25w tdp versus 10w tdp.
The form factor support will be totally different.
Since Toms HW measure 10w J1750 Baytrail Celeron at 20w, then Pentium will be same or higher with Kabini and new Beema Jaguar/Puma APUs.
Can you link? I've never seen Intel go over TDP. What does an equivalent Kabini measurement look like using the same methodology?
It is completely valid to prefer 25W over 10W CPUs in desktops, but I don't believe for a second these two CPUs are anywhere near each other in power consumption.
avtek: the text clealy says its whole system power. You understand the difference between system power measured at the wall and TDP?
The Kabini SOC alone is 25W TDP.
I also suspect the 15W TDP is perhaps a bit conservative, total platform power consumption with all CPU cores firing never exceeded 12W (meaning SoC power consumption is far lower, likely sub-10W).
From Toms, AMD Kabini and Pentium and i3 (17w), which one stays closest near TDP spec?
Sweclockers reporting new AMD Athlon Kabinis with four Jaguar cores and GCN 128 SPs will be selling first qtr next year for $35 to $54! Benchmarks and features in several charts.
Its another misfit product on AMDs road to being the new VIA.
Apparently these cpus will be socketable in a new socket (fs1b), so between AM3, FM2(+), and this socket, AMD will have 3 sockets targeting entry-level/mainstream desktops....
From Toms, AMD Kabini and Pentium and i3 (17w), which one stays closest near TDP spec?
Not really. Because Atom does very well for its intended(and rapidly growing) market, and Kabini really only fits the value Notebook, something that's increasingly being taken up by Tablets of all sorts, Chromebooks, Androidbooks.Why ATOM road is not a way for Intel to become a new VIA but Kabini is for AMD ?? Double standards ??
You can't compare the two together because the tests are not the same.
Also, its very likely the best Bay Trails are in the Tablet parts, and Celeron and Pentium parts are similar to Desktop Atoms that they are not optimized so much for efficiency and size, but for cost.
Not really. Because Atom does very well for its intended(and rapidly growing) market, and Kabini really only fits the value Notebook, something that's increasingly being taken up by Tablets of all sorts, Chromebooks, Androidbooks.
TDP is not power consumption. Anandtech clearly showed that A4-5000 which is 15W TDP rated never reached 15W for the entire system. A 25W TDP kabini will have lower than 20-15W power usage.
If i remember correctly, Desktop ATOMs are different dies at different manufacturing process than Tablet ATOMs.
The article did show that but it only measured power for CPU-only tests. Unlike a lot of Intel CPUs that have a wide turbo swing where either CPU or GPU can take it A4-5000 has no CPU turbo swing whatsoever and statically allocates a big chunk of its TDP towards GPU. Notebookcheck for instance showed that under GPU heavy tests the power consumption went way up.
That doesn't mean your statement is wrong if we look at GPU-light scenarios only, it just means that AMD hasn't exaggerated TDP as vastly as this alone suggests.
That was the case last generation but is probably not the case now. All of the parts called BayTrail are probably the same die.
Notebookcheck measures the entire Laptop power consumption including the Monitor. Do you have any links to check out ??
No PCI-e on the Desktop ATOMs ???
http://www.notebookcheck.net/Short-Review-AMD-A4-5000-APU-Kabini.93173.0.html
The point here isn't absolute power but the jump you get going from Cinebench to FurMark, then Prime95 + FurMark - it's that last one that gives you the best picture of the difference between a heavy CPU-only load and what would push the limits of TDP.
Or I/O fused off on different packages, I don't really know, could be either I guess.. we'd need to see die shots..
Is FS1B the first user-replaceable SoC socket? It'd be interesting to see the longevity of the motherboards considering there's no separate chipset that needs to be upgraded. If the socket supports multi-channel RAM from the start a single motherboard could potentially last 2-3 generations.
Yes but now give me a link to Pentium 2020m for the same workloads to check the power consumption difference between the two.
I will try to find out later on but I believe those two are different dies.