i.e. different error messages that will need new or updated support flowcharts, documentation, wiki entries, etc. for the non-technical frontline helpdesky people). New hardware to support, so instead of having 30k desktop of poduct X that all use the same BIOS/UEFI firmware images, drivers, installation routines, patches, etc., now they could have 20k product X and 10k product Y that also has to have the same effort put in (i.e. rather than effort X, it's now effort X + effort Y).
You are right in saying it is no different when doing a major desktop replacement, going between generations, where you'll have a desktop replacement program and for a year or two you'd have to support both as the migration progresses. But those are usually planned and budgeted process. The issue here is that this is an unplanned, un-budgeted situation. This could be happening mid-cycle, i.e. they expected (and budgeted for) only be supporting X, but because of the chip shortages they can't get any more (or enough) system X's for their current needs, so if they want to have enough desktops, they have to get system Y at a time in their budgeting cycle they hadn't budgeted to have to support two systems, so they might have to support 25k X's and 5k Y's, each of which take the same SOE-building and certification effort.
Or, even worse, they already were in the middle of a desktop replacement cycle, having to support (Intel) X and Y systems, but they are unable to get enough Y's to finish the migration, or enough X's to offset the Y deficit, therefore they now have to get (AMD) Z's to create a 3rd SOE and use Z's to complete the replacement of their outdated X systems. So now during their aborted Y migration, where they've had to substitute unfilled Y orders for Z's, they might have to be support 10k unreplaced X's which are being replaced by Z's, 15k Y's which they can't get any more of, and 5k Z's (whilch will eventually be 15k once all remaining X's have been replaced with Z's). So they've had to build an additional SOE on top of what they budgeted ($100k+, possibly way more) plus support three, rather than 2, SOEs during this transition. Which means instead of supporting just one SOE/firmware system for the 3 years between cycles when the replacement is completed to the beginning of the next replacement cycle, i.e. just Y, they now have to support Y and Z for the rest of the 3-year cycle until they can do a full refresh of only 1 type of platform - during which time they'd again have to support 3 SOEs until they'd done a refresh of Y+Z with a single new type of system.
When you are talking large corporations, thousands, tens of thousands, hundreds of thousands of desktops, it's a bit different then just going down to the local computer store, ordering 20 computers to pickup next week, and replacing all 20 of your business computers in one weekend with a single office IT nerd instead of a dedicated IT team doing all the work ...
Well it's down to supply alone if the demand grows 40% and all they can meet is 20%. I made those numbers up but the concept must be true at this point (even intel is presenting it).
25% increased waffers is great news, _IF_ people actually want their CPU. Let's face it, people were just used to buying intel and now that they've had to research alternatives, maybe they don't.
Intel needs a new CPU, it's as simple as that. They can't keep robbing Peter (7nm) to pay Paul (10nm).
"Well it's down to supply alone if the demand grows 40% and all they can meet is 20%. I made those numbers up but the concept must be true at this point"
Whatever the cause is, the vast majority of the delay in delivery is occuring in the low end of the market.
If you order high end desktop processors the delivery delays are much shorter.
"Whatever the cause is, the vast majority of the delay in delivery is occuring in the low end of the market."
Intel can't hit production targets because they are at capacity. Intel can hit revenue targets by altering the product mix to favour more expensive chips.
As long as investors and banks see revenue growth, Intel can obtain loans or sell bonds to finance more very expensive fabs.
Yes, it hurts the value end of Intel's market, but the reality is Intel has A LOT of other issues to worry about. Five years ago, Intel was planning to dominate the semiconductor world from CPU's to FPGAs to mobile comms. The inability to produce next-gen processes has killed their mobile dreams, severely impacted FPGA's and is now hurting CPU's. Intel will have pumped ~$40bn into fabs over 2018-2020 and if that doesn't provide a competitive process (ideally equivalent or better than TSMC's 5nm) as a minimum (a working Intel 10nm process would really help with capacity...), it will be hard to justify the next $20bn fab investment...
"Read one way Intel is increasing capacity 25% however it said PC so I am guessing that they are just diverting server capacity to PCs"
They are going from 4 to 5 fabs for 14nm so it is a genuine 25% wafer capacity increase. However...as the only high clock chips they can make are 14nm and their answer to AMD's core counts is to double up on cores/dies, the likely capacity increase on the basis of number of end user CPU's produced will depend on the dies per CPU. i.e. assuming all single die parts in 2018, 20% dual die parts in 2019 and 40% dual die parts in 2020 and 14nm being so mature that defect rates are close to equal across the three years:
2018: 100% effective 2018 capacity
2019: 80% effective 2018 capacity
2020: 85% effective 2018 capacity
Disclaimer: The product split numbers are pulled out of my butt - I haven't seen anything to suggest the actual breakdown.
At 10nm, again they are going from 4 fabs to 5 BUT it appears they may have 3 of the 5 actually able to produce usable chips based on rumours (versus <0.5 usable 10nm fabs in 2018/2019). They still can't get the top end frequencies that 14nm provides, but they can get above 4GHz which is good enough to allow them to use the parts for mainstream and ship in volume. These are only rumours and Intel has to address performance AND production (both defect rates and volume) so whether they will hit targets is unknown outside Intel...
If Intel can ship volume parts on 10nm, they can likely address the majority of the laptop market, avoid heavy losses on the high margin mobile parts and take a lot of pressure off 14nm.
14nm will continue to ship performance parts - the question is what is the split between single and dual die parts and whether there is an effective capacity increase inspite of more dies being produced?
Not quite as simple as is made out. Demand is for higher core counts and improved capabilities. Using the 14nm(+++) node then you get fewer chips per wafer (partly because you can fit fewer of these bigger chips on a wafer, and partly because your failure rate goes up as the chip size increases). With 10nm Intel have poor yields to start with, and they can’t match the performance of chips fabricated on the, now very mature, 14nm node.
AMD have been rather smart here, in that their CPUs are fabricated in batches of 8 cores on 7nm chiplets. These are small, so with binning of chiplets that have some but not all cores working, along with the fully operational parts, they get very high yields. They then use cheaper 14nm IO chips to stitch the chiplets together into a seamless whole, letting them scale cores for almost linear incremental cost. Intel are having to compete with 16 core and higher parts, when their fastest parts have to be made on 14nm, and are monolithic so need a huge area on a wafer to get anywhere close.
Um, look, I'm very happy that AMD is riding the wave right now, but you're fooling no one. AMD has had capable products before and not attained such levels. You and I both know that, despite AMD having an excellent product, it is Intel's failure to deliver over a long period that has pushed companies and consumers to switch their mindset and go to you.
And that's fine.
I’ve often thought that Intel make the slowest chips that they think they can sell, whereas AMD have had to make faster or better chips in order to sell anything at all. AMD haven’t always managed that of course, but right now technical superiority is paying dividends.
I also think that AMD’s decision to divest its fabs has paid off. It allows them to more readily survive the fallow times. Of course it means that they can’t ramp up massively because they have to buy manufacturing capacity on the open market, but that has given them access to the Taiwanese 7nm expertise that Intel still hasn’t fully replicated.
I don't think it's just the shortages, AMD have now managed to open quite a gap in the price/performance stakes.
That isn't just down to more cores for the money or a more intelligent/efficient way of managing the chips, that also has to do with the fact that current Intel users out there at present suffer significant drops in performance as a result of the patches to keep Intel's backdoor shut.
The result is a combined bang for buck and security problem that will take quite a while to address, and AMD is making a mint as a result. Good on them.
I agree with you sentiment. However AMD gained market share when they beat intel to 64bit. I think if AMD did not have a strong product line as Intel faltered we would not see these numbers.
IMHO AMD deserves a pat on the back for the Ryzen line, it's good to see them getting paid.
Intel, TSMC and Samsung are all investing between $10bn-15bn/annum into foundries at the moment because the three are so competitive and very close to the limit of what they can do.
While Intel have looked to outsource capacity, I don't believe they have any CPU products shipping from those ventures at present unless you count Nervana/MobileEye (produced at TSMC but acquisitions so yet to be moved to Intel fabs) and unconfirmed rumours around chipsets at Samsung (i.e. it maybe happening but no shipping products at present). At various points in time acquired companies have come with their own fabs or used third parties for chip production, but generally Intel has aimed to bring production in-house. Without wanting to belittle Intel's design side or downplay their 10nm issues, Intel's strength over the years has been production (both in high yield designs and low defect processes).
So, Intel lost just a small chunk of their share almost monopolistic of the market, but at the same time they sold a lot more in the high end segment, so I guess that this "shortage" didn't affect the balance sheet. In the meantime they are cutting their staff, so I also guess that the manager bonuses will remain unscathed. Are we sure that the shortage is accidental?
"Intel lost just a small chunk of their share almost monopolistic of the market"
The issue was and is that Intel had been falling way behind on demand for a long period before AMD launched the new CPUs - AND not bothering to address the issue - AND keeping pricing up.
So when the latest generation of AMD CPUs started sweeping the field, they really weren't prepared and they can't _buy_ enough outsourcing capacity to catch up because it's all booked out.
That's what hubris gets you.
"Being able to replace a full rack of Intel with two AMD boxes is music to the ears of any data center professional."
The "rack full" of dual CPU Intel boxes MAYBE being replaced by "less than a rack full" of AMD boxes.
Licensing makes high core counts hard to justify for Enterprises unless they have applications (like HPC) that aren't licensed per core.
The high core count chips are power/cooling hungry making it hard to sell them into cloud environments or enterprise blade chassis where density is more important. That's not to say moving from 20 servers per rack to 10 while lowering connectivity costs and increasing core counts doesn't make sense.
IO (typically storage) also kills your per rack density - you are likely limited in the amount of IO you can deliver to each rack. So moving from 20 x 8-cores to 2 x 128 cores likely doesn't unless you can avoid storage/IO/network/memory bottlenecks by scaling up connection speeds cost effectively.
Can you mix your new CPU's with your existing CPU's in a VM environment? Or do you need to start a new farm. That makes switching vendors hard unless your current environment is end of life and ditching it is an option. If you have a larger environment and replace a fixed percentage each year, changing may require considerable planning/budgetting.
And finally, what does it cost all up?
If any of the Super7+1 (or is that +2 now nVidia are trying to compete for mobile gaming?) announce major deals with AMD in 2020, AMD likely double their server market share.
TL;DR: high core counts look great, but AMD's pricing, power usage, low-to-mid range core counts and being able to deliver are more likely to win them this round.
"they can't _buy_ enough outsourcing capacity to catch up because it's all booked out."
To address outsourcing high performance chip designs to another fab - you will likely take 1-2 years to redesign a working CPU operating at >2GHz if you take a working design from one fab to another.
Intel/TSMC/Samsung all have significantly different processes - a working design for one fab does not instantly translate into a working design at another fab, and if you are dependent on a design that is outside of the conservative fab design rules, you risk low yields and poor performance. For examples, look at AMD/nVidia GPU's on identical TSMC processes where one card substantially under-performs or has low availability - typically the first GPU or largest GPU does not clock as highly as its rival because of design issues. A year later, we see a decent boost as the updated chip addresses the issue.
Gartner doesn't actually do any real analysis. CIOs pay them to say "the current market leader will continue to be the current market leader" which can be used for CYA purposes if and when the market changes.
AMD is absolutely killing it in the consumer market right now, and that success is starting to bleed into the server space. Massively multi-core silicon is perfect for dense virtualized computing environments. Being able to replace a full rack of Intel with two AMD boxes is music to the ears of any data center professional.
1. New docking station
2. Different Drivers
3. Different OEM
4. Different OS
These all happen when you change models etc, the cost of changing to a different type of docking station is huge compared to a different type of processor.
Change from Intel Processors to AMD ones makes no difference unless you are big in to vPro.
Images are done for each new device, it make little difference to deployment as long as the drivers are there, same testing same effort.
We support a mix of Dell and Lenovo kit and around 8 or more different laptop models and 4 or 5 different desktop models, adding AMD processors will make little difference
And when did any CIO have any knowledge of the real world at the coal face?
Intel are just deluding themselves.
Gartner analyst Mikako Kitagawa said CIOs still generally prefer Intel's Core processor family to AMD's Ryzen line
That probably due to a lot of CIOs not knowing what they are looking at.
...and that Chipzilla dominates the B2B portfolios of Lenovo, HP and Dell
I have long suspected Intel is up to it's old tricks again by giving PC builders discounts and back handers to not stock AMD.
"Adding AMD PCs to their PC portfolio would increase PC maintenance costs,"
Most companies are not going to stock spares anyway so I doubt it would increase maintenance costs.
I have long suspected Intel is up to it's old tricks again by giving PC builders discounts and back handers to not stock AMD.
That's my feeling, and what's happened every time AMD is thrashing Intel in technical performance.
Most companies are not going to stock spares anyway so I doubt it would increase maintenance costs.
I did read this and wonder. Who actually stocks spares at component level? I stock maintenance spares of entire desktops so I can just swap a dead PC with a working one and fix the dead one at my convenience. If the hardware is dead and it's out of warranty then you stick it in the "for disposal" pile and put a computer that's just been replaced via your replacement program that still works fine in the pile of maintenance spares.
When it comes to servers, pretty much the same story. If one of the CPU sockets does go then you failover to the backup and phone the support line and ask for an engineer to bring and fit a replacement part.
Fast moving organisations are ditching Intel left right and centre. Why beholden to higher prices, weaker performance and supply chain issues? When there's a ready-made alternative right there.
Of course, we all know that the supertanker-like organisations of this world with only teaspoons so steer them cannot nearly react so quickly. Why else is Win7 still on probably 75-80% of "big business" PC's? Maybe more? And Intel? Very large chunk of the market - even if it's legacy installs. There's a shipping container outside my office full of older Intel/Lenovo laptops that will be built, re-issued, and rebuilt-and-reissued for quite some time to come before any change in the hardware inventory will come to light.
There is a warped part of me that would like to see a genuinely widespread attack against the much talked about vulnerabilities, rather than just minor damage here and there. A good, hard slap might be just what the world needs to wake up.
"She claimed demand for enterprise PCs should fall this year, with the Windows 10 refresh cycle coming to an end, so Intel shortages may become less of an issue."
"It's okay, guys- demand is drying up anyways, so the fact that we can't deliver will stop mattering."
Not sure how this is supposed to make people feel good about Intel.
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2024
