Have you actually listened to the webcast? Here's the summary of what they did in the previous generations and the webcast was about how they'll improve it with Broadwell:
This is not nomenclature. Intel divides its graphics in (sub-)slices. With Haswell is was 2 subslices of 10 EUs, Broadwell changes this to 3 subslices of 8 EUs.and you guys say AMDs nomeclature is bad, what is 2x24 and 1x8?
Bad eyes, more like. I looked at a different shot of Broadwell, and I take my last comment back now. The cores appear to be more narrow, though.You must have eagle eyes. I can barely make out any details in the first die photograph and I know what to look for.
how ?It seems that Intel has fixed some of the bottlenecks that plagued the Gen 7/7.5 architecture in graphics/gaming workloads.
From what I remember from the Iris Pro review, its performance per GFLOPS was less than other architectures like Kepler, so there was a bottleneck that was not caused by the shaders.how ?
I see very small improvements, only uarch tweaks, not the big architecture redesign that many were expected.
From today's presentation, looks like I'm right that Cherry trail with gen8 GPU won't beat TK1 mobile Kepler...
From what I remember from the Iris Pro review, its performance per GFLOPS was less than other architectures like Kepler, so there was a bottleneck that was not caused by the shaders.
The improvements of Gen8 don't seem to be very big, but maybe Intel is holding information for IDF; there are a few presentations about graphics. But even if Gen8's improvements are not very big, CT still has 4X as many EUs.
What Intel is doing however is reiterating the benefits of their 14nm process in this case, noting that because 14nm significantly reduces GPU power consumption it will allow for more thermal headroom, which should further improve both burst and sustained GPU performance in TDP-limited scenarios relative to Haswell.
I hope you know that 4X EUs doesn't mean 4 times the performance, right ? Too many bottlenecks may occur (RAM bandwidth limitation, resource sharing efficiency, arch scaling, etc)From what I remember from the Iris Pro review, its performance per GFLOPS was less than other architectures like Kepler, so there was a bottleneck that was not caused by the shaders.
The improvements of Gen8 don't seem to be very big, but maybe Intel is holding information for IDF; there are a few presentations about graphics. But even if Gen8's improvements are not very big, CT still has 4X as many EUs.
how ?
I see very small improvements, only uarch tweaks, not the big architecture redesign that many were expected.
From today's presentation, looks like I'm right that Cherry trail with gen8 GPU won't beat TK1 mobile Kepler...
Sure. BT is quite competitive with 2013 GPUs, so 4X as much theoretical performance and a node shrink should put it around a leadership position.I hope you know that 4X EUs doesn't mean 4 times the performance, right ? Too many bottlenecks may occur (RAM bandwidth limitation, resource sharing efficiency, arch scaling, etc)
That might have actually happened if ASML had its EUV ready for the 14nm node.I wish the surface pro 3 I just bought had broadwell. More performance, less heat, less throttling.
I just started reading AnandTech's coverage and I see that, surprisingly, it isn't Anand, CPU editor, who wrote the 2 articles but Ryan smith, GPU editor. It seems Anand is busy reviewing smartwatches.
1x8 for low-end CT.
GT3 will be 2x24
From what I remember from the Iris Pro review, its performance per GFLOPS was less than other architectures like Kepler, so there was a bottleneck that was not caused by the shaders.
The improvements of Gen8 don't seem to be very big, but maybe Intel is holding information for IDF; there are a few presentations about graphics. But even if Gen8's improvements are not very big, CT still has 4X as many EUs.
I hope you know that 4X EUs doesn't mean 4 times the performance, right ? Too many bottlenecks may occur (RAM bandwidth limitation, resource sharing efficiency, arch scaling, etc)
With the great results I've found using a Z3770 Asus T100, I would be tickled with even a 50% bump. A lot of games are just barely playable and would get knocked over the line into good with even a small bump. Exciting times for tiny x86 gaming.
I'm most interested to see how much more performance core M gives over cherry trail- in this segment it feels like CPU's actually matter again.
You are not excited because you probably are thinking from a desktop perspective. Even then, the confirmed scaling on 14nm is a reason to not be disappointed that the trend of 40 years of semiconductor progress is coming to an end.Why can I not get excited about this?
Why can I not get excited about this?
Try not to read into it too much. Intel held this press briefing at their Oregon campus (as opposed to their Santa Clara campus), so it made a lot more sense to send me since I live in the area.I just started reading AnandTech's coverage and I see that, surprisingly, it isn't Anand, CPU editor, who wrote the 2 articles but Ryan smith, GPU editor. It seems Anand is busy reviewing smartwatches.
If we are talk a 25% improvement of gen 8 vs gen 7.5 the 16 eus of gen 8 cherry trail will be equal to the performance of Intel (20 eu) 4200 to 4600 depending on clock speed.
We can not assume this for there may be a bottleneck such as memory bandwidth or CPU speed.
Why can I not get excited about this?
Wait for desktop Skylake, that will surely be something to get excited about. Broadwell is just a preview of which general direction Intel are going, Skylake should be a performance-per-watt monster.