Okay, but that's not actually a good thing. A current laptop Celeron or Pentium is way faster than an Atom. So a Haswell-based Celeron/Pentium would be faster than Silvermont anything.
It depends. Core based Pentium and Celeron have only 2 cores. Silvermont has 4 cores. Multithreading performance should be way better on a Silvermont 4 core.
Sure, but Intel won't compete with itself, so there will never be Haswell-based Celerons/Pentiums
Intel will do the bare minimum required to make desktop Silvermont fast "enough". Combine that with the inherently low power consumption, and you have OEMs all clamoring for the chance to sell slimmer, sleeker machines to boost their slumping sales. At least buyers will now get a good die scratch-made for efficiency rather than a terrible die harvested for yields.
In a way, we're still advancing forward. Intel just decided to do so on another track. :)
Maybe Intel stops that now at Haswell, but up until now at least, they've been using de-featured (knobs turned off) versions of their high performance CPUs for Pentium & Celeron so they don't have to design a different IC for their lower end. This has been true for Sandy Bridge & Ivy Bridge too.
Given the terrible performance of some of those parts it's not inconceivable that this ends up being more competitive than you might expect. It's certainly less of a big deal than it would have been a few years ago since Intel crippled the ability to overclock the Pentium series; you're effectively getting something performing optimally for its die size instead of being a crippled variant.
The big cores will obviously still be faster per clock, but at 4 cores, turbo up to 2.4 GHz and ~30% IPC improvement over current Atom these CPUs seem like like nice "good enough" candidates. And the big-core Pentiums and Celerons hardly exceed 1 GHz in low power configurations..
Wibtek were the only motherboard manufacturers I could tell that had a desktop Silvermont SKU up on display. It was by chance as well - did not get much time to 'casually' look around this year, especially at the Chinese-only manufacturers, but this definitely caught my eye.
A 2W SDP means TDP will be around 5W (or more if the test involves significant idle time), similar to current ARM SoCs. It seems the 22nm process can't defy the laws of physics...
Should be wrong. Looks like 2W SDP and 3W TDP. Current Y-Ivy Bridge have a 7W SDP and 10W TDP. And also Silvermont should be much much faster than ARM offerings.
3W TDP seems unlikely at 2.4GHz. Anand's measurements on Z2760 show CPU+GPU having ~2W TDP. Now imagine doubling the core count, 33% higher frequency and a faster GPU. The 22nm process helps of course, but the disappointing reduction in active power from SB to Haswell shows it's not magical. So increased performance means increased power.
As for performance, based on the estimates given it looks like Silvermont will have worse IPC than Cortex-A15. Although it may win Sunspider, that's one of the few benchmarks where Atom appears to do well.
Of course not. The real questions are "what performance can you put into a given TDP" and "how much energy is used to complete a given task". TDP tells nothing about these.
Am I the only one wondering about the Z3770 branding? I suppose the idea is that this the top end Silvermont part, like how the i7-3770 was the highest performing Ivy Bridge part? I am unsure for whom this sort of marketing is for (I'd love to see a numbering system like what GPU manufactures use but from CPU companies).
I do not believe that the Z3770 is a Baytrail-T SoC. Baytrail-T lacks SATA and it's specifications lack reasonable functionality (eg a free PCI-e lane) to support the SATA solution shown on several of the motherboard images. More likely, this is a Baytrail-M SoC.
Intel's margins on this product are likely lots better than the feature stripped Core chips Intel traditionally pitched to this segment of the market. Slap this 2W SDP SoC into a tablet with a SSD that supports DEVSLP and this is a potent weapon in the battle for Android device market share. The price is higher than Baytrail-T solution but the SoC is better in ways some folks may be willing to pay for.
today arm is used in multiple times more devices then x86. every single calculator and soooo many more devices have arm chips in them. inside of just about every desktop there is atleast one arm chip. hard drives have arm chips inside of them. so please stop being silly
Yes, arm is the probably the dominant general purpose cpu in the world now.. provided they continue to innovate and offer a whole range of products covering a wide range of performance, power efficiency, cost and integrated features, then Intel have a near impossible task to break their stranglehold.
We’ve updated our terms. By continuing to use the site and/or by logging into your account, you agree to the Site’s updated Terms of Use and Privacy Policy.
17 Comments
Back to Article
flyingpants1 - Sunday, June 9, 2013 - link
Okay, but that's not actually a good thing. A current laptop Celeron or Pentium is way faster than an Atom. So a Haswell-based Celeron/Pentium would be faster than Silvermont anything.mikk - Sunday, June 9, 2013 - link
It depends. Core based Pentium and Celeron have only 2 cores. Silvermont has 4 cores. Multithreading performance should be way better on a Silvermont 4 core.mavere - Monday, June 10, 2013 - link
Sure, but Intel won't compete with itself, so there will never be Haswell-based Celerons/PentiumsIntel will do the bare minimum required to make desktop Silvermont fast "enough". Combine that with the inherently low power consumption, and you have OEMs all clamoring for the chance to sell slimmer, sleeker machines to boost their slumping sales. At least buyers will now get a good die scratch-made for efficiency rather than a terrible die harvested for yields.
In a way, we're still advancing forward. Intel just decided to do so on another track. :)
Hector2 - Monday, June 10, 2013 - link
Maybe Intel stops that now at Haswell, but up until now at least, they've been using de-featured (knobs turned off) versions of their high performance CPUs for Pentium & Celeron so they don't have to design a different IC for their lower end. This has been true for Sandy Bridge & Ivy Bridge too.JDG1980 - Sunday, June 9, 2013 - link
I suspect these Silvermont "Pentiums" and "Celerons" will be slower than the Ivy Bridge Pentiums and Celerons on the shelf now.Spunjji - Monday, June 10, 2013 - link
Given the terrible performance of some of those parts it's not inconceivable that this ends up being more competitive than you might expect. It's certainly less of a big deal than it would have been a few years ago since Intel crippled the ability to overclock the Pentium series; you're effectively getting something performing optimally for its die size instead of being a crippled variant.MrSpadge - Monday, June 10, 2013 - link
The big cores will obviously still be faster per clock, but at 4 cores, turbo up to 2.4 GHz and ~30% IPC improvement over current Atom these CPUs seem like like nice "good enough" candidates. And the big-core Pentiums and Celerons hardly exceed 1 GHz in low power configurations..IanCutress - Sunday, June 9, 2013 - link
Wibtek were the only motherboard manufacturers I could tell that had a desktop Silvermont SKU up on display. It was by chance as well - did not get much time to 'casually' look around this year, especially at the Chinese-only manufacturers, but this definitely caught my eye.Wilco1 - Monday, June 10, 2013 - link
A 2W SDP means TDP will be around 5W (or more if the test involves significant idle time), similar to current ARM SoCs. It seems the 22nm process can't defy the laws of physics...mikk - Monday, June 10, 2013 - link
Should be wrong. Looks like 2W SDP and 3W TDP. Current Y-Ivy Bridge have a 7W SDP and 10W TDP. And also Silvermont should be much much faster than ARM offerings.Wilco1 - Monday, June 10, 2013 - link
3W TDP seems unlikely at 2.4GHz. Anand's measurements on Z2760 show CPU+GPU having ~2W TDP. Now imagine doubling the core count, 33% higher frequency and a faster GPU. The 22nm process helps of course, but the disappointing reduction in active power from SB to Haswell shows it's not magical. So increased performance means increased power.As for performance, based on the estimates given it looks like Silvermont will have worse IPC than Cortex-A15. Although it may win Sunspider, that's one of the few benchmarks where Atom appears to do well.
MrSpadge - Monday, June 10, 2013 - link
Of course not. The real questions are "what performance can you put into a given TDP" and "how much energy is used to complete a given task". TDP tells nothing about these.frogger4 - Monday, June 10, 2013 - link
Am I the only one wondering about the Z3770 branding? I suppose the idea is that this the top end Silvermont part, like how the i7-3770 was the highest performing Ivy Bridge part? I am unsure for whom this sort of marketing is for (I'd love to see a numbering system like what GPU manufactures use but from CPU companies).dealcorn - Tuesday, June 11, 2013 - link
I do not believe that the Z3770 is a Baytrail-T SoC. Baytrail-T lacks SATA and it's specifications lack reasonable functionality (eg a free PCI-e lane) to support the SATA solution shown on several of the motherboard images. More likely, this is a Baytrail-M SoC.Intel's margins on this product are likely lots better than the feature stripped Core chips Intel traditionally pitched to this segment of the market. Slap this 2W SDP SoC into a tablet with a SSD that supports DEVSLP and this is a potent weapon in the battle for Android device market share. The price is higher than Baytrail-T solution but the SoC is better in ways some folks may be willing to pay for.
dwade123 - Tuesday, June 11, 2013 - link
ARM's future is only death. And maybe in cheap Chinese knockoff phones.MustafaK - Tuesday, June 11, 2013 - link
today arm is used in multiple times more devices then x86. every single calculator and soooo many more devices have arm chips in them. inside of just about every desktop there is atleast one arm chip. hard drives have arm chips inside of them. so please stop being sillyspeculatrix - Tuesday, July 23, 2013 - link
Yes, arm is the probably the dominant general purpose cpu in the world now.. provided they continue to innovate and offer a whole range of products covering a wide range of performance, power efficiency, cost and integrated features, then Intel have a near impossible task to break their stranglehold.