It would be hard to show actual performance difference - but since the 835 is emulation, I would expect that 835 will be significantly slower than these processors when run 835 application. It would be hard to say on .net application - that actually depends on Microsoft implementation of .net on ARM. In any case ,net applications on ARM should be significantly faster than Emulation.
You can buy a motherboard with the Apollo Lake J4205 for under $100 despite that having a list price of $160, so I assume the prices are for show only.
Windows 10 and installed/store apps run natively on the 835, only x86 apps that aren't yet recompiled will require emulation. Even then you'll typically run native code since all the system libraries are native, not emulated.
First of the site - you provided does not have much creditability in my opinion, they are extremely Anti-Bias
Secondary - I know there is some native C++ portions of the code - but a lot of windows code has been refactor with .net - this allows for integration into other platforms easier. But still there is a huge difference between native C++ code and Emulated code.
My guess performance wise the new Intel Atom here will be at same level or better than 835 - likely include in native C++ code and .net.
I heard that Microsoft is coming out with a Surface Phone, likely it will be 835 based.
Biased in what way? Plenty of sites have similar articles showing how Windows x86 emulation works. Are they all wrong/biased too? Is the video detailing how the x86 emulator works somehow faked?
Not all "apps" are native. Some are repackaged Win32 apps. Yes, the various MS libraries ARE native, which is awesome - but there's still a lot of code in an x86 program or "app" which is, well, x86. So the emulation cost is roughly an order of magnitude lower than it would be if everything was emulated, but it's still substantial and varies from program to program. Basically if you want to compare performance of the chips, you're better off testing native-compiled (mostly Store) software. If you're interested in seeing the performance impact of the partial emulation, test x86 programs. :D
Looks suspiciously like the Apollo Lake SoC is using single-channel memory, while the Snapdragon 835 is using double (its memory bandwidth is more than double). With dual-channel, the N5000 should make a much better showing.
Hard to make a direct comparison. The Intel chips have a higher power consumption. I believe the SD835 consumes up to maybe 4W, where all the Intel chips start beyond 4W up to 10W. So the Intel chips will likely perform better (or at least the CPU), but the SD835 will likely offer better battery life.
Theres a 6W variant of the chip that I figured would be put in netbooks 2n1's that would compete against things like the novago. x86 vs the 835 running Windows 10s.
Emulation isnt an issue bc MS/Qualcom have said so, so lets run it like that.
I want to see battery life, typical app usage performance.
1. I thought there were rumors that Intel was discontinuing the Atom series. It appears those rumors were wrong
2. This is first I heard of Gen10 Graphics core - and curious if more details - maybe I missed it somewhere else but latest I heard was Gen9.5
3. It good that they a speeding up these computers - they are significantly cheaper but performance is no where close to main line GPU including Y series which first came out Core M - which for most stuff is well accurate as long as you are not a serious gamer.
I hope this announcement means we here about next series Y processor with 10nm and possible even U series on 10nm. Intel has not updated Y series for 8th generation.
graphics wise, the gen I could be wrong is for example DX9, DX10, DX11, DX12, so maybe when they are saying "gen 9 with some gen10) is saying DX9 for sure, but not all of DX10.
Intel is not "known" for graphics leadership, as long as they can get an image on the screen, it is "good enough" for the heavy lifting, that is where graphics companies come into play.
One company cannot do everything perfectly, Intel might be massive, but, they are still but a small portion of "tech" in general.
HD Graphics is pretty good for this power envelope. When you're talking about ~50% of the desktop chips' performance in 10 W, that's sure not bad. Nvidia's Tegra is still faster, but by not as much as you'd think.
"CPU base frequencies have not changed compared to respective Apollo Lake SKUs and burst frequencies only increased slightly (by 200 MHz, less than 10%). Therefore, all general-purpose performance benefits that Gemini Lake may have over their immediate predecessors are going to be originated from larger caches and whatever microarchitectural optimizations the new Goldmont Plus cores may have. Certainly, new instruction set extensions will bring their benefits, but only after software starts to use them."
Moving from 2 MB L3 to 4 MB L3 usually only yields a few % performance on average. The peak clock speed increase is 7.7%. To rough approximation both effects are comparable, depending on the details.
And the new instructions: are they new in Goldmont Plus compared to Goldmont?
Yes, there was a link detailing them. However, the only ones that should help performance are the SHA-NI, which accelerate SHA encryption/decryption.
The base clock didn't change because the process didn't change - Goldmont+ is still made on the original 14 nm node. Anyway, since Goldmont has only been out for about one year, this should be thought of as a refresh.
I am similar question but with Sandy Bridge, so newer then Core 2, and surprisingly it is very close. So my guess it is now equal or faster then Core 2.
Few things. 1. The one SB shown is a 35W part, and compared to a 6W part. 2. The SB has higher score, mostly due to faster memory and lower latency, my guess is that the N5000 was skimming down on cost and only uses a single 8GB DIMM, using dual channel should be even faster then SB. 3. If we use the Desktop 10W version with higher frequency and dual channel memory, it should be similar in Single Core Performance. 4. Much faster in Multi Core Performance.
Agreed on the single-channel speculation. The memory bandwidth is oddly low, given that the i3-2350M supported DDR3-1333 and the Pentium Silver N5000 supports DDR4-2400.
I realize Intel has said they are out of mobile, but it would be nice if they developed a follow on to the Cherry Trail processors. I had an Asus T100 that eventually died with Bay Trail in it and I have a T100ha which has Cherry Trail in it. A slick little, nice traveling 2-in-1. I don't need gobs of performance, but I sure don't mind improved performance. Microarchitecture in the ~10" tablet/2-in-1 range being basically static the last 2 years kind of stinks and having to have to have a tablet with active cooling, or a massive battery in it (relatively speaking) are kind of a killjoy.
Intel dropped out of the 2W-4W TDP segment when they killed the Cherry Trail successor. I also upgraded from a Bay Trail Windows tablet to a Cherry Trail device, it's great carrying a very light machine that can still run full Windows. I also do gaming with older Need for Speed games and old Bioware RPGs - on a tablet, no less. I doubt Intel want to get back into the segment, especially when Qualcomm look dominant in that space.
Probably. That said, my wife's 12 inch Acer 2-in-1 isn't too terrible. I usually end up using my T100ha as a laptop 95% of the time. So maybe next year I can find something in the 11-12" range, with Core M that is still reasonably small and light.
Heck, maybe AMD will come out with something in that low power space.
The Goldmont cores are very power-efficient. Whether they're awful depends on what you're trying to do with them.
Remember that the 6 W and 10 W TDP power figures include the envelope of a rather substantial GPU. If you're only doing something CPU-intensive, they'll probably use much less than that.
Can someone explain the rationale of Gemini Lake? Is there really a market between Core 15/5 watt products to slide in Gemini at 10/6.5 watt? I thought Atom SOC was supposed to go down to 2.5 watt so fight ARM's incursion.
Atoms used to go as low as 2W but Apollo Lake SoCs all were 6W and above. It's sad that Cherry Trail tablets are still being made because Intel has nothing else in that power segment. These are great chips for thin, cheap tablets and I have one as my main ultralight travel machine - 600g for a Windows tablet with 8 hour battery life is pretty good.
14nm vs 14nm, 4 cores vs 4 cores, 3 issue vs 3 issue, 1.5GHz vs 1.5GHz, 12/18 Gen 9 vs 12/18 Gen 9... So basically the same thing as Apollo Lake. Brand engineering.
The CPU has double the L2 cache and new instructions (including SHA acceleration). I had read that it was supposed to be 4-way issue, but I guess it's not.
The GPU has 10-bit support for HEVC & VP9 and native HDMI 2.0.
The SoC also has built-in AC wireless.
So, think of it as more of a mid-generation refresh, rather than either a tick or a tock. IMO, the clock speed & cache size improvements are worth something. The cache size will be especially helpful for implementations with single-channel memmory, which are all too common, sadly (recall that integrated graphics likes memory bandwidth).
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.
Facebook
Twitter
RSS
56 Comments
Back to Article
Old_Fogie_Late_Bloomer - Tuesday, December 12, 2017 - link
There's just something about the irregularities in the ball grid array in that first image that really tweaks my OCD.yeeeeman - Friday, January 12, 2018 - link
What irregularity? I see everything is symetric.Manch - Tuesday, December 12, 2017 - link
While I have no thoughts on the irregularities of the ball grid array, I am wondering what its performance compared to an 835 is.HStewart - Tuesday, December 12, 2017 - link
It would be hard to show actual performance difference - but since the 835 is emulation, I would expect that 835 will be significantly slower than these processors when run 835 application. It would be hard to say on .net application - that actually depends on Microsoft implementation of .net on ARM. In any case ,net applications on ARM should be significantly faster than Emulation.ddrіver - Tuesday, December 12, 2017 - link
Emulation or not it would still be interesting to see the current state since they'll probably go in similar devices.But if you ever think you're asking for too much for anything just remember that Intel is selling Atom chips for $160 a pop.
BedfordTim - Tuesday, December 12, 2017 - link
You can buy a motherboard with the Apollo Lake J4205 for under $100 despite that having a list price of $160, so I assume the prices are for show only.HStewart - Wednesday, December 13, 2017 - link
Yes a big difference with other systems ( possibly also Y series ) is that includes motherboard.Wilco1 - Tuesday, December 12, 2017 - link
Windows 10 and installed/store apps run natively on the 835, only x86 apps that aren't yet recompiled will require emulation. Even then you'll typically run native code since all the system libraries are native, not emulated.HStewart - Wednesday, December 13, 2017 - link
You are basically saying the same thing - stores apps are .net application - most of OS is written in .netWilco1 - Wednesday, December 13, 2017 - link
No, Windows is native C++. The key point is that even when running an emulated x86 application, you'll still run native code most of the time. See: https://www.extremetech.com/computing/249292-micro...HStewart - Wednesday, December 13, 2017 - link
First of the site - you provided does not have much creditability in my opinion, they are extremely Anti-BiasSecondary - I know there is some native C++ portions of the code - but a lot of windows code has been refactor with .net - this allows for integration into other platforms easier. But still there is a huge difference between native C++ code and Emulated code.
My guess performance wise the new Intel Atom here will be at same level or better than 835 - likely include in native C++ code and .net.
I heard that Microsoft is coming out with a Surface Phone, likely it will be 835 based.
Wilco1 - Wednesday, December 13, 2017 - link
Biased in what way? Plenty of sites have similar articles showing how Windows x86 emulation works. Are they all wrong/biased too? Is the video detailing how the x86 emulator works somehow faked?Alexvrb - Thursday, December 14, 2017 - link
Not all "apps" are native. Some are repackaged Win32 apps. Yes, the various MS libraries ARE native, which is awesome - but there's still a lot of code in an x86 program or "app" which is, well, x86. So the emulation cost is roughly an order of magnitude lower than it would be if everything was emulated, but it's still substantial and varies from program to program. Basically if you want to compare performance of the chips, you're better off testing native-compiled (mostly Store) software. If you're interested in seeing the performance impact of the partial emulation, test x86 programs. :DA5 - Tuesday, December 12, 2017 - link
Yeah it will be interesting to see (better) comparisons once Windows is on the 835/845.Wilco1 - Tuesday, December 12, 2017 - link
Geekbench 4.2 results for the N5000 vs 835: http://browser.geekbench.com/v4/cpu/compare/532602...So they are pretty close despite slightly lower frequency of the 835 (2.45GHz vs 2.7GHz max).
mode_13h - Saturday, December 16, 2017 - link
Looks suspiciously like the Apollo Lake SoC is using single-channel memory, while the Snapdragon 835 is using double (its memory bandwidth is more than double). With dual-channel, the N5000 should make a much better showing.MonkeyPaw - Tuesday, December 12, 2017 - link
Hard to make a direct comparison. The Intel chips have a higher power consumption. I believe the SD835 consumes up to maybe 4W, where all the Intel chips start beyond 4W up to 10W. So the Intel chips will likely perform better (or at least the CPU), but the SD835 will likely offer better battery life.Manch - Wednesday, December 13, 2017 - link
Theres a 6W variant of the chip that I figured would be put in netbooks 2n1's that would compete against things like the novago. x86 vs the 835 running Windows 10s.Emulation isnt an issue bc MS/Qualcom have said so, so lets run it like that.
I want to see battery life, typical app usage performance.
mode_13h - Friday, December 15, 2017 - link
Gemini Lake has a SDP of only 4.8 W. And I wonder if that's still with two memory channels.GhostOfAnand - Tuesday, December 12, 2017 - link
Where can I get me one of them "bigger accumulator"?solnyshok - Tuesday, December 12, 2017 - link
$160 for an Atom. Ouch. It might be cheaper to put Pentium G into mini-ITX board and limit frequencies to keep power consumption low.MrSpadge - Tuesday, December 12, 2017 - link
Yeah, the official prices have always been this mad. I don't think anyone is paying as much for them, though.FwFred - Tuesday, December 12, 2017 - link
Look at what full systems cost. That alone tells you OEMs aren't paying the list price.vovillia - Tuesday, December 19, 2017 - link
Agreed, as the 8350K is about the same price. Hopefully the listed price is just for show.mode_13h - Tuesday, December 19, 2017 - link
It's not comparable. Gemini Lake is a SoC, while the i3 still requires a South Bridge.Also, that i3 has a TDP of 91 W!
Jorgp2 - Tuesday, October 30, 2018 - link
That Pentium G would cost $160 by itself.For that price you can get this, a motherboard and memory.
HStewart - Tuesday, December 12, 2017 - link
A couple of things that come to mind1. I thought there were rumors that Intel was discontinuing the Atom series. It appears those rumors were wrong
2. This is first I heard of Gen10 Graphics core - and curious if more details - maybe I missed it somewhere else but latest I heard was Gen9.5
3. It good that they a speeding up these computers - they are significantly cheaper but performance is no where close to main line GPU including Y series which first came out Core M - which for most stuff is well accurate as long as you are not a serious gamer.
I hope this announcement means we here about next series Y processor with 10nm and possible even U series on 10nm. Intel has not updated Y series for 8th generation.
MrSpadge - Tuesday, December 12, 2017 - link
Re 1.: they're not branded as Atoms, that's what the "discontinuation" was about.Re 4.: those CPUs should appear sooner than later, but I don't think their schedule has anything to do with these "Atoms".
mode_13h - Friday, December 15, 2017 - link
What was discontinued were their SoC's aimed at cell phones and tablets. These are aimed at laptops and budget mini desktop systems.BTW, they still use Atom branding for embedded products, like the Atom x7-E3950:
https://ark.intel.com/products/96488/Intel-Atom-x7...
Dragonstongue - Tuesday, December 12, 2017 - link
graphics wise, the gen I could be wrong is for example DX9, DX10, DX11, DX12, so maybe when they are saying "gen 9 with some gen10) is saying DX9 for sure, but not all of DX10.Intel is not "known" for graphics leadership, as long as they can get an image on the screen, it is "good enough" for the heavy lifting, that is where graphics companies come into play.
One company cannot do everything perfectly, Intel might be massive, but, they are still but a small portion of "tech" in general.
extide - Tuesday, December 12, 2017 - link
Gen9/10 refers to the GPU arch, not anything to do with DirectX versions. I believe Gen 9 and 10 support DXFL 12_1mode_13h - Friday, December 15, 2017 - link
HD Graphics is pretty good for this power envelope. When you're talking about ~50% of the desktop chips' performance in 10 W, that's sure not bad. Nvidia's Tegra is still faster, but by not as much as you'd think.MrSpadge - Tuesday, December 12, 2017 - link
"CPU base frequencies have not changed compared to respective Apollo Lake SKUs and burst frequencies only increased slightly (by 200 MHz, less than 10%). Therefore, all general-purpose performance benefits that Gemini Lake may have over their immediate predecessors are going to be originated from larger caches and whatever microarchitectural optimizations the new Goldmont Plus cores may have. Certainly, new instruction set extensions will bring their benefits, but only after software starts to use them."Moving from 2 MB L3 to 4 MB L3 usually only yields a few % performance on average. The peak clock speed increase is 7.7%. To rough approximation both effects are comparable, depending on the details.
And the new instructions: are they new in Goldmont Plus compared to Goldmont?
mode_13h - Friday, December 15, 2017 - link
Yes, there was a link detailing them. However, the only ones that should help performance are the SHA-NI, which accelerate SHA encryption/decryption.The base clock didn't change because the process didn't change - Goldmont+ is still made on the original 14 nm node. Anyway, since Goldmont has only been out for about one year, this should be thought of as a refresh.
Ken_g6 - Tuesday, December 12, 2017 - link
So, is Atom Desktop faster than Core 2 Desktop yet? Wake me up when it is.iwod - Wednesday, December 13, 2017 - link
I am similar question but with Sandy Bridge, so newer then Core 2, and surprisingly it is very close. So my guess it is now equal or faster then Core 2.http://browser.geekbench.com/v4/cpu/compare/540979...
Few things.
1. The one SB shown is a 35W part, and compared to a 6W part.
2. The SB has higher score, mostly due to faster memory and lower latency, my guess is that the N5000 was skimming down on cost and only uses a single 8GB DIMM, using dual channel should be even faster then SB.
3. If we use the Desktop 10W version with higher frequency and dual channel memory, it should be similar in Single Core Performance.
4. Much faster in Multi Core Performance.
mode_13h - Saturday, December 16, 2017 - link
Agreed on the single-channel speculation. The memory bandwidth is oddly low, given that the i3-2350M supported DDR3-1333 and the Pentium Silver N5000 supports DDR4-2400.anon4832 - Wednesday, January 10, 2018 - link
Some Geekbench4 results have started to show up. Here's a nice one between GeminiLake and a Core2 Duohttps://browser.geekbench.com/v4/cpu/compare/27791...
Jorgp2 - Tuesday, October 30, 2018 - link
Current Atoms outperform a Q6600azazel1024 - Tuesday, December 12, 2017 - link
I realize Intel has said they are out of mobile, but it would be nice if they developed a follow on to the Cherry Trail processors. I had an Asus T100 that eventually died with Bay Trail in it and I have a T100ha which has Cherry Trail in it. A slick little, nice traveling 2-in-1. I don't need gobs of performance, but I sure don't mind improved performance. Microarchitecture in the ~10" tablet/2-in-1 range being basically static the last 2 years kind of stinks and having to have to have a tablet with active cooling, or a massive battery in it (relatively speaking) are kind of a killjoy.serendip - Wednesday, December 13, 2017 - link
Intel dropped out of the 2W-4W TDP segment when they killed the Cherry Trail successor. I also upgraded from a Bay Trail Windows tablet to a Cherry Trail device, it's great carrying a very light machine that can still run full Windows. I also do gaming with older Need for Speed games and old Bioware RPGs - on a tablet, no less. I doubt Intel want to get back into the segment, especially when Qualcomm look dominant in that space.azazel1024 - Wednesday, December 13, 2017 - link
Probably. That said, my wife's 12 inch Acer 2-in-1 isn't too terrible. I usually end up using my T100ha as a laptop 95% of the time. So maybe next year I can find something in the 11-12" range, with Core M that is still reasonably small and light.Heck, maybe AMD will come out with something in that low power space.
Pork@III - Tuesday, December 12, 2017 - link
How to quadruple price of mizerable ATOM CPU?Simply by adding the adjective "Silver" to the name.
Pork@III - Tuesday, December 12, 2017 - link
Pardon, it's not a processor, it's SoCextide - Tuesday, December 12, 2017 - link
? cost is same as old gens?MrSpadge - Tuesday, December 12, 2017 - link
Exactly. Although the meaning of these official prices is almost non-existent.Alistair - Tuesday, December 12, 2017 - link
Let's not beat around the bush. These are pretty awful. If you've ever built or been forced to use a machine based on these CPUs I feel sympathy.Just release a low power version of the G4560 and forget these exist.
mode_13h - Friday, December 15, 2017 - link
The Goldmont cores are very power-efficient. Whether they're awful depends on what you're trying to do with them.Remember that the 6 W and 10 W TDP power figures include the envelope of a rather substantial GPU. If you're only doing something CPU-intensive, they'll probably use much less than that.
wr3zzz - Wednesday, December 13, 2017 - link
Can someone explain the rationale of Gemini Lake? Is there really a market between Core 15/5 watt products to slide in Gemini at 10/6.5 watt? I thought Atom SOC was supposed to go down to 2.5 watt so fight ARM's incursion.serendip - Wednesday, December 13, 2017 - link
Atoms used to go as low as 2W but Apollo Lake SoCs all were 6W and above. It's sad that Cherry Trail tablets are still being made because Intel has nothing else in that power segment. These are great chips for thin, cheap tablets and I have one as my main ultralight travel machine - 600g for a Windows tablet with 8 hour battery life is pretty good.iwod - Wednesday, December 13, 2017 - link
I just looked it up and the 10W seems to have performance of a single core Sandy Bridge and much faster Multi Core in Geekbench.HEVC 10Bit Support.
Since Metal 2 is supported on Gen9 iGPU, i am guessing this is going to be the chip for next Mac Mini?
peevee - Wednesday, December 13, 2017 - link
14nm vs 14nm, 4 cores vs 4 cores, 3 issue vs 3 issue, 1.5GHz vs 1.5GHz, 12/18 Gen 9 vs 12/18 Gen 9... So basically the same thing as Apollo Lake. Brand engineering.mode_13h - Friday, December 15, 2017 - link
The CPU has double the L2 cache and new instructions (including SHA acceleration). I had read that it was supposed to be 4-way issue, but I guess it's not.The GPU has 10-bit support for HEVC & VP9 and native HDMI 2.0.
The SoC also has built-in AC wireless.
So, think of it as more of a mid-generation refresh, rather than either a tick or a tock. IMO, the clock speed & cache size improvements are worth something. The cache size will be especially helpful for implementations with single-channel memmory, which are all too common, sadly (recall that integrated graphics likes memory bandwidth).
raywzou8 - Wednesday, April 11, 2018 - link
Does anyone have first hand experience if 2x 8gb stick of ram will work on nuc7pjyh, or an asrock board with either j4005, j4105, or j5005Jorgp2 - Tuesday, October 30, 2018 - link
This is a really late reply.But i have been running a 4GB + 8G; for a while now, i don't see why 8+8 wouldnt work