'AMD has 30% of the CPU market for the first time in over a decade...' - Techradar

[UsandThem]

https://www.techradar.com/news/amd-...time-in-over-a-decade-should-intel-be-worried

While none of the sites can know the exact market share percentage, this should be somewhat of a decent indicator of AMD's market share increase. When AMD first began talking about Ryzen, I honestly didn't think AMD had a chance (or that it would be a competitive product). It's really amazing how far they've come in 3 years, and that they were able to gain a sizable market share.

I know there are strong feelings from both camps, but just remember healthy competition is good for all consumers.

 

[jpiniero]

No way does AMD have anywhere near 30%. They'd be making a lot more money if they were.

 

[amrnuke]

No way does AMD have anywhere near 30%. They'd be making a lot more money if they were.

Exactly. Of users on PassMark running benchmarks on their CPU, 30% are AMD users. Which makes sense since they just had a novel release.

 

[Topweasel]

Exactly. Of users on PassMark running benchmarks on their CPU, 30% are AMD users. Which makes sense since they just had a novel release.

I think you guys are missing on how the numbers are collected its not going to be 1:1, and there are Intel dominated markets that this this test wouldn't be run on. But its straight forward a person gets a machine they run tests in the past this means 80-85% of them done on Intel now 70%. We are already seeing markets where AMD is selling 3:1. It pretty conceivable that AMD has grown in share sold to 30% they were never as bad as they were on server, this is a new spike for the new hotness, but it isn't incredibly higher than it had been in the past.

 

[beginner99]

We are already seeing markets where AMD is selling 3:1

true but these are niche markets and in the one that matters for volume (laptop), AMD is far less competitive and their 7nm APU isn't coming anytime soon, liek 9 months.

 

[BigDaveX]

If that chart were accurate to the actual marketshare of Intel and AMD, it would mean that right before the launch of Conroe, AMD's marketshare was nearly equal to Intel's.

Spoiler alert: It wasn't. Even in those days when AMD was curb-stomping Intel in just about every market segment, their market share never got outside of the 20-25% range.

 

[Nothingness]

Spoiler alert: It wasn't. Even in those days when AMD was curb-stomping Intel in just about every market segment, their market share never got outside of the 20-25% range.

And as I like references:

Intel's Conroe processor roils chip market

Intel is planning to launch its "Conroe" chip for desktops this week, triggering a cascade of steep price cuts in the chip market and a raft of new PC designs.

www.arnnet.com.au

AMD's market share by units sold has risen from 19.1 percent in the first quarter of 2004 to 20.6 percent in the first quarter of 2005 and 24.7 percent in the first quarter of 2006, according to research firm Gartner.

 

[soresu]

Spoiler alert: It wasn't. Even in those days when AMD was curb-stomping Intel in just about every market segment, their market share never got outside of the 20-25% range.

In case you weren't aware, back then Intel was engaged in a systematic campaign of bribery and strongarm tactics against OEM's (and retailers?) to get them to keep pushing Intel no matter the advantage AMD had in K8's design.

Intel has less room to be sneaky now because of this, couple that with an obvious lead in power efficiency (which matters to certain customers, myself included) and well priced chips - AMD are simply becoming the DIY choice du jour.

I wouldn't expect that effect to propagate into the OEM and laptop spaces as quickly mind you.

 

[soresu]

true but these are niche markets and in the one that matters for volume (laptop), AMD is far less competitive

How so?

Picasso seems like a solid product to me, if anything the moves Microsoft is making with Surface signals a change there.

 

[scannall]

How so?

Picasso seems like a solid product to me, if anything the moves Microsoft is making with Surface signals a change there.

Intel has done well in the laptop market for a few reasons. OEM's can have just 1 motherboard design for example to cover most of their lines, and 1 bios. And they've done pretty well at battery life. I'm not saying AMD can't break that wall, the way they smashed through other walls. And I think they will. It'll just take a bit of time. Laptops are not a DIY kinds thing, though that might be cool to do. (Hmm, new standards/business plan thing here?)

To really bust through, I think they'd have to provide a 'reference' platform to OEM's. Not just CPU, but a standardized motherboard, ram etc.

 

[chrisjames61]

No way does AMD have anywhere near 30%. They'd be making a lot more money if they were.

Marketshare and installed base are two different sets of figures.

 

[Guru]

They probably are outselling Intel in most markets by 3 to 1. Though I believe Intel still has advantage in smaller markets and smaller countries. But at this point I don't know anyone who is getting Intel over AMD if they can help it.

 

[Ajay]

I'd like the see some report summary from a PC market data company in Jan/Feb and see where 2019 ended.

 

[scannall]

They probably are outselling Intel in most markets by 3 to 1. Though I believe Intel still has advantage in smaller markets and smaller countries. But at this point I don't know anyone who is getting Intel over AMD if they can help it.

Intel still has most of the OEM's locked up. That can and will change over time, but it isn't an overnight thing.

 

[beginner99]

How so?

Picasso seems like a solid product to me, if anything the moves Microsoft is making with Surface signals a change there.

Because AMD is still on 14nm GF which simply is worse than 14nm Intel and it shows mostly in battery life and/or performance. AMD simply made the smart decision to focus on high margin segment (server) instead of mass market (laptop). What we get on desktop is more or less just the trickeled down stuff from server (AMD only needed to do the separate IO die). laptop gets the scraps of the resources. It's always at least 1 year behind.

 

[soresu]

Because AMD is still on 14nm GF which simply is worse than 14nm Intel

Picasso is 12nm, they seem to think 12nm competitive with Intel 14nm:

 

[beginner99]

Picasso is 12nm, they seem to think 12nm competitive with Intel 14nm

What AMD or GF puts on marketing slide is irrelevant. Intel 14nm is clearly better than whatever GF has to offer.

This is easily seen on laptop battery life tests in which intel simply beat AMD even if the CPUs have the same TDP ranking. Why? Because intel idles at lower power use.

 

[LikeLinus]

Picasso is 12nm, they seem to think 12nm competitive with Intel 14nm:

The new 10th gen Intel CPU in the surface is 10nm.

 

[maddie]

What AMD or GF puts on marketing slide is irrelevant. Intel 14nm is clearly better than whatever GF has to offer.

This is easily seen on laptop battery life tests in which intel simply beat AMD even if the CPUs have the same TDP ranking. Why? Because intel idles at lower power use.

In a broad sense, what does lower idle power have to do with node? I would imagine design is more relevant here (gating, etc).

 

[Gideon]

What AMD or GF puts on marketing slide is irrelevant. Intel 14nm is clearly better than whatever GF has to offer.

This is easily seen on laptop battery life tests in which intel simply beat AMD even if the CPUs have the same TDP ranking. Why? Because intel idles at lower power use.

Yes, Raven Ridge was particularly bad in this area (inherently buggy probably). Picasso is a lot better.

But I'm pretty sure the discrepancy @ idle has nearly nothing to do with process. Intel just has years of more experience doing that stuff, has lower power states, does more power gating
Even more importantly, Intel has worked years with OEMs helping to implement better Firmware, BIOSes, etc to keep Idle Power down (not to mention OS optimisations) ...

Just look at power efficiency charts on notebookcheck.net for Lenovos with a really similar chassis (and battery) and both AMD and Intel processors.


T495 vs T490 power numbers from this T495 review and this E495 review (I added the i5 from the search-box below the table).

Power Consumption (W)	Lenovo ThinkPad T495-20NKS01Y00 PRO R5 3500U, Vega 8, Samsung SSD PM981 MZVLB512HAJQ, IPS, 1920x1080, 14	Lenovo ThinkPad T490-20N3S02L00 I7 8565U, GeForce MX250, Toshiba XG5 KXG50ZNV512G, IGZO IPS LED, 1920x1080, 14	Lenovo ThinkPad T490-20N2004EGE I5 8265U, UHD Graphics 620, Toshiba KXG5AZNV256G, IGZO IPS LED, 1920x1080, 14	Lenovo ThinkPad E495-20NE000JGE 3500U, Vega 8, SK hynix BC501 HFM256GDHTNG, IPS, 1920x1080, 14	Lenovo ThinkPad E490-20N8000RGE 8265U, UHD Graphics 620, SK hynix BC501 HFM256GDHTNG, IPS, 1920x1080, 14
Idle Minimum	5.1	2.16	2.34	4.1	2.76
Idle Average	7.7	5.04	5.22	5.86	5.96
Idle Maximum	10.7	8.64	8.82	6.36	6.57
Load Average	35.8	63.2	35.8	30.4	28.2
Load Maximum	43.2	67.5	47.3	50	43.6

I bolded the winning scores for both the T and E chassis separately. Now, some observations:

• So much about "the same " TDP between i5 and i7. i7 T490 uses almost twice the burst power briefly for 10% gain.
• T490 vs T495 Intel with i5 has similar power draw to AMD while scoring 13% better in Cinebench R15 ST burst load and is equal in Multi Threaded one.
• E490 vs E495 AMD in reverse has better idle numbers and slightly worse load ones compared to i5 . This in turn means 25% better MT performance for Ryzen while it's still 15% worse in ST.

Moving on to Battery Life:

In battery life the AMD model actually does better than the Intel one:
Lenovo T495 (Ryzen R5 3500U)


Lenovo T490 (Core i7 8565U)



TL;DR
12nm and 14nm+ actually seem quite similar. AMDs worse Idle number are probably more connected to other factors. We will know better once we have Surface Laptop 3 Reviews (as it's full-stack optimized by Microsoft).

 

[beginner99]

In a broad sense, what does lower idle power have to do with node? I would imagine design is more relevant here (gating, etc).

OF course as Gideon says a lot matters including bios/firmware. idle power for sure also depends on the process but not the specific node size.

 

[Guru]

Intel still has most of the OEM's locked up. That can and will change over time, but it isn't an overnight thing.

It doesn't even matter for AMD at this point, as OEM's are becoming less and less of the market, 10 years ago OEM's accounted for 70%+ of all desktops built/sold, now it's about 50% and declining. More and more people are building their own PC's, the large OEM's are actually replaced by smaller OEM's, a lot of smaller online shops actually have paid services where for just $50 they assemble your parts and send you the PC.

So its not a big deal for AMD, and a lot of OEM's are using AMD more and more.

 

[jpiniero]

It doesn't even matter for AMD at this point, as OEM's are becoming less and less of the market, 10 years ago OEM's accounted for 70%+ of all desktops built/sold, now it's about 50% and declining. More and more people are building their own PC's, the large OEM's are actually replaced by smaller OEM's, a lot of smaller online shops actually have paid services where for just $50 they assemble your parts and send you the PC.

So its not a big deal for AMD, and a lot of OEM's are using AMD more and more.

Haha no, if anything the PC sales report suggest there's a consolidation happening around the large OEMs. DIY is doing better than the consumer market as a whole relatively speaking, but it's still a tiny market. Corporate is where it's at.

 

[Zucker2k]

In battery life the AMD model actually does better than the Intel one:
Lenovo T495 (Ryzen R5 3500U)

Shouldn't the 3700U/3750H be a better comparison here, since you're comparing against an i7 that turbos to 4.60GHz? The Core i7 8565U won't be the most power optimized chip of the Intel lot in that form factor compared to a R5 3500u.

 

[beginner99]

I'm digging this out again. For work I'm looking for a new workstation due to various reasons. Mostly for general ML, data crunching tasks. But also interested in deep learning. Anyway I go looking around and all the common OEMs seem to offer is the xeon + NV option with base config something stupid as a 6-core xeon but already cascade-lake.

Some more niche shops I found some threadripper 2 options. Did not find a single workstation option with an epyc rome option (only normal rack servers) and in general OEMs liek dell or hpe still seem to be deeply entrenched in intel. This is an uphill battle for AMD as if a buyer isn't looking for epyc/TR he will almost certainly not find it / get a product with it.