Intel brings a six-core i9 CPU to laptops
anandtech.com
anandtech.com
https://ark.intel.com/products/134903/Intel-Core-i9-8950HK-P...
Which means they'll be stuck using LPDDR3-2133 memory with the same bandwidth as previous generations in any power conscious design.
For example, Apple's MacBook Pro line continues to use LPDDR3-2133: https://www.apple.com/macbook-pro/specs/
I wish AMD would add LPDDR4 support to their mobile chips, if only to light another fire under Intel to have feature parity like they did with core count.
LPDDR4 also uses a different physical signaling scheme than LPDDR3 so it's not a minor change to adapt current processors, if it is even possible without a packaging change.
Second question, do modems exist that could handle the bandwidth of existing designs?
It seems like something that would be extraordinarily expensive, and there wouldn't be much of a benefit considering Intel doesn't have competition on this front.
Intel can update their DRAM controller independently of the CPU core microarchitecture, just like they can update the integrated GPU microarchitecture independently of the CPU core microarchitecture. The only reason why they might not be releasing a CPU with LPDDR4 yet is if they never expected to need an LPDDR4 controller before their 10nm process was ready, and never started designing a 14nm LPDDR4-capable controller. If so, that's a clear miscalculation on their part and a sign that the processor architects are probably insufficiently skeptical of what the fab guys are telling them.
I don't even shut down my multiple JetBrains IDEs and gazillion browser tabs or bloated Slack when I take a break and pin the CPU with Ableton Live and a bunch of soft synths on my MBP. Nothing skips a beat.
Is anyone seriously running into issues with only 16GB of RAM?
I showed up at the company and noticed you couldn't run the vm for a test and compile separately at the same time. Next I found out I couldn't have a bunch of tabs open while doing dev. So that was painful, and I got a desktop. The existing people who were used to how things worked said you shouldn't have a bunch of tabs open, don't do that and it works fine (oh and don't run any tests while compiling).
Then as memory use kept up even the at most 4 tabs people found they kept running out of memory and they bought a few 32 gig laptops and suddenly things worked again.
A few of us have desktops, most people are struggling with 16 gig notebooks and they started buying 32 gig notebooks for people that want them.
Where they using them for their own use out side of working hours? not usually a good idea you want to keep your personal device use separate from work equipment.
I'm a data scientist and regularly work with multiple datasets simulataneously that require the RAM usage. Both Python and R rely on in-memory processing. Loading on/off disk is substantially slower and does not fit with what I am trying to do. For really large datasets I also have a 28 core Xeon with 196GB that I can remote into, but it is nice to not have constraints on my laptop.
Of course, you could go with Hadoop or Spark to process some of these datasets, but that requires quite a bit of overhead and its easier (and cheaper) to just buy more RAM
They've probably updated it since then...
You can still write your data analysis code in Python, but you get to leverage multiple machines and intelligent compute engine that knows how to distribute your computation across nodes automatically, keeping data linkage and parentage information, so computation is moved closest to where data is located.
That is the kind of spot where you max out everything you can max out and just go take a break when something intensive is running.
If you still can run on one machine, it's almost always a win. 32Gb is a perfectly reasonable amount of memory to expect. 64Gb isn't outlandish at all for a workstation.
It's severely limited the freedom of range with our dev environment and we're constantly fighting to stay within that 16GB spec. Do we deviate heavily from our staging/prod environments?
What's the sweet spot? A bunch of us have built hackintosh desktops at this point so we can have 32-64+ and more cores so we're not constantly fighting resource contension with all of docker containers we need to run.
[0]: https://www.ifixit.com/Teardown/iMac+Pro+Teardown/101807
That sounds an awful lot like the infamous "640 kB ought to be enough for anybody" quote
> Is anyone seriously running into issues with only 16GB of RAM?
Every day.
You're in luck because your working set happens to fit into RAM, and the rest is written to swap out gracefully and doesn't pull itself back into memory. But as soon as you're actually working with more than 16GB of data at once, you're in trouble.
Of course, we can argue how many of these will be executed from a laptop but there are people who use a laptop as their main rig so I guess every possible scenario is on the table.
Yes, all the time. I wouldn't touch even a laptop with less than 32Gb these days, but YMMV with workload. JetBrain IDE's and Slack are a far cry from volumetric image processing or lots of data science loads.
I should have rephrased my original question - is any significant share of the market running into issues? Because everyone acts like this 16GB limit is something a huge chunk of people currently need.
In my experience it's pretty easy to run up against the 16 GB limit on the MBP if you're running Slack, a browser, a couple of IDEs, and Docker.
Meanwhile, Visual Studio Code uses ~400M when freshly opened and ~550M with the same file (with similar plugins where available). Admittedly, VSC offers far more functionality, but the memory increase is still sizable.
I know that those (Slack and VSC) are vastly different programs with vastly different purposes, but even a minimal Electron app is going to have ~100M baseline memory, which is going to be used again with each and every Electron app that gets launched, in addition to the runtime overhead.
A common response to this is that "RAM is there to be used", but that RAM would have been used anyways for caching (which would have increased overall IO performance across the system) if these apps didn't hog it all. This fact becomes especially relevant when doing tasks that require lots of data (machine learning, compilation, etc).
That being said, I acknowledge that browser runtime based apps make it much easier to develop cross-platform applications, a fact for which I am grateful for as I run Linux. I think that a reasonable solution going forward would be if Electron (or another similar runtime) offered a way for multiple installed apps to share one running application. Ideally, of course, this would be offered natively by the browsers themselves, but given the technical hurdles to doing that _safely_, I'd easily settle for the former.
• Software Development
• 4k+ Video Editing
• High Resolution Image Editing
• 3D CAD
• GIS
• AR/VR
• Data Science
• Machine Learning
• the list goes on…We'd also like a few TB of VRAM as well, but that's another order of magnitude expense...
I've used a machine with 8GB of RAM and a swap partition and hard drive cache on the fastest SSD (Intel Optane SSD DC P4800X). Responsiveness still takes a huge hit when processes are actively using more data than fits in RAM.
Fast SSDs can help when you have more RAM than you need but not as much as you'd like, but they don't help when you don't have as much RAM as you need.
https://www.youtube.com/watch?v=WDIkqP4JbkE&feature=youtu.be...
Intel's profits?
six to 8 iphone's "glued together" would be somewhat appealing.
I already have a thinkpad with 32gb that I use for ML and minikube.
Who knows though, maybe they'll reintroduce the xserve? ;)
With more cores memory bandwidth does start to become more and more of an issue...
"internals of the XPS 15 model 9570 have been upgraded: [...]a fully-unlocked six-core Core i9-8950 HK."
"Since the new processors support DDR4-2666 memory, Dell will equip its new XPS 15 with 8 – 32 GB of DDR4-2666. "
It even scores 4 lanes PCIe Thunderbolt, finally.
(I'm looking at upgrading my original X1 Carbon with a new one in the next couple of years myself so will probably wait for the LPDDR4).
Sigh, the same old intel, they accuse others of selling desktop parts as server cores, but they are the ones that don't have a proper desktop lineup. Rather their focusing entirely on power/thermal constrained mobile parts, and then packing as many as possible into a server part. Desktop users get whatever random dies are left over. At least the new "workstation" series xeon's acknowledge that there are users for which single threaded desktop performance is still important.
So where the market is? Sounds like smart for a for-profit.
If anything, one could accuse them that they didn't do that enough (e.g. missed on the mobile phone market), not that they did it.
Every benchmark i know from intel desktop cpu versus mobile cpu is the same: desktop wins by big margins.
I'm guessing that the mobile version does have less execution units.
Let's compare a 8550u and 8700k. The desktop has a base clock of 3.7 instead of 1.8. it has 12mb cache instead of 8mb. Twice the bus speed. Faster supported memory and more bandwidth. Desktop also has 16 pcie Lanes vs 4 (both can access more via the chipset, but then there's that bus bandwidth issue). Instruction and feature support is almost identical (except some things like vpro).
Sustained clocks, bandwidth, and more cache make a huge difference.
That's why desktop high end CPUs have such a large gap in performance in desktop vs. mobile. The gap seems small if you just look at the specs, but is much larger in practice.
They don't. The architecture of their mobile CPUs is exactly the same as the desktop versions. (They use the same die.) The difference is in the thermal limits, which determine how long the CPU can spend at each power state.
That's disregarding TDP/TPL adjustments/overclocking, of course (which is disabled by BIOS on workstations anyway).
it was pretty interesting!
Core i7-8850H
Core i7-8750H
Core i7-8700T
Just not clocked as high as the i9.
I wonder if any of these 6 core laptop CPUs will have the AMD integrated graphics - it appears not at this moment. I was looking forward to that on an upgrade to my Dell XPS 15.
Who is getting the AMD integrated graphics CPUs then? Apple? I really wanted one in my Dell as I was hoping it would be faster and more power efficient than the NVIDIA 1050/1050 Ti.
Dell are doing preorders on XPS15 2-in-1; HP doing same on the 15" Spectre x360.
Doesn't really fit Apple's line. The 13" probably doesn't have the thermal headroom for it at all, and it's not an improvement on the discrete AMD stuff already used in the 15" MBP. Might work for a low-end 15" MBP, but it's not like it's a particularly cheap part, anyway.
The i7-8750H has the amount of cores and thread it is rated at 2.2-4.2 GHZ. But it's 45 W and 395 dollars. That's a lot of money and heat to pay for a paltry 5% speedup at the very top -- and let's not forget the non Turbo speed is 8% lower.
Later they do some "sleuthing" to quote them and come up with this chart which again indicates turbo across all cores: https://images.anandtech.com/doci/12607/Turbos.png
If anything, most cores are underutilized in laptops even today...
Instead, the major market pressure was on power efficiency.
(My cores are pegged too -- I run DAWs and NLEs -- but me and you are irrelevant in the grand scheme of things, if there are not enough of us. And that's not determined by counting us, but by the emergence of a market).
Anyone try out these Optane chips yet and seen a significant difference? This is apparently the other big announcement coming to laptops...
I also expect most high-end NVME SSDs are going to be in a similar ballpark, maybe "only" 4.5x faster or something.
That said it's quite possible NVMe drives will be "good enough" where PCI-e Optane never makes major inroads, even just in terms of pure volume. The dollar-per-GB needs to come down a lot, still.
Optane in DIMM form is also still MIA. I'm extremely skeptical it will live up to the original hype it was put through ("thousands of times faster") but I imagine it will be able to outclass many competitors.
Intel is also updating their Windows drivers to enable Optane caching of non-boot drives. For the past year, Optane Memory caching has only been usable for the boot volume. This driver update should be available for both the new platforms and for all existing Kaby Lake platforms that support the original Optane Memory implementation, since there's no motherboard firmware functionality that needs to be updated for non-boot volumes.
As a former user of laptop workstations, there is nothing worse than having to always lug around a bulky UPC as well as worry about overheating.
The other reason I could see them doing this is maybe related to wanting to push people to use more graphic related applications (3D rendering), but even that is slowly going to cloud.
That's not... how that works. The "U" indicates a low-power part, while the "H" indicates a higher power part.
For example, the i7-8650U (the U is the important bit) uses 15W TDP: https://ark.intel.com/products/124968/Intel-Core-i7-8650U-Pr...
The i5-8400H uses 45W TDP: https://ark.intel.com/products/134877/Intel-Core-i5-8400H-Pr...
So the i7-8650U will have better power-usage than the i5. Heck, due to binning and Turbo (the "race to idle"), the i7-8650U likely will have better power-usage all around than the equivalent i5-U class.
The i3 / i5 / i7 is just a marketing trick by Intel. The "important" bit is the U, or H on the end of the chip number, telling you whether its a low-power, mid-power, or high-power version. The "U" chips tend to be around 2GHz base, while the "H" chips turbo to 4.5+ GHz and are almost desktop class power-usage.
The most important tidbit about "TDP" is that its "THERMAL design power", not actual power draw. So strictly speaking, TDP measures the size of the heatsink needed to keep the CPU functional.
This means that different chips, even with the same TDP, offer grossly different power consumption rates. Besides, every chip's actual power draw is slightly different, even chips of the same design (see "binning" and "silicon lottery").
-----------
When you understand how chips are made, then you see why things get complicated. Its a well known fact that the entire AMD Zen series uses only two die designs: Zeppelin, and a 2nd design with the iGPU.
That's right, the singular "Zeppelin" design covers AMD EPYC, AMD Ryzen3, 5, 7, and Threadripper (1900X, 1920X, 1950X).
How?? The difference between the dies is in "binning". If one or two cores are broken because of manufacturing defects, AMD sells it as a Ryzen 5 (6-core model) instead of the Ryzen7 (8-core model).
Very broken designs (3 or 4 broken cores) are sold as a Ryzen3 (4-core model). But when manufactured, these are all 8-core chips.
Between chips with full functionality between all eight-cores, some will be able to reach higher clocks than others. The 8-core designs are tested, and the ones that reach the highest clocks are 1800X, and the ones that reach the lowest clocks are sold as a 1700.
And that's how ONE design gets turned into 10 different SKUs sold to the customer. Because the manufacturing process is innately variable.
-------
Intel is known to bin for power-efficiency, high clocks and so forth. The marketing is a way to sell "broken chips" to people who don't care as much about high speeds or low-power draw.
Your caps are misplaced here. The key word here is DESIGN power - inherently an approximation. It's more of a "power category" than an attempt at a rigorous measurement - really they're saying that CPU X needs to be paired with cooler Y. For example, a 6800K and a 6950X are both listed as a 140W TDP, but a 10-core processor is obviously going to pull more power than a 6-core processor under a full load.
Also, what you're measuring can significantly impact the number you get as a result. Measuring at all-core AVX, all-core base, single-core turbo, etc will all give you different numbers. There is frankly more marketing that goes into this than actual technical backing.
If you're implying that heat != power then no, that's incorrect. A CPU is essentially a nearly-perfect resistive load and all power that goes in is converted to heat.
AMD has attempted to spin this one in the past (I believe it was AMD_Robert or AMD_James who made a handwavey post about how TDP was not actually about electrical input but about heat output, as if they are somehow different), simply put they are being misleading. Turboing above average power consumption means that you need to reduce power later to average things out. Otherwise, you need to dissipate a greater amount of heat. That's why it's an average power measurement, and not a hard cap. And it certainly does not imply that "power in != heat out".
edit, found it: https://www.reddit.com/r/Amd/comments/6svy1a/tdp_vs_tdp/dlg8...
Yeah, simply put, thermal watts and electrical watts are the same thing. There is no power that goes into a CPU that is not converted into heat, and there is no heat that is not generated from the electrical input (or the ambient temperature of the room). AMD did not disprove the laws of thermodynamics.
The rest is some handwaving over a formula that he claims doesn't include power, but it's right there in the ϴca term (°C/W), the W is power. It's true that you can sprint above the cooling capacity of your cooler for a short time, but then the heatsink will heat up, and at some point you'll have to reduce power to compensate. Again, that's why it's an average and not a hard cap.
Pretty cringey stuff for an engineer to be spouting and it should not be repeated. Again, there is a grain of truth that TDP is typically not reported as an accurate, measured number but rather more of a general category, but the idea that electrical watts and thermal watts are different things is horseshit.
> How?? The difference between the dies is in "binning". If one or two cores are broken because of manufacturing defects, AMD sells it as a Ryzen 5 (6-core model) instead of the Ryzen7 (8-core model).
This is how virtually all CPUs do it, not unique to Ryzen. What is unique about Ryzen is that all products are constructed from a single die, from laptop to server products, whereas Intel has five: laptop, client, LCC server, HCC server, and XCC server. But within each die, everyone employs die-harvesting to increase yields.
> Very broken designs (3 or 4 broken cores) are sold as a Ryzen3 (4-core model). But when manufactured, these are all 8-core chips.
Side note, but there are typically not enough broken dies to fulfill all the demand for the very cheap SKUs, so many of these are actually fully-functional dies that are locked at the factory.
Back in the day, you used to get Phenom X3s that could be unlocked to the full 4 cores, and you also get GPUs that could have additional shaders unlocked (most recently on Fury, I believe).
There's some Ryzens that are sold with more cores enabled than they're supposed to have, but that is different because they come from the factory like that. Pretty sure it falls into the category of manufacturing errors - someone screwed up and didn't blow all the fuses they were supposed to.
The common use case for these machines: a developer who needs a fast CPU and lots of RAM (and often a dedicated video card) coupled to a 15" or better yet a 17" screen and a ton of hard drive space. You may unplug to go to a meeting for an hour or two, but you'll be back to your desk (and docking station) pretty quickly. You're not working from a coffee shop or a couch because you need a real mouse and a real desk. You occasionally take your work computer home when you're the on-call resource that week.
In that case you don't need a big battery (or rather you need a big battery but not a lot of battery life). Five hours is more than good enough, but you can't trade that performance for anything you don't need.
I know because I have one sitting unused beside me (Thinkpad W530) from back in the days before I started traveling for work.
Was only a few years ago where any decently performant machine didn't get more than 3 or 4 hours anyway. Everyone has a different use case.
I think for most people that's not the purpose of a laptop:
- 80% of the time people will work from a desk.
- It may be a hot desk
- They want to take calls with computer in front of them, but not from their desk (due to open plan offices)
- A couple of times a month they need to take something with a keyboard into a vendor/client meeting/presentationI just want a portable/luggable workstation, not a laptop.
I am typing this on one: it's still technically a laptop, and I can still unplug it from the plethora of peripherals and take it for a walk - it's just that I rarely do that. Still, it does happen every month or so, so there's some value in the laptop factor. (Haven't moved my desktop computer for years, am using both; and on the road, the smartphone is powerful enough.)
You want power efficiency when you don't load the computer much, while you might also want to be able to load it a lot to get e.g. fast builds. If you can get both (and now it seems that you can) it does not make sense to only get power efficiency while never allowing high perf for transient high loads (or even sustained load for when you are plugged in).
The notebook i9 is still an i9 relative to their notebook processors. It's still a 45W TDP processor, just like the other i7 6 core mobile chips.
It may run a bit hotter if you run it at sustained load, but mostly you're paying extra for the higher-efficiency chip that can handle the extra clock speed.
An i7 from the 15W TDP series can still perform slower and use less power than an i5 from the 45W TDP series.
Intel® vPro™ Technology is a set of security and manageability capabilities built into the processor aimed at addressing four critical areas of IT security: 1) Threat management, including protection from rootkits, viruses, and malware 2) Identity and web site access point protection 3) Confidential personal and business data protection 4) Remote and local monitoring, remediation, and repair of PCs and workstations
https://www.cpubenchmark.net/cpu.php?cpu=AMD+FX-8350+Eight-C... https://www.cpubenchmark.net/cpu.php?cpu=Intel+Core+i7-3770+...
I'm not sure in what world anyone could think the FX series ever "dominated" Intel chips at the time, much less current ones, in the datacenter -- unless they are literally delusional.
The i7 was introduced as a 4-core part and only now have they started to shift that to 6-core by default.
If their initial trend with Core continued we'd be at 16+ by now.
Next-gen Intel chips will support LPDDR4, making the whole thing a bit of a non-issue.
My dev team is finally abandoning mac books for lenovo laptops with 32 gig.
How many months it's out from this point is presumably something only Intel knows. Or maybe not even Intel.
The only issue is the missing Spectre / Meltdown silicon fixes. And even when they are available, its probably pure luck to get a model having it. :(