Kabini IPC, what generation are we talking about?

[NTMBK]

You can also replace the 53W with a 35W T model. not to mention again, the Celeron idles at lower power.

And how much of a premium do you pay for those T models? The only place I can find selling the G1820T in the UK, Amazon, are charging £46.62 for it without a heatsink. The cheapest widely available, non-OEM T-series part I can find is an £86 Core i3!

How much performance does the T series Celeron sacrifice? CPU clock goes 2.7GHz->2.4GHz, but the GPU clock goes from 350MHz down to 200MHz. That's almost half its performance gone!

And yet again, you're forgetting the extra 4W for the Intel chipset. It's 25W vs 39W.

 

[ShintaiDK]

And how much of a premium do you pay for those T models? The only place I can find selling the G1820T in the UK, Amazon, are charging £46.62 for it without a heatsink. The cheapest widely available, non-OEM T-series part I can find is an £86 Core i3!

How much performance does the T series Celeron sacrifice? CPU clock goes 2.7GHz->2.4GHz, but the GPU clock goes from 350MHz down to 200MHz. That's almost half its performance gone!

And yet again, you're forgetting the extra 4W for the Intel chipset. It's 25W vs 39W.

The MSRP for the T model is the same for the non T model.

The G1840T is 2.5Ghz 200Mhz/1.05Ghz GPU. The G1850 is 2.9Ghz and 350Mhz/1.05Ghz GPU.

When talking about performance, you also have to remember that these Celerons offers twice the performance in 1-2 threads compared to Kabini. Its really really hard to justify Kabini unless you shoehorn it into some ridiculous scenario. Remember something like the J1900 with board also sits below in TDP at 70$ for CPU+board. Almost the same price as the 5350 CPU alone.

 

[monstercameron]

The MSRP for the T model is the same for the non T model.

The G1840T is 2.5Ghz 200Mhz/1.05Ghz GPU. The G1850 is 2.9Ghz and 350Mhz/1.05Ghz GPU.

When talking about performance, you also have to remember that these Celerons offers twice the performance in 1-2 threads compared to Kabini. Its really really hard to justify Kabini unless you shoehorn it into some ridiculous scenario.

it isn't hard at all, you are making it hard.

 

[ShintaiDK]

it isn't hard at all, you are making it hard.

Please elaborate.

I assume its something along the line of needing more GPU than a J1900, willing to pay more, and then anything over 25W TDP is still too much? While 10W TDP is unattractive? Oh, and you run 4 threaded applications only as well while performance/$ doesnt matter much?

 

[monstercameron]

The kabini system takes it for me because of the added gpu perf.

 

[NTMBK]

The MSRP for the T model is the same for the non T model.

What does MSRP matter if you can't actually buy it at that price?

 

[NTMBK]

The J1900 certainly makes sense, by the way, and is the sensible part to compare Kabini to. Haswell? No. Really no.

 

[ShintaiDK]

What does MSRP matter if you can't actually buy it at that price?

The G1840T on geishalz is 10euro cheaper than the 5350. better? I am sure you can find a 35W cooler for 10euro.

 

[ShintaiDK]

The J1900 certainly makes sense, by the way, and is the sensible part to compare Kabini to. Haswell? No. Really no.

So you want to compare 10W with 25W. But 25W with 35W or 53W is completely out of the question. You dont see the issue yourself?

 

[ShintaiDK]

The kabini system takes it for me because of the added gpu perf.

What added GPU perf? Over Haswell?

 

[Enigmoid]

The kabini system takes it for me because of the added gpu perf.

128 gcn cores cannot catch up to intel.
The 10 EU in HW pentiums is a substantial increase over IVB pentiums.

 

[SPBHM]

128 gcn cores cannot catch up to intel.
The 10 EU in HW pentiums is a substantial increase over IVB pentiums.

more like single channel ram, low clock and slow CPU cores...

128 gcn cores at the same clock as the 10 EUs and with the same memory bandwidth, and a faster CPU could be a good fight!?

but yes, Kabini IGP looks good compared to Bay Trail HD Graphics, not Haswell HD Graphics

 

[AtenRa]

The Celeron will complete the task much much faster and return to idle.

So now that the Power Consumption graphs are not in your favor, you are talking about Energy consumption.
Funny that you dont consinder Energy consumption when AMD products are faster than Intel.

 

[AtenRa]

What added GPU perf? Over Haswell?

A single game?? really??

 

[SPBHM]

A single game?? really??

http://www.techspot.com/articles-info/806/bench/Gaming_06.png

5+ games
http://pclab.pl/art57195-7.html

G1820 is normally a little faster, but to be honest both Kabini and Haswell GT1 are really bad for gaming anyway.

 

[AtenRa]

5+ games
http://pclab.pl/art57195-7.html

G1820 is normally a little faster, but to be honest both Kabini and Haswell GT1 are really bad for gaming anyway.

There are 6 games in the techspot review but he only chose the one that shows the Athlon in the worst possition.

Yes i will agree those games are not playable with either CPU.

 

[Enigmoid]

more like single channel ram, low clock and slow CPU cores...

128 gcn cores at the same clock as the 10 EUs and with the same memory bandwidth, and a faster CPU could be a good fight!?

but yes, Kabini IGP looks good compared to Bay Trail HD Graphics, not Haswell HD Graphics

Intel's architecture is made to run the EUs at 1+ ghz. AMD's does not reach clockspeeds that high. Clock for clock is largely worthless.

A single game?? really??

Because somehow the fact that HD 4600 performs worse than HD graphics (HW) doesn't send up a red flag?

 

[Roland00Address]

45% more power usage in CPU + GPU load for the Celeron G1820 than TOP-end Athlon 5350

That is when you load the cpu at complete load. Since the celeron G1820 is often much faster than the 5350 it does not need to be at 100% load to do the same amount of work.

For example if the G1820 is twice as fast as the 5350 at task X and task X for the 5350 uses a 100% of the cpu, then the G1820 is using only 50% of the cpu potential at task X.

------

No one is doubting that the Haswell Desktop Pentiums and Celerons are capable of using more power at 100% load, but that does not mean they will use more power than the athlon at the same load.

What you need to do is do a test where you have a fixed workload, such as game with a frame rate cap and see if the Haswell uses more power or less power during that work load.

So now that the Power Consumption graphs are not in your favor, you are talking about Energy consumption.
Funny that you dont consinder Energy consumption when AMD products are faster than Intel.

You think he is being a fan boy, but there are very little scenarios where the cpu can have power over 10 watts, but it must be under 50 watts.

Something like a router, a firewall, a nas would need a very low power consumption cpu so they need a sub 10 watt cpu.

Any itx form factor case can easily run a 50 watt haswell.

You have to be doing something silly to need a 25 watt cpu, something like an itx form factor case and you want to run the cpu completely passive and you do not want to have a big passive heatsink and the heat sink has to use cheap materials.

You can run even a 53 watt pentium completely passive, but it would require a bigger heatsink than a 25 watt cpu.

------

Lets be honest, the 5350 is a laptop chip that is socketed and can be put into a desktop motherboard. It compares very badly to desktop chips. There are some circumstances where it makes sense, but those circumstances are extremely limited.

 

[AtenRa]

That is when you load the cpu at complete load. Since the celeron G1820 is often much faster than the 5350 it does not need to be at 100% load to do the same amount of work.

For example if the G1820 is twice as fast as the 5350 at task X and task X for the 5350 uses a 100% of the cpu, then the G1820 is using only 50% of the cpu potential at task X.

G1820 is not that much faster than Athlon 5350 in every day loads. I can tell you that zooming in and out in Google Maps using Chrome(im just testing it now) the G1820 has up to 96% of CPU Utilization when Athlon 5150 (1.6GHz Quad Core) only has up to 76%.

No one is doubting that the Haswell Desktop Pentiums and Celerons are capable of using more power at 100% load, but that does not mean they will use more power than the athlon at the same load.

Nobody test at the same load, every review test at 100% load for each CPU.

What you need to do is do a test where you have a fixed workload, such as game with a frame rate cap and see if the Haswell uses more power or less power during that work load.

That measures Power Consumption at the same Performance. It is not the maximum power consumption of the CPU.

Lets be honest, the 5350 is a laptop chip that is socketed and can be put into a desktop motherboard. It compares very badly to desktop chips. There are some circumstances where it makes sense, but those circumstances are extremely limited.

At its TDP Its the best APU so far. It is not made to compete against the Haswell Celerons and Pentiums. Only people in Forums compares them, AMD only compare them against Intel ATOM based Celeron/Pentiums.

 

[ShintaiDK]

At its TDP Its the best APU so far. It is not made to compete against the Haswell Celerons and Pentiums. Only people in Forums compares them, AMD only compare them against Intel ATOM based Celeron/Pentiums.

Why do you think AMD only compares them to CPUs with 40% the TDP of the Kabini chips?

Nomatter if you like it or not, they also have to compete against Haswell based Celerons. The buyers dont care what AMD wish to position them against to show the most favourable benchmarks.

 

[AtenRa]

Why do you think AMD only compares them to CPUs with 40% the TDP of the Kabini chips?

Nomatter if you like it or not, they also have to compete against Haswell based Celerons. The buyers dont care what AMD wish to position them against to show the most favourable benchmarks.

You know very well that AMD has 45/65/95 and 100W TDP SKUs to compete against those Haswell Celerons, Pentiums, Core i3 and i5s.

 

[Sweepr]

Why do you think AMD only compares them to CPUs with 40% the TDP of the Kabini chips?

Nomatter if you like it or not, they also have to compete against Haswell based Celerons. The buyers dont care what AMD wish to position them against to show the most favourable benchmarks.

Exactly. At similar price and slightly higher power consumption Haswell Pentiums/Celerons are all around superior, and Bay Trail-D gives you pretty much the same CPU performance as Kabini at 10W TDP and easily allows passive cooling. If you want Bay Trail-D CPU performance + better graphics (but still far too slow for many games BT-D also cant handle) at 25W TDP, can't live with 10-25W extra watts for much better performance on the Haswell side and dont plan too add a reasonably powerful discrete gfx card at any time, sure, go Kabini.

 

[SPBHM]

Intel's architecture is made to run the EUs at 1+ ghz. AMD's does not reach clockspeeds that high. Clock for clock is largely worthless.

you said 128 GCN is no match for 10 Haswell EUs, without other informations this might not be true,
in this case it's 600 vs 1050MHz, GCN can normally work at higher clocks, maybe not 1GHz with the GloFo 28nm process, but still should be able to go higher with a higher TDP, and as I said 64bit vs 128bit memory is affecting the comparison,

You know very well that AMD has 45/65/95 and 100W TDP SKUs to compete against those Haswell Celerons, Pentiums, Core i3 and i5s.

single module FM2 CPUs are no match for the dual core Haswell (higher power draw, lower performance, no cost advantage), and the 2 module CPUs are more expensive.

 

[Roland00Address]

G1820 is not that much faster than Athlon 5350 in every day loads. I can tell you that zooming in and out in Google Maps using Chrome(im just testing it now) the G1820 has up to 96% of CPU Utilization when Athlon 5150 (1.6GHz Quad Core) only has up to 76%.

Could that be the software you are now testing only stresses 3 threads and thus you are only using 3/4s of your quad core. The other core is just sitting there being useless? Your google maps software wants even more power than the haswell celeron or the 5350 can provide, but since it can't get that power it will just run the cpu at near maximum load.

Nobody test at the same load, every review test at 100% load for each CPU.

You did not get my point. You are correct most reviews do not test a cpu at the same load. It is quite easy for them to test a cpu at the exact same load, but most reviews choose not to.

But you did not get my point in the real world some tests do not use a cpu to the maximum potential, and the 5350 and the haswell pentium or celeron will perform equally for they are not using the full potential of the cpu. In such a situation you would not get the haswell pentium using 53 watts of power but instead a lesser amount. They may even use roughly the same amount of power.

But when you need as much power as possible, you need as much speed, the haswell pentium can be 50 to 300% faster for those extra 25 watts of power.

You do not get that dynamic range is a good thing, and just because a cpu has high dynamic range does not mean it can't be more efficient than a cpu with low dynamic range. If the cpu is not running at 100% the cpu that actually uses the least amount of power is not the lower tdp chip or the higher tdp chip but the chip that is the most efficient.

That measures Power Consumption at the same Performance. It is not the maximum power consumption of the CPU.

That is my entire point. Power Consumption at the same performance. The haswell system does need a very very slightly beefier cooling for it can go roughly 25 watts higher in tdp. But it rare if ever produces 53 watts of heat. If you throw the same type of tasks at the haswell system the haswell system will do those tasks faster and at a lower power usage when it performs those tasks.

If you have something such as a dynamic fan with different fan speeds, or you are measuring temperature at the heatsink the fan will need to be used less with the haswell and thus can be used at lower rpms, or if you are passively cooled the temperature of the cooler will be lower on the haswell.

At its TDP Its the best APU so far. It is not made to compete against the Haswell Celerons and Pentiums. Only people in Forums compares them, AMD only compare them against Intel ATOM based Celeron/Pentiums.

When people compare computers, they compare them via how people use computers, they do not compare the engineering merits of a cpu.

Put another way, the 5350 is effectively an atom 10 watt cpu with a 15 watt video card. Well what is the point of a 15 watt video card when it can't play any games at any playable settings.

Well the 5350 has a lower possible power consumption than the haswell pentiums and celerons. Well when you have a desktop that can easily cool said cpu, and it cost the same as the 5350 system, why would you go for the slower desktop cpu?

The 5350 has a purpose, in laptops and all in ones, that want better graphics than the intel offerings, and want to keep the price low. The 5350 should not be in any desktop, for other cpus do that task so much better. The 5350 is a laptop chip it shines as a laptop chip, it really has no place in a socketed desktop world, besides AMD made it and is willing to sell it to OEM for peanuts due to volume purchasing discounts but AMD is not willing to pass those same prices to the DIY.

 

[monstercameron]

Put another way, the 5350 is effectively an atom 10 watt cpu with a 15 watt video card. Well what is the point of a 15 watt video card when it can't play any games at any playable settings.

Well the 5350 has a lower possible power consumption than the haswell pentiums and celerons. Well when you have a desktop that can easily cool said cpu, and it cost the same as the 5350 system, why would you go for the slower desktop cpu?

The 5350 has a purpose, in laptops and all in ones, that want better graphics than the intel offerings, and want to keep the price low. The 5350 should not be in any desktop, for other cpus do that task so much better. The 5350 is a laptop chip it shines as a laptop chip, it really has no place in a socketed desktop world, besides AMD made it and is willing to sell it to OEM for peanuts due to volume purchasing discounts but AMD is not willing to pass those same prices to the DIY.

beg to differ, the Athlon 5350 is quite capable.