For those talking about Dota 2, this is it running on HD 5300 on a Core M 5Y70.
https://www.youtube.com/watch?v=rNI07tAUVh4
10-20fps, but that is running on just 4.5W....
Just realized they used 2+3 (GT3) Haswell Iris Pro die picture that is 180mm2.
The gains that qualcomm, apple, nvidia, and AMD are putting out will leave this GPU in the dust. 20% after 1.5 years is not going to cut it. At this point I'm hoping for some sort of magic driver update.
Next year's mobile graphics probably will pummel whatever Intel graphics there is.
You have insider infos about the future? Share with us.
Considering how poorly Intel are doing despite their supposedly awesome 14nm process, and that mobile guys aren't letting off steam at all, I'd say the chances are pretty large.
Considering how poorly Intel are doing despite their supposedly awesome 14nm process, and that mobile guys aren't letting off steam at all, I'd say the chances are pretty large.
I find it amazing how guys like Nvidia put out so much better perf/watt gains on their GPU at the same process generation than Intel did with a process jump. So much for their process lead, when they can't take advantage of their supposedly marvelous tech.
Somebody needs to tell me how the "40% transistor performance gain" they claimed with their 14nm process isn't showing in any of their product lines, even in embarassingly parallel GPUs. That means a) Intel's process sucks b) Intel's process is good but the CPU is horrible. If their CPU sucks and process is good, than how much better are competitor designs if they were on Intel's 14nm? If Intel's process sucks and CPU is ok than they should wrap up their process guys and go send it to TSMC/Samsung.
Do you want a real proof of what a good product was? Core 2. It was genuinely faster than everything else. Synthetics, Games, "Competitor optimized" applications, 64-bit, you name it. The power use was awesome. Another decent one was Sandy Bridge, it was even priced right!
The Sisoft GPGPU scores are very good from Skylake, low clocked old ES already faster than much higher clocked Haswell.
You have no infos on what Intel is doing. Considering that Intel is going from PowerVR to its own Gen9 in their smarpthones Soc by nex year, there could be some good reason for this change, otherwise they better should rely on PowerVR.
Really? Tell me Intel. What makes Gen 8 so different from Gen 7? Gen 4 was radically different going from fixed function to a unified pipeline. Performance often sucked having to resort to SOFTWARE rendering but at least it was an exciting change.Broadwell graphics "dwarf any other silicon iteration during my tenure, and certainly can compete with the likes of the gen3->gen4 changes."
- Can you get more specific on that? Numbers?
Fair point, but I've heard similar arguments for Gen 8 last year.
elemein said:The iGP is a different story and the performance disappointments there are easily attributed to lack of experience.
Intel easily claimed 2x improvements in perf/watt with their 14nm process. I remember that they even said Core M's specific improvements will result even better, in fact, 2x over traditional non-optimized 14nm of their own! At least we were led to believe.
If even the 2x number was true, we would have the GPU at least being 2x faster. Especially when things are so thermally bound nowadays.
You can't say 20% on a next GPU, even on the same process, let alone next gen, is good at all. Did they know that they messed up Broadwell so bad they want to "save up" gains for 14nm on Skylake? Or maybe they want us to see it that way? Or maybe they just don't have the expertise anymore.
-"14nm is really impressive, trust us"
-"Gen 8 is revolutionary"
You have no infos on what Intel is doing. Considering that Intel is going from PowerVR to its own Gen9 in their smarpthones Soc by nex year, there could be some good reason for this change, otherwise they better should rely on PowerVR. They have to work on low power efficiency. To me it looks like Gen9 is Intels real next gen (because of important missing features in Gen8 like HEVC Hardware encode/decode and FL_12). The Sisoft GPGPU scores are very good from Skylake, low clocked old ES already faster than much higher clocked Haswell.
-snip-
You are just about the only person seeing this.Considering how poorly Intel are doing despite their supposedly awesome 14nm process
That's not how transistors work.If even the 2x number was true, we would have the GPU at least being 2x faster. Especially when things are so thermally bound nowadays.
You know someone is really grasping for straws, when they resort to straw men...-"14nm is really impressive, trust us"
-"Gen 8 is revolutionary"
I think Broadwell-U looks great, personally.Remember, if Intel had its way, Broadwell would be well into its life cycle at this point. It looks not-so-great because it's oh-so-late.
Considering that Intel is going from PowerVR to its own Gen9 in their smarpthones Soc by nex year, there could be some good reason for this change, otherwise they better should rely on PowerVR. They have to work on low power efficiency.
http://www.cpu-world.com/news_2014/2014060901_Some_features_of_Skylake_graphics_architecture.htmlSkylake GPUs will be built on "Generation 9 Low Power" architecture
Intel easily claimed 2x improvements in perf/watt with their 14nm process. I remember that they even said Core M's specific improvements will result even better, in fact, 2x over traditional non-optimized 14nm of their own! At least we were led to believe.
If even the 2x number was true, we would have the GPU at least being 2x faster. Especially when things are so thermally bound nowadays.
You can't say 20% on a next GPU, even on the same process, let alone next gen, is good at all. Did they know that they messed up Broadwell so bad they want to "save up" gains for 14nm on Skylake? Or maybe they want us to see it that way? Or maybe they just don't have the expertise anymore.
-"14nm is really impressive, trust us"
-"Gen 8 is revolutionary"
That's a GT3 U die, so its either HD 5000 or Iris 5100. The GT3e, or Iris Pro, is at ~250mm2. The GT2 Quad Core CPU Haswell is at 180mm2.
I believe a Tegra X1 is already at the point of 15W U's graphics performance. Next year's mobile graphics probably will pummel whatever Intel graphics there is.
That's a GT3 U die, so its either HD 5000 or Iris 5100. The GT3e, or Iris Pro, is at ~250mm2. The GT2 Quad Core CPU Haswell is at 180mm2.
Do you have a link or source for a claim from intel o a 2x increase in performance per watt? My recollection is that the claim was a 30% increase.
Do you have a link or source for a claim from intel o a 2x increase in performance per watt? My recollection is that the claim was a 30% increase.
30% less power consumption at the same performance as 22nm.