Aww yisss, finally, tangible performance gains and huge value increases for consumers in the CPU space. Intel getting curb stomped is just a bonus.
I'm worried about AMD since AFAIK, the enthusiast PC builder/Gamer for who these chips are addressed is a niche market and most volumes are in OEMs laptops (Dell, HP, Lenovo) that companies buy in bulk for their workforce and that's where AMD has zero presence while intel has a deathgrip on those OEMs.
Everyone I know building a new PC is using an AMD chip. It's cheaper for the performance. In addition, if the gains (both energy and performance) can be moved to the mobile CPUs, then most people would switch there as well.
In terms of OEMs, they have some super computers being built with their CPUs, Google using their GPUs for their gaming engine, and I suspect with that confidence placed in AMD we'll see even more. Just recently, Lenovo also released the T495, which has an AMD chip.
I suspect the move to AMD will be slow, but steady until Intel has an answer for this. AND even if Intel has an answer (10nm), they have had several large security bugs. I suspect many companies and cloud providers will move to or at least explore a more diverse set of chips just because of that. This means more people trying the AMD chips, at the very least.
I've been thinking about getting a T495, so I'm watching for reviews and whatnot (esp since the A495 had throttling issues), as well as watching for news on Zen 2 mobile processors.
It would've been nice to get the product launched in time for the fall, but hopefully we'll see it launch soon.
The T495 may beat the T490 in most things (if other laptops are any representation) but now is not the time to buy it in the U.S. For us, there’s only 3 prebuilt models to choose from as customizing isn’t an option yet.
I wanted to get one as well, however I might hold off until the next gen as there’s a chance of a Ryzen 2 chip and USB 4. And maybe a bigger battery ;)
I’d be curious of how the builds using the 3700 are doing, though. They pack a punch.
>I’d be curious of how the builds using the 3700 are doing, though. They pack a punch.
If I'm looking at user-submitted benchmarks properly, there's only a very slight difference compared to the 3500U. Seemingly not worth the upgrade, or am I mistaken about this?
The Australian store did not have the customisation options I wanted, so I purchased mine via the Hong Kong store and got it delivered to a family member there.
Be careful though - the only way I got my E585 stable was to add idle=poll to my kernel parameters (which means I get pretty poor battery performance).
I asked Lenovo support about this, and the rep said a customizable version would be available withing two months. I'm in Canada, but I imagine the US won't be later than here.
I looked at the T495(s) and all the best AMD-based laptops recently. They aren't that good, certainly not better than Intel-based offerings, except for the improved integrated graphics. notebookcheck.net has some decent comparisons.
But the laptop parts (3700U) are still old-gen CPUs, not the ones everybody is happy about this year. Hopefully in 2020 we get some decent 7nm options with Thunderbolt, HiDPI screens and LPDDR4.
While they may be cheaper for performance if you compare just cpu, but if you include a GPU then it isn't that clear. The fastest ryzen with integrated graphics card is 2400g and it's only 4 cores. That's an extra $75-100 for the you.
I had 2400g and it isn't cutting it anymore. And I don't need a gaming video card to work with visual studio at 4k@60.
If you're spending $300+ on a CPU, you're probably buying a GPU as well. And if you're not, and you just need something that's as good as onboard graphics you can probably get a old card for $20-30 on CL.
My last two builds were an i7-7700k desktop and an i3-8100 FreeNAS (FreeBSD based) build. The i3 will become a Linux box for other purposes once cheap Xeon Es become available on eBay.
The onboard graphics support is great. Now I can use my PCI slots for a 10gb SFP+ backbone and some other stuff.
Watch out for 3600G then. I don't know if there's a concrete plan to launch it immediately, but it looks like the additional $20 on it may give you what you need. The Navi should do 4k@60.
I realize that this proofs you from getting a PCIe 4.0 GPU of your choice much later, when you have the budget for it.
Of course, don't expect it to give you ultra level graphics for that small die size and heat dissipation.
I'm somewhat skeptical about the 3600G. The 2400G is already a bit bandwidth starved. Dual 64-bit DDR4 channels simply can't keep up with the CPU plus a large GPU. Adding channels isn't possible without changing sockets. There doesn't appear to be enough room on package for an HBM stack like Intel did with Hades Canyon. Running your RAM at 4400MHz with that 0.5x Infinity Fabric multiplier could probably allow a bump from 11CU to 16CU (or 20CU at lower clockspeed). A decent bump and enough to eat the entire low-end GPU market, but not incredible.
I'm left wanting a heterogeneous next-gen system using shared HBM for everything.
Intel currently sells the i7-8809G. It contains a substrate, 4GB HBM2, 24CU Vega M (14nm), and quad-core (3.7-4.7GHz, 14nm) with a TDP of 95w for an MSRP $546. If the substrate and HBM cost $100, then that's MSRP $446 for quad core plus 24CU GPU.
AMD currently sells the 2400G for $130 (launch MSRP for 3400G is $150). The 1300X (100MHz slower, but double the cache) costs $110 giving us around $20 for the 11CU GPU (the 2200G can be had for about $90 indicating the GPU price may actually be lower than that).
Let's build a next-gen APU for both desktop and gaming laptops.
HBM3 spec was finalized almost 3 years ago. Estimates for Vega put 2 HBM2 chips (8GB) at around $150 with a substrate adding another $25. We'll actually work within the constraints of HBM 2 at the moment. Just realize you should get 2x the RAM for around the same price (die shrinks and advances in 2.5D chip construction) along with lower voltages, higher speeds per pin, and potentially fewer stacks needed. We'll use HBM because it has enough bandwidth and uses simpler memory controllers while also cutting latency in HALF vs DDR [latency source](http://www.pdsw.org/pdsw-discs17/slides/stocksdale-pdsw-disc...).
We'll need 16GB of HBM2 at a cost of $300 (remember, that should become 32GB of HBM3 or lower costs closer to $150 for 16GB). If we cost $20 for 11CU, we'll assume $100 for 40CU (adding around a 25% markup for the larger chip though 36CU and 24CU lasered versions could greatly reduce that). That puts us up to $400. A 3600 6-core CPU is selling at $200. We'll add another $50 to the MSRP just because for a final price of $650.
That is $100 more than the MSRP for the i7-8809G, but has a TON of advantages. It has 2 more CPU cores and almost 2x the GPU power (with access to more RAM if it needs). There's no need to spend another $100-200 for additional DDR4 plus the CPU performance will be better due to reduced latency. The lack of RAM on the motherboard and getting rid of all those traces should reduce motherboard cost and size. Finally, at 7nm, this should all fit inside a 95w TDP suitable for both desktops and gaming laptops (which usually have 35-45w CPUs plus 50w+ GPUs).
But it’s up to the manufacturers to decide what ports to include.
PS: That said it does seem like there is a premium associated with DisplayPort, are you sure you are not also paying that premium when looking at Intel motherboards?
GPU is compute. The list of use-cases that require lots of CPU and negligible GPU performance are shrinking by the day. Video and photo editing are heavily GPU accelerated, as is 2D and 3D CAD/CAM. Scientific computing is increasingly reliant on GPU performance and even your browser is GPU accelerated. I suspect that an increasingly large proportion of developers will start demanding fast GPUs as the usefulness of low-precision computing expands.
It just doesn't make sense for AMD to waste a lot of very expensive die area on a crappy GPU that most of their customers will never use, for the benefit of a very small minority of customers who resent paying extra for a cheap dGPU. The Zen 3 APUs should offer up to 8c/8t or 8c/16t with a fairly capable Vega GPU, which is a far more compelling value proposition for the vast majority of consumers.
They announced the 3000-series APUs, but it looks like they're basically the same generation CPU-wise as the 2000-series CPUs, so I guess 7nm APUs with more cores will be sometime next year and be called 4000-series.
The new ones are zen+, so about 5% faster per clock due to cache improvements. They are also 12nm instead of 14nm. The other zen+ chips shrunk the transistors, but left gaps between them so the physical size was the same, but with a GPU shrink, a few more months, and all the 3xxx mobile chips also being zen+, they may have gone ahead with a physical shrink too.
I find this is a good move by AMD. Have 1-5x CPU models with IGPs, and keep the lineup clean and compact. Not like Intel, with their, what, 50 iterations of different types per generation? I think it could even be hundreds.
I'm not sure if it was part of their business strategy to overwhelm, but they sure did, and this didn't make them any more sympathetic, on the contrary...
Realistically developers are the only niche that needs as much CPU as they can get with virtually no GPU requirements. All other CPU-intensive applications tend to heavily profit from powerful GPUs.
> I'm worried about AMD since AFAIK, the enthusiast PC builder/Gamer for who these chips are addressed is a niche market and most volumes are in OEMs laptops (Dell, HP, Lenovo) that companies buy in bulk for their workforce and that's where AMD has zero presence while intel has a deathgrip on those OEMs.
Intel's deathgrip rests on four factors:
- Intel's "deal sweeteners" (preferential pricing et al), which they can't afford to do forever.
- Intel's ability to provide the whole stack (chipsets->boards).
- Corporate IT's sclerotic processes (not always a bad thing BTW though personally I find it stifiling).
- Server chip architecture.
It will all come down to money. If AMD parts have good performance/cost/power consumption points, you'll get AMD laptops (is thunderbolt 3 silicon available for AMD? I don't know).
If they beat Intel on server performance why wouldn't everybody switch? AMD doesn't appear to have the full AVX instruction set but I'm not sure how many people really use that.
Corporate IT's process is mostly about money, holistically (does it give us volume pricing? Does it reduce support cost?) as corporate IT is always a cost center. So they'll switch if it's worth it.
So you're right that Intel will continue to be dominant for a (relatively) long time, but for the reasons above that's more about market hysteresis than fundamental superiority, which Intel still has in certain places but which has eroded mostly across the board for a long time. And the kicker of hysteretic phenomena is when you reach the critical point you can't really get back.
That is so far from truth. Cooperate IT is all about not getting fired and doing no wrong. It has very little to do with performance, power, or cost. Your CTO could make an excuse of not buying AMD because it is not stable. That was the reason in the old days AMD not getting any traction.
There is a huge brand stickiness in Corporate purchase, and people are willing to pay premium for it. And that is why everyone want these client.
> Cooperate IT is all about not getting fired and doing no wrong. It has very little to do with performance, power, or cost.
From what I have seen this rule is mostly about software which is vastly more complex than hardware for typical IT departments. So that 'not getting fired' is mostly about software. E.g our company still buy very expensive licenses from IBM, Oracle, SAP and so on. However as far was hardware goes it is super crappy HP laptops despite devs complaining about disk crashing, freezing and myriad other problems everyone still keep getting those.
It’s “available” but hasn’t been adopted yet by most motherboard manufacturers. Maybe we’ll see laptops with TB3 when the 4000-series mobile chips come out.
There's no need to worry about laptop & mobile CPU market share. It's a very low-margin market, and Intel has such deep pockets that they can bleed competitors out if they need to. What AMD needs is server CPU market share. As more computation moves to the cloud, that's where all the money is. It's also in the datacenter where AMD's price/performance and performance/watt advantage makes the most difference for potential customers.
Now that AMD has a solid foot in the enthusiast mindshare with Ryzen and ThreadRipper, it's time for them to strenghthen the EPYC lineup and seriously compete with Xeon.
This. I recently made the leap (financed by work, of course) to "server grade" hardware after years of cobbling together my own boxes from consumer parts. The price difference is phenomenal. Like eating at McDonald's vs. catering a wedding. The idea that essentially one company had this market locked up for all those years boggles the mind.
Well AMD has a "death grip" in the console market. That's kind of their equivalent. If they gain access and spit the Intel monopoly in the laptop market then AMD can become Intel faster than anyone could think of.
Nobody wants the console market in the first place, it's low-margin high-touch garbage. It's a good base-load, but if you could, you would rather sell that wafer worth of chips to datacenter applications or in a glossy box to gamers.
I wonder if the wafer angle is less the case because AMD is fabless. Intel, indeed, would lose the capacity, but AMD can farm out the Xbox One to a second-tier manufacturer or process while still cashing the cheque.
I also suspect there's a market positioning long game in controlling console chips. Every AAA title and major engine for the next 5-10 years is going to target 8-thread AMD CPUs, so you'd expect lazy ports to perform well on Ryzen for free.
It's not so much targeting 8-thread AMD CPUs but targeting multicore x64 CPUs in general. The differences between operating systems account for way more than the differences between Intel and AMD CPUs.
There are a few things to keep in mind. The absolute vast majority of the $1000+ laptops market belongs to Apple. And most of the other laptop CPU are low price and comparatively low margin. AMD is already competing in that area with their APU using GF 12nm, Those 12nm wafer capacity first and foremost goes to I/O chip for Zen 2 especially EPYC, which is where all the money are made.
Their next Gen 7nm APU, will be going head to head against Intel's 10nm IceLake. So while Zen 2 could enjoy its performance lead on Desktop, it is highly likely to lose against Icelake. And Icelake is suppose to have laptop on shelf by this Xmas, I don't expect the 7nm APU to be announced until next CES and product shipping in Q2 2020.
Sometimes I do wish AMD could work out a deal with Apple to use their CPU / APU in their product line. It will have a lock on effect across the industry, if it is good enough for Apple. It is good enough for everyone else.
Uh, Apple has less than 8% of the market[1]. Why do you think it has the "vast majority" of sales? I'm assuming there's confirmation bias here but the vast majority of laptop sales are definitely not Apple, and not even at the high end—even if we exclude corporate sales.
> The absolute vast majority of the $1000+ laptops market belongs to Apple. And most of the other laptop CPU are low price and comparatively low margin
Eh? You make this a statment, but without a source, I really fail to believe it.
I would guess that most laptop purchases are by business, and I would guess that most of those are over $1000ea, and I would further guess the overwhelming majority of those are not Macbooks.
New Ryzen Notebooks are appearing. A friend of mine has one and is pretty happy with it (apart from some touch screen issues on Linux, which have been fixed in the latest kernel).
I don't know much about AMD's OEM strategy, but it seems like they're making some progress.
What incentives can Intel still offer OEMs to go with them?
Not just Thinkpad either, but T series, T is the mainstay for businesses (certainly by volume), that shows a lot of trust, E lines are entry (and largely to be avoided if you can), X is thin/light and P is workstation class (Xeon's and Quadro type stuff) - Though the newer T's aren't that much thicker/heavier than the earlier X's so the lines are blurry.
All I've seen says otherwise. The X1 is more expensive and has a more premium construction. The T4x0s approximates it in some specs and form-factor but is still behind on both. The T4x0s seems positioned so that enterprises can distribute it broadly and get the size/weight advantage without the cost of the X1. At least the enterprise I work for does that, using the X1 only for the executives.
No, the X1 series is Lenovo's answer to the MacBook Air and other ultraportables, whereas the T4XX series is their answer to the MacBook Pro, with the T4XXs series being the more premium variants.
Also (though they don't do it now) the T4xxP which I think I bought the last model they made(i7-7700HQ w/ 14" 2560x1440 and support for up to 32GB of DDR4 - was a little speed demon two years ago and still is really).
I had a low end one like 7 years ago and I was pretty happy with it. The integrated graphics were way better than Intel had back then and it worked pretty well with that GPU accelerated UI Windows got starting with Vista.
Lenovo has at least introduced a bunch of AMD laptops in the Thinkpad lineup this year. T495, X395, E595, E495 at least. These correspond to T490, X390, E590 and E490 which come with Intel CPU.
You'll have a while to wait. Ryzen 3xxx APU and mobile chips use zen+ on 12nm. Zen 3 launches early next year, so you'll probably have to wait until then for zen 2 mobile chips.
Enterprises tend to move slowly, but leaving performance and money on the table is stupid, the small companies will adopt, then it will percolate slowly to larger companies.
"When it comes to gaming performance, the 9700K and 9900K remain the best performing CPUs on the market"
"Ultimately, while AMD still lags behind Intel in gaming performance, the gap has narrowed immensely, to the point that Ryzen CPUs are no longer something to be dismissed if you want to have a high-end gaming machine. Intel's performance advantage is rather limited here – and for the power-conscientious, AMD is delivering better efficiency at this point – so while they may not always win out as the very best choice for absolute peak gaming performance, the 3rd gen Ryzens are still very much a very viable option worth considering."
This isnt to take anything away from AMD and people will finally consider them now as a serious alternative, but I would not call this a "curb stomping"
AMD processors are reasonably cheaper, draw less power and get less hot for extremely competitive performance (especially in multi-core where they exceed).
I guess it depends on where you're defining 'curb stomped'. As a whole, I think AMD is besting Intel here.
Does anyone have any insight on whether AMD being included in the 2-3 major gaming consoles being produced in 2019 will help to sustain marketshare? I believe Microsoft, Sony, and Google have all announced their next gen. consoles / cloud platforms will use AMD.
Dell ships at least 2 systems with Ryzen 'pro' CPUs, and servers with Epyc CPUs. I know because I've ordered the desktop system for our lab and they work fine, and they were a good value.
I'm worried about AMD since AFAIK, the enthusiast PC builder/Gamer for who these chips are addressed is a niche market and most volumes are in OEMs laptops (Dell, HP, Lenovo) that companies buy in bulk for their workforce and that's where AMD has zero presence while intel has a deathgrip on those OEMs.
Can AMD still recover in this market?