Bay Trail-D Atoms (especially 2GHz+ quads) deliver very similar CPU performance, and reasonable IGP performance for the basic office use you'd expect from such chips. Kabini's better IGP comes at a cost: higher TDP (the 25W models are definitely not suitable for passive cooling). Also if you take a look at PCLab tests Kabini delivers <2x better gaming performance in some tests so while it is faster it's still way too slow for a lot of PC titles and Intel's could close this gap real fast with Braswell (but that's 2015 talk).
AtenRa
Lifer
- Feb 2, 2009
- 14,003
- 3,362
- 136
There are only one or two workloads/applications that Baytrail even comes close or is faster than Kabini.
You can play plenty of games at low resolutions (720p or lower) with kabini, but even at those low resolutions the same games are unplayable with Baytrail.
You know that Kabini replacement in Tablets and Mobile is coming close to 6 months before Braswell.
Any way, this is a Haswell topic, ATOM and Kabini has no businesses here (only for reference).
On the desktop? Sure. I think both Bay Trail AND Kabini are a joke for a desktop platform. You can get a Haswell based Pentium or Celeron which destroy both and costs a similar amount of money. Last I checked, this thread referred to *mind blown* HASWELL BASED PENTIUM & CELERON. And yeah, anyone buying either desktop BT or Kabini for desktop is a little special in the head.
Simply put, if you want a 80-100$ mobo + CPU you shouldn't be looking at BT or Kabini. The best choice is the Haswell celeron or pentium. Or you can get a 90$ Kabini platform which is inferior in every way. BT isnt' much better unless you get it in a tablet.
Now where Bay Trail excels is for superior performance per watt in tablets. And that's precisely why it has so many design wins, while Kabini has nearly none. AMD has their niche, but mobile products is just not an area they excel in. Their performance per watt for tablets and ultra portables is just not on intel's level. Which relegates AMD to the desktop for the most part.
Now AMD has some pretty good desktop products. Obviously they don't perform nearly as good as intel on the high end, but some of their APUs are good values for light / HTPC use. The only outlier I can think of here is the 7850k. At 190$ for the chip itself and 100$ for an overclocking motherboard, it is too expensive. Like said a Haswell 3220 + H81 mobo is right around 110$ right now with combo deals. The 7650 APU could be nice but I think it was delayed, can't find it anywhere at the moment.
Bay Trail was built from the ground up as a tablet product. And it excels there. Bay trail tablets are excellent with outstanding battery life. But if you want a desktop platform, you'd have to be nuts to get either BT or Kabini over a Haswell celeron or Pentium which costs the same and perform WAY WAY better. Period.
Simply put, if you want a 80-100$ mobo + CPU you shouldn't be looking at BT or Kabini. The best choice is the Haswell celeron or pentium. Or you can get a 90$ Kabini platform which is inferior in every way. BT isnt' much better unless you get it in a tablet.
Now where Bay Trail excels is for superior performance per watt in tablets. And that's precisely why it has so many design wins, while Kabini has nearly none. AMD has their niche, but mobile products is just not an area they excel in. Their performance per watt for tablets and ultra portables is just not on intel's level. Which relegates AMD to the desktop for the most part.
Now AMD has some pretty good desktop products. Obviously they don't perform nearly as good as intel on the high end, but some of their APUs are good values for light / HTPC use. The only outlier I can think of here is the 7850k. At 190$ for the chip itself and 100$ for an overclocking motherboard, it is too expensive. Like said a Haswell 3220 + H81 mobo is right around 110$ right now with combo deals. The 7650 APU could be nice but I think it was delayed, can't find it anywhere at the moment.
Bay Trail was built from the ground up as a tablet product. And it excels there. Bay trail tablets are excellent with outstanding battery life. But if you want a desktop platform, you'd have to be nuts to get either BT or Kabini over a Haswell celeron or Pentium which costs the same and perform WAY WAY better. Period.
Last edited:
Those are just a few games you can play with Bay Trail-T, which has a slower iGPU than Bay Trail-D
http://forums.anandtech.com/showpost.php?p=35912839&postcount=1936
Not an optimal experience, but huge improvement from the Clover Trail days and I bet Bay Trail-D's successor will also boost graphics performance considerably.
Only for reference then, I am not aware of any Mullins/Beema AM1 announcement, but 14nm Braswell is confirmed (this thread is about desktop products, we're not talking about tablets).
Last edited:
I just cannot think of any scenario where either of these chips (desktop bay trail or kabini) would be desirable for a desktop platform over the pentium or celeron. Sure, battery life and power use means a LOT for portables. But if your'e talking 25W on desktop vs 40W, who cares. I just can't see it. Don't understand it.
Now BT is good for mobile of course. As I mentioned power use matters for tablets, which is why BT was made primarily for portables. I'm trying to think of a compelling use scenario for either BT or Kabini on a desktop platform , and i'm just coming up with nothing.
No doubt, if you don't mind the extra platform power consumption (or active fan noise vs Bay Trail-D) they are all around superior desktop chips.
witeken
Diamond Member
- Dec 25, 2013
- 3,899
- 193
- 106
Maybe embedded bay trail could make sense in house devices. Smart TVs. Stuff like that? Although I haven't seen design wins, I could see why intel would design such an embedded chip for that purpose.
But yeah. For a desktop platform. I can't imagine who would buy this stuff. BT or Kabini, they both suck for desktop. Bt is primarily a tablet product, and it's a fantastic tablet chip. Unfortunately, a good tablet chip doesn't always turn into a great desktop chip.
But again, i'm thinking that intel had bigger goals in mind with their embedded bay trail chips. At least i'd hope so, but what do I know.
But yeah. For a desktop platform. I can't imagine who would buy this stuff. BT or Kabini, they both suck for desktop. Bt is primarily a tablet product, and it's a fantastic tablet chip. Unfortunately, a good tablet chip doesn't always turn into a great desktop chip.
But again, i'm thinking that intel had bigger goals in mind with their embedded bay trail chips. At least i'd hope so, but what do I know.
coercitiv
Diamond Member
- Jan 24, 2014
- 7,378
- 17,487
- 136
Both BT and Kabini are well suited for small, low power, silent HTPCs or home servers.
Even if prices don't show this yet, these platforms also have the potential to be very cheap, both in acquisition and usage costs, making them ideal in education, administration, services and other areas where that ST performance we love so much is not really a factor.
Make no mistake, a cheap and low power desktop platform will certainly find it's way on today's global market.
AtenRa
Lifer
- Feb 2, 2009
- 14,003
- 3,362
- 136
Well, the problem lies with you and others comparing Kabini to Haswell. They are not directly competing.
Secondly,
Cheapest Haswell Celeron G1820 is at $50 + cheapest Socket 1150 Mini-iTX motherboard is at $60.
Total of $110
Now, cheapest AM1 APU is Sempron 2650 at $41 + Cheapest Mini-iTX motherboard with USB3 and HDMI is at $35.
Total of $76
Haswell Celeron is almost 45% more expensive.
So again, Kabini is not directly competing with Haswell Celerons/Pentiums.
frozentundra123456
Lifer
- Aug 11, 2008
- 10,451
- 642
- 126
Since they are both in desktops, as well as BT, it is perfectly valid to compare them. I agree with Blackened that it makes little or no sense to put BT/Kabini in any sort of conventional desktop. Just too much performance sacrifice for that platform, even if you consider a small cost savings in absolute dollars, not percentage, since the absolute cost is pretty low to begin with.
I could see a place for them though in small devices like for mall kiosks, POS terminals, etc., but for conventional desktops, not really.
TreVader
Platinum Member
- Oct 28, 2013
- 2,057
- 2
- 0
And that's (haswell) for a POS H81 board. Put a decent Z87 board (because you're going to upgrade to a 4770K right, i mean that's the argument for 1150?) and even an 1820 is going to run you $250-350. That's more than Kabini will cost for a full system.
AtenRa
Lifer
- Feb 2, 2009
- 14,003
- 3,362
- 136
Well, Celeron G1820T and Pentium G3220T are also Desktop SKUs at 35W TDP, why dont you and the others compare them to Kabini ??
There are AIOs with those 35W Celerons and Pentiums and im more than 100% sure that Kabini Athlon 5350 will be faster, cheaper and consume less power than those Haswell SKUs.
The only thing you can with Z87 is features that you may not need for a low end platform. Suggesting that everyone gets Z87 with a Haswell Pentium is assinine. With Z87 you get more USB/SATA ports, , maybe a better audio chip, SLI/CF, more expansion , and overclocking.
These are things that I do not care about for my HTPC, and I certainly do not plan to upgrade it to a 4770k. Ever. I overclock my desktop but I will not be overclocking my HTPC anytime soon. H81 as a platform also does not perform differently than any other chipset. Z87 gives you more features and overclocking, which is desirable for overclocking SKUs such as the 4770k. These features aren't needed by buyers of light use daily driving systems.
Total cost? 100$. Honestly you'd have to be stupid to get Z87 with a Haswell Celeron. It makes sense with a 4770k. It doesn't make sense with a Pentium or Celeron. But. By all means. If someone wants to buy an overclocking Z87 motherboard for a chip that doesn't overclock? Sure they can do it. Further, If someone wants a low performance chip such as the Kabini, that's their prerogative. I do believe most would opt for the much higher performing and similarly costing platform with the celeron or pentium.
I also do not see desktop SKU power draw being a valid argument in this case; all of these SKUs are very close to each other. You can run any and all of these chips with a 300W PSU. Therefore, even if the Pentium G3220 uses perhaps 15-20 more watts (not sure, just an example) it doesn't matter. All of these run on a 300W PSU. This is not a situation of something like going from a iGPU to a R9-290X which adds 300W and requires a bigger power supply - power consumption can be a valid argument in that context because the differences in graphics power use are vast. But comparing kabini, BT, haswell, celeron on the desktop: These are all usable CPUs on a 300W PSU therefore it does. not. matter. These are all low power consumption chips, and are not a mobile platform, therefore the power differences between kabini and celeron/pentium means little if anything in a desktop context. In a mobile context it might matter. Not here though.
These are things that I do not care about for my HTPC, and I certainly do not plan to upgrade it to a 4770k. Ever. I overclock my desktop but I will not be overclocking my HTPC anytime soon. H81 as a platform also does not perform differently than any other chipset. Z87 gives you more features and overclocking, which is desirable for overclocking SKUs such as the 4770k. These features aren't needed by buyers of light use daily driving systems.
Total cost? 100$. Honestly you'd have to be stupid to get Z87 with a Haswell Celeron. It makes sense with a 4770k. It doesn't make sense with a Pentium or Celeron. But. By all means. If someone wants to buy an overclocking Z87 motherboard for a chip that doesn't overclock? Sure they can do it. Further, If someone wants a low performance chip such as the Kabini, that's their prerogative. I do believe most would opt for the much higher performing and similarly costing platform with the celeron or pentium.
I also do not see desktop SKU power draw being a valid argument in this case; all of these SKUs are very close to each other. You can run any and all of these chips with a 300W PSU. Therefore, even if the Pentium G3220 uses perhaps 15-20 more watts (not sure, just an example) it doesn't matter. All of these run on a 300W PSU. This is not a situation of something like going from a iGPU to a R9-290X which adds 300W and requires a bigger power supply - power consumption can be a valid argument in that context because the differences in graphics power use are vast. But comparing kabini, BT, haswell, celeron on the desktop: These are all usable CPUs on a 300W PSU therefore it does. not. matter. These are all low power consumption chips, and are not a mobile platform, therefore the power differences between kabini and celeron/pentium means little if anything in a desktop context. In a mobile context it might matter. Not here though.
Last edited:
Wrong.
Intel Haswell Pentium G3220 + H81 mobo:
http://www.newegg.com/Product/ComboD...=Combo.1596228
112$
I bought it even cheaper than that with a promo code. You have to ask yourself if saving 25-35$ is worth getting a kabini chip. Or an embedded BT for that matter. Both of these options suck for desktop. I don't buy this power argument either. This does not matter , they're desktop SKUs, and we're talking 15-30W. I can see the power argument coming into play when the differences are vast, such as with some GPUs. GPUs range from 60W to 500W in TDP. Certainly power use can be an argument to made for GPUs. But when the differences are this small, on a desktop, and for low end desktop SKUs? Come on. Who the heck cares. All 4 of these chips (Haswell celeron, pentium, BT and Kabini) these are ALL efficient chips. The difference being that the celeron and pentium perform significantly better without a huge jump in power use. Certainly not enough of a jump to matter. The added performance warrants it.
This matters for mobile. This does not matter for a desktop. Now, bay trail has its uses; it's a great chip for tablets with long battery life. But as a desktop platform it is not desirable. Kabini, on top of being terrible for a desktop chip, isn't really even desirable on a mobile platform.
Last edited:
AtenRa
Lifer
- Feb 2, 2009
- 14,003
- 3,362
- 136
That is not a mini-itx board, not to mention that combo it is not even available to purchase.
Again, why dont you compare the 35W TDP Haswell SKUs against 25W TDP Kabini.
Even the slowest 35W ''T'' model (2.4GHz Celeron G1820T) would be faster than Kabini most of the time CPU-wise (MT-stuff included), just take a look at 2.7GHz Celeron G1820's performance. Lets not forget the 2x better per core performance, which does make a difference. Considering that regular Celeron/Pentium ''Haswell GT1'' put Kabini to rest in PCLab's iGPU gaming tests (not even Mantle saved the day) it wouldnt surprise me if they still beat the AM1 Athlon when it comes to actual gaming performance (unlike what synthetic GPU-only benchmarks could sugest). The Celeron would also be a much better pick to pair with discrete graphics, and allows you to upgrade to Core i3/i5/is later while AM1 Kabini will allow you to upgrade to... another AM1 Kabini (or Beema/Mullins if it ever comes to this platform).
Last edited:
The Celeron would also be a much better pick to pair with discrete graphics, and allows you to upgrade to Core i3/i5/is later while AM1 Kabini will allow you to upgrade to... another AM1 Kabini (or Beema/Mullins if it ever comes to this platform).
Excuse me, but exactly who in the target market for Kabini is going to pair a discrete card with it? Kabini doesn't even have a full-size PCIe x16 slot, its only a x4 link...
Kabini with 260(x)/750ti... riiiight...
(It'd actually be worth doing it to see how much performance you can squeeze out of the Athlon 5350)For discrete cards, AMD has the superb Athlon 750/760K.
Because no one tested them? also, if you gona compare the 25W Kabini vs the 10W BT, its fair to also compare it to the 54W Haswell. You wanna check the %?
But the 25W kabini use about the same amount of power as BT unless you use as bench PClab who use a 550W PSU for Kabini and a 90W one for BT in the same review while the Haswell based celeron plateforms are clearly in another TDP range.
I think you entered the wrong topic, your comment makes no sense at all, it's quite bizarre, $350 for Celeron + MB!?
Z87 for Pentium/Celeron is wasteful, and a 4770K works 100% fine on H81, performance is the same if you don't OC.
if you think H81 is bad look at the cheaper AM1 boards, no PCIE 16x, no 4 sata ports and so on...
Kabini full system and Haswell full system can cost the same,
if you have a regular case the G1820 Haswell + h81 board you cost you around $90 which is the same or lower than the 5350 Kabini + AM1 board, and the H81 board offers more, the CPU is faster.
they've used the same PSU for Haswell and Kabini, the power usage range is close enough, it's a pretty valid comparison.
A 550W PSU, there will not be a single configuration in the world,even home made, that will use such an arrangement with a Kabini plateform, actualy its only relevance is to provide truncated power numbers that help promote plateforms being more power hungry and much less power efficient.
both are using the same PSU, both are under 50w most of the time, simple.
use numbers from the same source.
Haswell can also go lower with more efficient PSU and MB than they used, you seem to ignore this only for Haswell, not for Kabini, why?
I dont ignore this and indeed that is the point that interest us and you are right but it wont go as low as a Kabini while it will need much more when loaded given that it has inherently lower perf/watt than Kabini by a sizeable margin even with an efficient PSU for HW.
Now using your PClab numbers we can see that whatever the supply losses the G3220 plateform consume 8W more at iddle than Kabini s.
http://pclab.pl/zdjecia/artykuly/radek/2014/am1/wykresy/energia_spoczynek.png
From Golem.de review we know that a Be Quiet Straight Power E9 400 with high efficency at low loads has obviously 6W less losses at thoses levels than PClab s FSP Raider 550 so this would bring down the G3220 number to 22W, a nice number for such a plateform but still on the order of Kabini/BT full load power, besides it would require a 100-120W PSU while the low power solutions are content with 40-50W ones that will help cut the costs corners even more.
http://www.golem.de/news/kabini-fue...lte-jaguare-fuer-jedermann-1404-105670-2.html
Well Kabini is more of a BayTrail competitor rather than Celeron / Pentium one.
Cheap 'big core' A4/A6 Richland&upcoming Steamroller APUs & Athlons are Celeron / Pentium competirors.
Just saying. It is true that Intel is doing good job and hopefully Intel is planning to continue to improve cheap Celeron / Pentium Core processors in future as well with strong pace.
-----------
PS. If you're willing to read tests through translators then you can take a look at these as well:
G1820
http://www.purepc.pl/procesory/test_intel_celeron_g1820_haswell_najtanszy_procesor_lga_1150
G3220
http://www.purepc.pl/procesory/test_intel_pentium_g3220_jaki_procesor_za_200_zlotych
Cheap 'big core' A4/A6 Richland&upcoming Steamroller APUs & Athlons are Celeron / Pentium competirors.
Just saying. It is true that Intel is doing good job and hopefully Intel is planning to continue to improve cheap Celeron / Pentium Core processors in future as well with strong pace.
-----------
PS. If you're willing to read tests through translators then you can take a look at these as well:
G1820
http://www.purepc.pl/procesory/test_intel_celeron_g1820_haswell_najtanszy_procesor_lga_1150
G3220
http://www.purepc.pl/procesory/test_intel_pentium_g3220_jaki_procesor_za_200_zlotych
TRENDING THREADS
-
Discussion Zen 5 Speculation (EPYC Turin and Strix Point/Granite Ridge - Ryzen 9000)- Started by DisEnchantment
- Replies: 25K
-
Discussion Intel Meteor, Arrow, Lunar & Panther Lakes + WCL Discussion Threads
- Started by Tigerick
- Replies: 22K
-
Discussion Intel current and future Lakes & Rapids thread- Started by TheF34RChannel
- Replies: 23K
-
-
News NVIDIA and Intel to Develop AI Infrastructure and Personal Computing Products- Started by poke01
- Replies: 411
Facebook
Twitter