AMD CPU sales outpace those of Intel at leading German etailer

[HEXUS]

Beats Chipzilla in both units sold and revenue generated metrics.

Read more.

[DanceswithUnix]

Interesting that the r3 isn't selling in big quantities, wonder if the i3 is the same though as it isn't on these charts despite the r3 being on there at the same price point.

[lumireleon]

LET ME ASK: the R5 and R3 are cheaper than intel counterparts because they lack iGP, apparently iGP in i3 cost upwards of $30? So would you ditch more cores for iGP?

[Tabbykatze]

LET ME ASK: the R5 and R3 are cheaper than intel counterparts because they lack iGP, apparently iGP in i3 cost upwards of $30? So would you ditch more cores for iGP?

No, not until APUs come out anyway. And tbh the A10 series are still pretty good or I'll think about an i3. Considering the low powered nature, if someone came to me just wanting a word processor and browsing, I would send them a NUC which is Intel i-series atm (until AMD Ryzen-Vega APUs). However if I were to be building a full form PC and it were for any form of gaming, I would go Dedicated 100% of the time.

[CAPTAIN_ALLCAPS]

Honestly it doesn't surprise me that the R3 isn't selling too well - at least in Europe. Just about the only reason to buy a cheap CPU and a separate graphics card is a cheap gaming computer, and in games the i3 does pretty well. Plus at that end of the price spectrum there is always competition from either buying a second-hand i3/i5 office desktop and throwing a graphics card at it or spending a little extra and getting the much better R5.

I doubt the R3 will fare too well in the near future either unless AMD release a native quad, as a half-disabled octocore chip is going to struggle to be price competitive versus Coffee Lake i3 quads.

Still it is good to see AMD doing so well with Ryzen. They cannot rest idly though as Coffee Lake is going to make desktop Ryzen's life very hard - unless AMD can get their Vega APUs to run GPGPU loads such as transcoding and game physics to give it an edge - and if Intel get their CPU PCIE links upgraded to PCIE4 significantly in advance of AMD then AMD's big plus on the server/workstation market (where they look like crushing Intel over the next few months) will fall by the wayside.

[lumireleon]

as we have seen AMD is not asleep, Lisa Lu and many others have said the Zen 2 architecture is already baked.

[scaryjim]

Worth noting that the R3s were only released in August, and in their first month collectively sold more units than the i3 7100 in any single month. In fact, it's interesting to see just how few i3s sold compared to every other processor listed, pretty much. I wonder if there should be sales for other i3s on there that aren't being picked up, or whether people just aren't buying i3s any more (perhaps due to the Pentiums getting HT this time round?)

[Xlucine]

...Still it is good to see AMD doing so well with Ryzen. They cannot rest idly though as Coffee Lake is going to make desktop Ryzen's life very hard - unless AMD can get their Vega APUs to run GPGPU loads such as transcoding and game physics to give it an edge - and if Intel get their CPU PCIE links upgraded to PCIE4 significantly in advance of AMD then AMD's big plus on the server/workstation market (where they look like crushing Intel over the next few months) will fall by the wayside.

There's been nothing to indicate that coffee lake will bring any more heat on AMD - gaming benchmarks show that CPU performance (above the very bottom end of the range) is a hygiene factor, with little influence on game performance, and that adding more cores (even to the point of i9/TR) just runs into the same GPU bottlenecks. AMD provides this acceptable gaming performance at a lower cost, and there's every indication that the cost ratio will be further in AMD's favour with coffee lake, so how does it change the maths?

[DanceswithUnix]

There's been nothing to indicate that coffee lake will bring any more heat on AMD - gaming benchmarks show that CPU performance (above the very bottom end of the range) is a hygiene factor, with little influence on game performance, and that adding more cores (even to the point of i9/TR) just runs into the same GPU bottlenecks. AMD provides this acceptable gaming performance at a lower cost, and there's every indication that the cost ratio will be further in AMD's favour with coffee lake, so how does it change the maths?

If CL comes in as basically a factory overclock compared to previous generations then the out of the box performance will look a lot better in some benchmarks even if the cpu isn't really any better. I will be interested to see if the overclocking headroom seems to drop in this generation as Intel eats into what for a while now has been quite a margin. The existence of superpi as a benchmark shows how broken some of the reviews are (apologies to anyone who runs superpi for a living )

[stoatwblr]

"a half-disabled octocore chip is going to struggle to be price competitive"

They don't take fully working octocores and nuke a few cores. These animals are lower-binned devices where the disabled cores were faulty in some way. It's a way of salvaging as much off the fabrication cost of a wafer as possible and the alternative is much more expensive octocores.

As production yields rise, you'll see supply of these cut-down units simultaneously dwindle and decrease in price as the flagship units come down - but they won't decrease as fast as the top end does or sales will cannabilise that top end. AMD would prefer to sell full fat systems and unless they're achieving fantastic yields they won't be disabling perfectly functional cpus on a die simply to sell at a lower price when they can sell them as octos for a lower overall price and still make decent profit.

If yields don't rise, expect to see R3 pricing sink faster. They have to sell what they make - unsold inventory is lost income, but R3 is something that won't be around in large quantities or for a prolonged period if AMD gets things the way it wants them. I wouldn't be at all surprised to hear that R3 sales numbers are already constrained by supply rather than demand.

[CAPTAIN_ALLCAPS]

"a half-disabled octocore chip is going to struggle to be price competitive"

They don't take fully working octocores and nuke a few cores. These animals are lower-binned devices where the disabled cores were faulty in some way. It's a way of salvaging as much off the fabrication cost of a wafer as possible and the alternative is much more expensive octocores.

As production yields rise, you'll see supply of these cut-down units simultaneously dwindle and decrease in price as the flagship units come down - but they won't decrease as fast as the top end does or sales will cannabilise that top end. AMD would prefer to sell full fat systems and unless they're achieving fantastic yields they won't be disabling perfectly functional cpus on a die simply to sell at a lower price when they can sell them as octos for a lower overall price and still make decent profit.

If yields don't rise, expect to see R3 pricing sink faster. They have to sell what they make - unsold inventory is lost income, but R3 is something that won't be around in large quantities or for a prolonged period if AMD gets things the way it wants them. I wouldn't be at all surprised to hear that R3 sales numbers are already constrained by supply rather than demand.

Of course chip companies cut down perfectly good chips to fit lower price point SKUs. Otherwise we would be looking at something like 50% of Ryzen die having 1 or 2 bad cores, yet a relatively small percentage have serious defects in the enormous L3 cache area that would turn them into an R5 1400 or lower. Not to mention so many of Intel's chips having a peculiar condition where the hyperthreading silicon has a weirdly high defect rate.

Sure many of those chips are defective or lower binned, but by no means all.

My bet is that unless AMD releases a native quad the R3 will largely fall by the wayside as a small volume die scavenge for whatever the market rate is and the R5 will be pitted against the Coffee Lake i3s.

[CAPTAIN_ALLCAPS]

There's been nothing to indicate that coffee lake will bring any more heat on AMD - gaming benchmarks show that CPU performance (above the very bottom end of the range) is a hygiene factor, with little influence on game performance, and that adding more cores (even to the point of i9/TR) just runs into the same GPU bottlenecks. AMD provides this acceptable gaming performance at a lower cost, and there's every indication that the cost ratio will be further in AMD's favour with coffee lake, so how does it change the maths?

If CL comes in as basically a factory overclock compared to previous generations then the out of the box performance will look a lot better in some benchmarks even if the cpu isn't really any better. I will be interested to see if the overclocking headroom seems to drop in this generation as Intel eats into what for a while now has been quite a margin. The existence of superpi as a benchmark shows how broken some of the reviews are (apologies to anyone who runs superpi for a living )

Gaming performance on current titles shows that Ryzen will bottleneck high end graphics cards. Not by much, but it is there and it is on current games - especially if you seek reliable 120fps+ performance. At the moment you can make a compelling case that getting the extra cores with Ryzen is probably worth it in the long run. Once Coffee Lake hits that argument no longer holds as Coffee Lake will offer Kaby Lake single threaded performance and solid multi-threaded performance. That will definitely make Ryzen's life hard.

[scaryjim]

... My bet is that unless AMD releases a native quad the R3 will largely fall by the wayside ...

Of course, AMD will be releasing a native quad core - it'll just have a big slab of GPU cores in it as well. The interesting thing will be how they brand the CPU-only derivatives of the APUs (if they release any). We could end up with a situation where a quad-core Ryzen 3 could be either a 2+2 arrangement hewn from a higher range Ryzen chip, or a native quad core hewn from an APU with a failed IGP...

[shaithis]

Of course the i3 doesn't show (or in very low numbers) this is the sales from one retailer.....Statistics are easily skewed.

How many i3's did Intel sell to OEMs that month though I wonder?

[scaryjim]

... How many i3's did Intel sell to OEMs that month though I wonder?

I'm more worried about how many they rebadged as i5s and i7s and stuck in laptops in the same period....

But yeah, the simple fact that the i3 73x0 series aren't in those figures is reason for pause - seems like an odd omission given the fairly comprehensive figures for all the other processor series...

[msroadkill612]

EDIT, maybe a mispost, meant for same author but re using a single zen 4 core ccx as a cpu - half a raven ridge apu e.g.

http://www.anandtech.com/show/11551/...pyc-analysis/2

"EPYC (Naples) Thread Ping Connections
Latency Bandwidth
Within A Core 26 ns -
Core-to-Core, Same CCX 42 ns -
Core-to-Core, Different CCX, Same Die 142 ns -"

So a single ccx could be a little rocket.

It would only give them 16 lanes total tho afaik.