That is highly unlikely. Ring bus is pretty much signal time of flight delay and basically distance minimal. In contrast, dual die Ryznen is going to have hit edge of die, hit a phy layer, transit basically board trace, hit phy layer, transit from edge of die, etc.I was under the impression the latency hit Ryzen gets is less than the ringbus drawback xeon faces for those 8c plus cpu?
They buy Cannon because that's what the pro's buy because there is a lot more to a camera than just the sensor, esp since pretty much all the top end cameras aren't sensor limited anymore. For one, Cannon tends to have a lot better features and support than Sony, much much better lens availability, etc. AKA Sony simply cannot compete with the existing Cannon ecosystem (or Nikon for that matter). Sony will have to remain cheaper and competitive for literally decade in order for the eco systems to equalize.I think Canon is a strong example. Sony undoubtedly makes by far the best sensors as shown by dxo marks numbers. Now there is far more to a camera than just the sensor and ergonomics, old glass, whatnot. That can be perfectly rational argument for buying a new Canon body. Yet i think most consumers still buy out of old habit and canon sales is therefore nr one. Customers is by nature conservative and reactionary.
thanks man.Latest leak here .
Possibly more than that (new cache structure and more cache than regular SKL/KBL, 4-channel DDR4) but I'm being conservative till we get some more substantially leaks.thanks man.
Expectations on SKY-X performance difference vs Broadwell-E? Will it be similar to the mainstream cores, skylake was only like 5% faster than broadwell-c?
Yeaa makes sense. But where does scalability matters for servers regarding cpu in the current and future market? - is it 16 32 64 128 256...9096 clusters?That is highly unlikely. Ring bus is pretty much signal time of flight delay and basically distance minimal. In contrast, dual die Ryznen is going to have hit edge of die, hit a phy layer, transit basically board trace, hit phy layer, transit from edge of die, etc.
You are woefully behind times to make that camera analogy and expecting people to believe it. I suggest a look over at dpreview to see how fast the reality is changing.They buy Cannon because that's what the pro's buy because there is a lot more to a camera than just the sensor, esp since pretty much all the top end cameras aren't sensor limited anymore. For one, Cannon tends to have a lot better features and support than Sony, much much better lens availability, etc. AKA Sony simply cannot compete with the existing Cannon ecosystem (or Nikon for that matter). Sony will have to remain cheaper and competitive for literally decade in order for the eco systems to equalize.
A lot of the same attributes apply to computers and esp servers as well. The eco system shift take quite a while.
Well it shows even enthusiast and pros are conservative in buying behavior.You are woefully behind times to make that camera analogy and expecting people to believe it. I suggest a look over at dpreview to see how fast the reality is changing.
www.pcper.com/reviews/Storage/Intel-Optane-SSD-DC-P4800X-375GB-Review-Enterprise-3D-XPoint/Write-Pressure-ConclusiThe P4800X is a beast. I mean seriously. We're talking 1/10th the latency of the fastest competing NAND products and 10x the IOPS performance at lower Queue Depths. Endurance is multiple times higher than anything with NAND flash in it. It's just a monster. The only catch? Software needs to catch up a bit in order to realize the full potential here, but that is very much a solvable problem, and future iterations of XPoint will be packaged to speak to the CPU directly via DIMM slots, further removing the legacy bottlenecks associated with modern OS kernels. As it stands now, it looks like the P4800X has certainly met Intel's expectations for XPoint in PCIe NVMe form. Now we can all start waiting to see what XPoint DIMMs are capable of!
They still buy it for picture quality. For all intents and purposes, these is no quality difference between the various high end camera bodies (what differences that exist are well below the noise floor these days that even professionals care about). What there is, is feature differences. Canon is ahead both in continuous and burst picture rates, lens availability, etc.Well it shows even enthusiast and pros are conservative in buying behavior.
When they originally bought the Canon dslr it was because of picture quality.
It takes time because a camera body is a *minor* investment for most people actually doing or into photography. Any even mild enthusiast has spend multiple times more just in lenses they any given camera body. And that is but just one category of accessories for cameras.But it takes time. Sony is catching up slowly in sales number. But look at the difference in technology. The basic product quality play a minor part for many users. Thet simply play safe and always come late to the party. Most people have that profile.
Scalability will always matter. As far as flexibility, they way pretty much everyone is handling that for the cases that work is based having lots and lots of servers available via various cloud vendors. In other cases, flexibility comes from being able to have large single image system (aka in memory DB). Multi-die sockets doesn't really increase the flexibility for the customer, it is basically a cost saving methodology for the manufacturer that comes with some inherent draw backs.Yeaa makes sense. But where does scalability matters for servers regarding cpu in the current and future market? - is it 16 32 64 128 256...9096 clusters?
I mean from an investment point of view what you want is something where you can just add extra capacity or move capacity as you fit. Flexibility reduces risk. Risk equal money.
I dont think its actually different from production or renting hotel rooms.
What gives you bad sleep is the risk of unused capacity.
Core i5-8250U entry popped up at Gfxbench: https://gfxbench.com/device.jsp?D=Intel(R)+Core(TM)+i5-8250U+CPU+with+HD+Graphics+620&os=Windows&api=gl&benchmark=gfx40
This is most likely KBL-R because as we know 8th generation will be KBL-R and HD graphics naming didn't change in this entry. Base frequency seems to be 1600 Mhz and it features 4 threads - looks like Core i5 gets 4 cores without HT for KBL-R.
That's true but I would think that the die size would have to be bigger than 143 or so if it had 12 MB L3.Traditionally Intels Core i5 features 6MB L3, 1.5 MB per Core. 4/8 Coffee Lake also gets 1.5 MB per Core according to Sisoft, although it's surely not Core i7 labeled anymore, so it might be expected. Only Core i7 gets 2MB L3 per Core traditionally.
Definitely KBL-R. Nice to see it up an running, hopefully we will see it in actual products later this year.Core i5-8250U entry popped up at Gfxbench: https://gfxbench.com/device.jsp?D=Intel(R)+Core(TM)+i5-8250U+CPU+with+HD+Graphics+620&os=Windows&api=gl&benchmark=gfx40
This is most likely KBL-R because as we know 8th generation will be KBL-R and HD graphics naming didn't change in this entry. Base frequency seems to be 1600 Mhz and it features 4 threads - looks like Core i5 gets 4 cores without HT for KBL-R.
Looks like the future Core i7 version. 4C/8T @ 1.9 GHz will wipe the floor with 7500U/7600U's 2C/4T @ 2.7-2.8 GHz in MT applications. Huge boost, also an impressive Turbo. This is just the first taste of 14nm++.Another entry which I believe belongs to KBL-R because of this string : Intel Kabylake R DDR4 RVP
http://ranker.sisoftware.net/show_run.php?q=c2ffcee889e8d5e3d4e1d2e1d4f280bd8dabceab96a680f3cef6&l=en
Genuine Intel(R) CPU 0000 @ 1.90GHz (4C 8T 4GHz, 3.74GHz IMC, 4x 256kB L2, 8MB L3)
This is a 4/8 SKU. Base clock seems to be 1.90 Ghz, Turbo up to 4 Ghz.
That's true but I would think that the die size would have to be bigger than 143 or so if it had 12 MB L3.
If the 4C8T model is the i3 K; that would be pretty crazy but I guess it would make sense that it has 6 MB. The die probably still has 8 MB on it though.
AMD slide at ISSCC 2017 stated that the 8 MB L3 cache is 19.1 mm2 for SkylakeL3 Cache isn't really big: https://www.techpowerup.com/img/15-08-18/77a.jpg
How big is one SKL/KBL core? 10 mm²? 8 MB L3 shouldn't bigger than 5-6 mm², adding 4 MB more L3 would be tiny overall. CFL 6+2 is 149 mm² big according to Benchlife.