News for hackers. Stuff that matters.
Re:Intel made a huge mistake 10 years ago by Retric on Wednesday April 20, 2016 @07:27PM (#11538492)
Intel's largest mistake was integrating graphics on their CPU. This cost them more than the entire cellphone CPU market is worth.

This ate up valuable chip real estate, RAM bandwidth, thermal overhead etc. Worse it cemented the idea that Intel was crap at graphics while slowing down the PC upgrade cycle.

Re:Intel made a huge mistake 10 years ago by dragontamer on Wednesday April 20, 2016 @07:33PM (#11538519)
What the hell else do you think Intel should put on those chips?

Note that ~15% of Steam Players are gaming on Intel iGPUs, as awful as they are.

http://store.steampowered.com/hwsurvey/

Re:Intel made a huge mistake 10 years ago by chadgeidel on Wednesday April 20, 2016 @07:44PM (#11538588)
And I'm pretty happy about that actually. Having indie games on my laptop with good battery life is preferable to no games or really bad battery life.
Re:Intel made a huge mistake 10 years ago by dragontamer on Wednesday April 20, 2016 @08:11PM (#11538731)
Exactly. I think the more casual gamers find the iGPU to be "good enough" (even if it doesn't match my personal preferences). So if Intel's customers are seeing the benefit, good for Intel.
Re:Intel made a huge mistake 10 years ago by Retric on Wednesday April 20, 2016 @07:39PM (#11538556)
A pure blank space would work as they save on manufacturing costs and get faster CPU's due to heat limits and more RAM bandwidth.
Re:Intel made a huge mistake 10 years ago by chadgeidel on Wednesday April 20, 2016 @07:51PM (#11538630)
No, that's not how it works.

Intel had the ability to put more transistors on the same die size with the same power requirements. This was long after they reached thermal/clockspeed limits (with the P4). They started putting additional cores in there and bumped up the L2 and L3 caches, but there was still space left on the die.

What do you do with those extra transistors? It would be absurd to "leave them blank" as you are basically throwing money down the drain.

Re:Intel made a huge mistake 10 years ago by im3w1l on Thursday April 21, 2016 @12:37AM (#11539730)
Even more cores?
Re:Intel made a huge mistake 10 years ago by masklinn on Thursday April 21, 2016 @03:57AM (#11540317)
Most non-server systems are barely able to use 2 cores let alone 4, what would they do with 8 or 16? More efficient & powerful graphics built-in are a much better use of the die space. Even more so considering the rise of GPGPU and hardware decoding.
Re:Intel made a huge mistake 10 years ago by dragontamer on Wednesday April 20, 2016 @08:20PM (#11538762)
> manufacturing costs

The iGPU needs to be manufactured anyway. And with memory controllers integrated into the CPU, it seems rather complicated to push the iGPU off the chip.

> and more RAM bandwidth

Cheap iGPUs on Motherboards would recycle the CPU's memory controller anyway. Did you work with computers in 2006 or so? Its definitely cheaper and more efficient to just integrate everything into the die in contrast to the designs of the past.

Re:Intel made a huge mistake 10 years ago by Chlorus on Wednesday April 20, 2016 @07:36PM (#11538538)
You realize motherboards had shitty integrated graphics long before Intel decided to put them on-die? How did they slow down the PC upgrade cycle?
Re:Intel made a huge mistake 10 years ago by touristtam on Thursday April 21, 2016 @11:01AM (#11542656)
Intel basically pushed the integrated graphic chips out of the motherboard like they did for the north bridge. I think that was a smart move.
Re:Intel made a huge mistake 10 years ago by Retric on Wednesday April 20, 2016 @07:43PM (#11538584)
Significantly reducing how powerful the next generation of CPU's where.
Re:Intel made a huge mistake 10 years ago by touristtam on Thursday April 21, 2016 @11:00AM (#11542645)
Weird I've seen figures contradiction this in specific cases.
Re:Intel made a huge mistake 10 years ago by pcwalton on Wednesday April 20, 2016 @07:38PM (#11538548)
> Intel's largest mistake was integrating graphics on their CPU. This cost them more than the entire cellphone CPU market is worth.

Er… you mean like every single mobile SoC?

Re:Intel made a huge mistake 10 years ago by Retric on Wednesday April 20, 2016 @07:42PM (#11538580)
These chips are 3D stacked with ram, but produced seperatly.

Which means if one part is defective you can use a different part boosting yields. But, they also get much lower latency.

Now recent Phone Chips have included on die GPU's but phones are so power limited it's not an issue.

Re:Intel made a huge mistake 10 years ago by dragontamer on Wednesday April 20, 2016 @08:14PM (#11538741)
Mobile SoCs use DDR3 or DDR4 RAM, just like Desktops and Laptops.

HBM is only being used on AMD's R9 Fury Graphics Card, as far as I'm aware. A rather high-end solution that simply doesn't exist in the phone market right now.

Re:Intel made a huge mistake 10 years ago by Retric on Wednesday April 20, 2016 @08:43PM (#11538860)
It's not a question of ram type. Physical distance is a large chunk of desktop RAM latency.
Re:Intel made a huge mistake 10 years ago by ajross on Wednesday April 20, 2016 @11:43PM (#11539534)
No it isn't. Physical distance change line capacitance and limits clock rates (or requires higher voltages, same thing). DRAM latency has always been completely dominated by precharge time inside the array; the whole idea of synchronous DRAM was to take advantage of that to stream data across many clock cycles (that is, get a very "wide" interface to that can transfer a whole cache line at once on a few dozen wires) as it's happening.
Re:Intel made a huge mistake 10 years ago by Retric on Thursday April 21, 2016 @09:55AM (#11542125)
You get ~1/2 the speed of light propagating though copper or ~150million meters per second. It's around two feet from CPU to edge of last chip if you trace the wires. So, round trip adds ~1nanosecond to latency. https://en.wikipedia.org/wiki/CAS_latency Shows latency's in the 6.3ns to 13 ns range for DDR3. Thous ~10% of latency is from distance.

PS: I could get more accurate numbers, but with such a wide range of latency number that's pointless.

Re:Intel made a huge mistake 10 years ago by dragontamer on Thursday April 21, 2016 @11:00AM (#11542644)
Do you not know what HBM memory is? You're talking about "stacked memory" and stuff, so I thought you'd know.

As I noted before: Mobile CPUs are using standard DDR3 or DDR4 RAM, on a motherboard drawn with copper wires to the CPU. The ONLY use of "3d stacked memory" that I know of is the HBM RAM of the AMD Fury X.

http://www.legitreviews.com/wp-content/uploads/2015/05/HBM-D...

I'm laying counter to your fact-claims. Only high-power GPUs have stacked RAM right now.

Anyone who has opened up a phone can see the DDR3 or DDR4 RAM on the phone's motherboard, with the phone SOC designed to go "inefficiently all the way" to the DDR3 RAM over "long latency" copper wires.

Read the spec sheet yourself if you don't believe me.

http://system-on-a-chip.specout.com/l/1107/Qualcomm-Snapdrag...

Re:Intel made a huge mistake 10 years ago by Retric on Thursday April 21, 2016 @11:39AM (#11542952)
Yes, it's a similar idea. These terms get repeated SoC used to refer to a single chip CPU, and they keep reusing the term as other components get integrated.

Anyway, I am referring to packaging like: http://cdn.arstechnica.net/wp-content/uploads/2013/04/ASIC_+... Yes, HBM also does more or less the same thing. The issues are twofold heat which is vastly less of a problem in the mobile world so with HBM it's a stack of ram chips next to a processer, with Mobil the CPU is under what might only be one RAM chip and they call it SoC.

http://arstechnica.com/gadgets/2013/04/the-pc-inside-your-ph...

Note: CPU's are already 3d with many layers depending on how you count.

Anyway after die size, heat, latency, and manufacturing costs are really the largest factors. Some PC ram chips run hot enough they really can't be stacked. You can slow them down and trade latency for total bandwidth. ed:Less of an issue with active cooling and a heat sink. However, they can call this HBM, because you can have more chips so even if each chip is slower overall you get higher speeds.

PS: Don't forget HBM is designed for a GPU/CPGPU without a large cache. CPU's have a huge cash which let's them have very different approach aka lot's of cheap but slow RAM. Benchmarks show faster ream really does not add much for a modern CPU with most workloads. You could use this for a CPU but it's harder to swap in more ram.

Re:Intel made a huge mistake 10 years ago by chadgeidel on Wednesday April 20, 2016 @07:46PM (#11538601)
In your opinion what should they have spent that die space on? Moore's law was about transistor count too. They didn't have a lot of architecture changes they could work with and the on-die memory was pretty much maxxed out.
Re:Intel made a huge mistake 10 years ago by Retric on Wednesday April 20, 2016 @07:49PM (#11538615)
Nothing was a real option. Thus significantly reducing manufacturing costs, boosting yields and with a better thermal envelope processing speeds. Remember the last few die shrinks have run into huge heat issues.
Re:Intel made a huge mistake 10 years ago by dragontamer on Wednesday April 20, 2016 @08:18PM (#11538756)
> significantly reducing manufacturing costs

I am pretty sure that if keeping the memory controller, northbridge, and iGPU separate, Intel would have kept these things separate.

My bet? I think the cheaper manufacturing costs associated with motherboard construction beat-out the slightly increased cost of tighter integration. But I'm not an Intel Engineer. Just a hunch... considering that Apple A9, Qualcomm Snapdragon, AMD A10 "APUs" , and Intel have all decided to integrate everything together.

> better thermal envelope processing speeds

What, are people actually using iGPUs for difficult tasks all the time? As far as I can tell, most people surf Facebook all day and ping the CPU.

iGPU isn't slowing down anything with Facebook surfing. Besides, a crappy integrated GPU on the MoBo would still use up RAM bandwidth.

If anything, integrating the GPU has made the process cheaper for Intel. The Northbridge, Memory Controller, and iGPU are all integrated into modern CPUs, because Intel has done the number crunching and has determined that its cheaper to do it this way.

Re:Intel made a huge mistake 10 years ago by chadgeidel on Wednesday April 20, 2016 @07:54PM (#11538648)
"Nothing" (i assume you mean "blank space on the die") does not reduce manufacturing costs. Boosting yields has nothing to do with leaving "part of the die bank". Thermal envelope is inherent to the process size, not the use (or unuse) of transistors on the die.
Re:Intel made a huge mistake 10 years ago by Retric on Wednesday April 20, 2016 @08:01PM (#11538688)
It's more complex than that as you want to avoid local hot spots etc. Also, you can change the layout so a void on one chip is used as part of a second chip. https://en.m.wikipedia.org/wiki/Tessellation. They can use lasor cutters which can trace out complex shapes.

It's much easer to cut rectangular chips though. Still, even if the used the same overall design and just ignore any defects over X area would boost yield.

Re:Intel made a huge mistake 10 years ago by Maken on Wednesday April 20, 2016 @07:42PM (#11538578)
They should instead have integrated a CPU into their graphics card like AMD do with those APUs, which they sell to every consoler maker, and NVIDIA do with their Tegras, for high spec smartphones.
Re:Intel made a huge mistake 10 years ago by touristtam on Thursday April 21, 2016 @11:03AM (#11542669)
I think Intel was hoping to be able to license (virtually for free) Nvidia technology at some point, during the legal battle they had with Nvidia.
Re:Intel made a huge mistake 10 years ago by Retric on Wednesday April 20, 2016 @07:46PM (#11538603)
This can really help CPU-RAM latency which often slows things to a crall. However, modern Intel CPU's have rediculus amounts of cache.
Re:Intel made a huge mistake 10 years ago by pervycreeper on Wednesday April 20, 2016 @07:30PM (#11538501)
Could you elaborate on this?
Re:Intel made a huge mistake 10 years ago by Retric on Wednesday April 20, 2016 @07:37PM (#11538547)
Graphics cards have a small amount of very high bandwidth dedicated RAM close to the GPU. PC ram is further from the CPU which increases latency due to the speed of light lag. CPU's get around this by using a lot of L2/3 cache. Which means PC's don't really care as much about RAM bandwidth. Instead they want lots of RAM.

Thus main memory is much slower than GPU bandwidth. So, now your on chip GPU is memory starved and reducing the bandwidth avalabel to the CPU. There are some advantages if you want to pass stuff back and forth, but the standard rendering pipeline is designed for GPU's.