AMD Ontario performance numbers leaked

[Skurge]

http://www.bit-tech.net/news/hardware/2010/07/30/amd-ontario-performance-numbers-leaked/1

Looking good, disappointing thats it a BGA chip though.

 

[veri745]

...snip... disappointing thats it a BGA chip though.

Why. BGA means smaller form factors. Isn't that exactly what Ontario is targeting?

 

[maniac5999]

It doesn't look too bad. Assuming they were running at the same speed, it's 85% as powerful as a L3-less K10.5, the article doesn't seem to specify the speed of the Bobcat, so it could even be running at 1.4ghz, which woud indicate the same IPC as K10.5.

I have a U230 netbook (1.6ghz K8 x2 and a 3200HD) and today this seems to be the sweet spot, or slightly below the sweet spot for an ultraportable performance-wise. I can play WoW on medium and SC2 on low, and can play most 1080p videos on CPU power alone. (I have a very demanding rip of Slumdog Millionaire that stutters badly tho) If not for the bad battery life (4 hours internet/idle, 2 hours gaming with a 6 cell) it's be the perfect laptop. If ontario can increase CPU/GPU power by 15% while increasing battery life 50% and still keeping the price under $500 I think that they'll have a major winner on their hands.

 

[busydude]

SA has got hold of the numbers for both Llano and Ontario.

Link to the article.

Link to the Llano's BIONIC benchmark using a custom Linux kernel.

 

[Martimus]

This looks like a nice Netbook processor. It will probably take at least 2 more iterations (I would guess 3) to get it down to a smaller form factor processor though (like a smart phone).

I really wonder when it will be released, and what types of devices will run it. Based on what I see here, it looks like a netbook processor that isn't dog slow.

 

[cbn]

Why. BGA means smaller form factors. Isn't that exactly what Ontario is targeting?

How does BGA mean smaller form factors?

P.S. I noticed the word "lidless" mentioned during my internet search of BGA. Is this done to reduce the size/weight of the hsf needed? Do all BGA socketed processors come lidless?

 

[Martimus]

Why does BGA mean smaller form factors?

P.S. I noticed the word "lidless" mentioned during my internet search of BGA. Is this done to reduce the size/weight of the hsf needed? Do all BGA socketed processors come lidless?

It fits smaller form factors, because there is no need for a socket to fit the pins into. The chip can be soldered directly to the PCB.

Although this chip is a little too power hungry (at 18w if the OP is correct) to really be used in most of the really small form factors.

EDIT: Wikipedia has a good overview of Ball Grid Array (BGA): http://en.wikipedia.org/wiki/Ball_grid_array

 

[khon]

Would be interesting to see how its performance compares to i5 processors with the same 18W TDP.

As it is 18W seems a bit high for netbooks, and the performance is probably on the low end for laptops, so it might be stuck somewhere in the middle.

 

[cbn]

EDIT: Wikipedia has a good overview of Ball Grid Array (BGA): http://en.wikipedia.org/wiki/Ball_grid_array

Thanks. The Wikipedia link does a good job at explaining how BGA helps transfer heat into the PCB reducing or eliminating the need for a HSF.

 

[cbn]

I did some more searching on the Ball Grid array thermals and it does appear some BGA designs like H-PBGA and HL-PBGA have something described which sounds like a "lid".

http://www.intel.com/Assets/PDF/pkginfo/ch_14.pdf

14.10.1 Thermal Performance
In general, three factors affect the thermal performance of the BGA: package and board materials,
package geometry, and use environment. Obviously, the more thermally conductive the materials,
the better the package dissipates heat. On a PBGA package, the molding compound that surrounds
the chip, which provides mechanical protection as well as a surface for marking, is typically 45mils
in thickness and is the main barrier to the elimination of the heat from the BGA package.
In general design-related factors have greater thermal effect on the PBGA than material-related variables.
A large die spreads heat easier, as does a larger package size and thus a higher ball count. Substrate
design features have tremendous effect on the package’s ability to dissipate heat. Vertical vias
running through the substrate help to transfer heat from the die to the solder balls. Four-layer designs
often incorporate conductive 2-ounce copper ground planes, which have a significant, positive effect
on thermal performance. The Enhanced PBGA has thermal balls under the die while H-PBGA and
the HL-PBGA utilize a heat spreader or slug across the top of the package to dissipate heat even
more efficiently. PBGAs with center thermal balls dissipate considerable heat into the board. A considerable
increase in thermal effectiveness of a BGA package can be obtained by using boards that
are thermally efficient, increasing the airflow, or providing thermal paths from the board. Remember,
with PBGAs, the board is your primary heatsink.
Environmental conditions play a critical role in the thermal performance of PBGAs. Ambient conditions,
junction and case temperatures, the device’s placement and orientation on a board, in conjunction
with the volume and temperature of air flowing past the unit present a broad range of
possible thermal solutions and problems for IC packaging. Typically a package cannot be capable
of handling a given power requirement unless the environmental conditions allow heat to dissipate.
When environmental or geometric constraints limit a BGA’s ability to dissipate heat, a copper or
aluminum heatsink is often used to provide an additional method for heat transfer. As with other
types of packages, the heatsinks for BGAs vary in design and methods of attachment. Most applications
recommend a maximum case temperature for the package. Various factors effect the case
temperature including ambient conditions and airflow. If the case temperature exceeds the recommended
rating, a heatsink may be required. Contact an Intel applications engineer for the product to
determine if a heatsink has been developed for the particular package or application.

 

[Chiropteran]

As it is 18W seems a bit high for netbooks, and the performance is probably on the low end for laptops, so it might be stuck somewhere in the middle.

There are currently already netbooks selling that use 18W+ CPUs. I can't see how it would be too high if it's already proven to be low enough.

 

[nonameo]

obligatory "I wonder how it will overclock" post.

So, I wonder how it will overclock. I know this defeats the purpose of a low power cpu, but something in me just has to know

 

[ShawnD1]

I don't understand why anyone would be impressed by low power chips. Aren't low power chips just the same as high power chips but they're severely underclocked and undervolted?

Seriously, I want someone in this thread to actually try this.
-Underclock your CPU so it's only running at 1GHz or less.
-Undervolt it until it's on the edge of stability.
-Unplug your CPU fan.
-Check the CPU temperature during Linpack testing.

I've actually done this before with an OG Phenom 9600 and the result is that they run at room temperature. Yes even a horribly designed Phenom X4 9600 Agena runs at room temperature when it's crippled to the point of being as slow as an Intel Atom. They also run at 0.9V in that case.

 

[busydude]

What about TDP of the chip? I am pretty sure it will not be <20W.

 

[cusideabelincoln]

Would be interesting to see how its performance compares to i5 processors with the same 18W TDP.

As it is 18W seems a bit high for netbooks, and the performance is probably on the low end for laptops, so it might be stuck somewhere in the middle.

Does 18W sound too high when this number includes both the CPU and GPU? Seems to be what bit-tech was insinuating.

 

[cbn]

I don't understand why anyone would be impressed by low power chips. Aren't low power chips just the same as high power chips but they're severely underclocked and undervolted?

I found this Hotchips 2004 link over at XS. Apparently low power and High power designs are quite different from each other.

http://www.hotchips.org/archives/hc16/3_Tue/9_HC16_Sess8_Pres2_bw.pdf (fourth slide illustrates the difference in efficiency)

My guess is that Ontario is a High power design.

 

[cbn]

obligatory "I wonder how it will overclock" post.

So, I wonder how it will overclock. I know this defeats the purpose of a low power cpu, but something in me just has to know

I'd like to know that also (especially considering Bobcat cores have OoO execution).

 

[JFAMD]

I don't understand why anyone would be impressed by low power chips. Aren't low power chips just the same as high power chips but they're severely underclocked and undervolted?

Not everyone in the world cares about performance.

I have a workstation that is on 24x7 for a variety of reasons. Wtih a quad core opteron (2.9GHz) and 4GB of memory I have it around 40W at idle.

I want something really quiet and really cool because 95% of the time it is not doing much of anything (but unfortunately needs to be on).

At $.11/KwH it is ~$38/year to power that. Anything that I could do to reduce it further would be good. I am already underclocked on the CPU and memory, it uses a notebook HD and is optimized in the BIOS to reduce power as much as possible.

And no, true low power chips can be very different.

 

[khon]

Searching through some benchmarks for comparison I gotta say the bobcat is not looking so good anymore.

For comparison I'm using the i3-330UM and the i5-520UM, both rated at 18W TDP for CPU+GPU, exactly the same as the AMD Ontario. Furthermore I'm using the AMD Athlon II X2 250u mentioned in the leak as being slightly faster than the Ontario

PassMark score:

Intel core i5-520UM: 1605
Intel core i3-330UM: 1295
AMD Athlon II X2 250u: 994

http://www.cpubenchmark.net/cpu_list.phphttp://www.cpubenchmark.net/cpu_lookup.php?cpu=AMD+Athlon+II+X2+250u

 

[cbn]

true low power chips can be very different.

JFAMD,

Can I ask you how many processes AMD uses for its chips?

Are we looking at just "high power" and "low power" in 2010 or do such things as "medium power" exist?

For example, are the current 2010 mainstream mobile chips "high power" designs binned for lower voltages or are they specially designed for "medium power"? How about the recently launched "Black Edition" laptop cpus?

 

[extra]

I don't think bobcat will be 18 watts. More like 10.

 

[cbn]

I don't think bobcat will be 18 watts. More like 10.

Yep, I can't imagine two bobcat cores using that much power. It might be the GPU plays a significant role in that 18 watt TDP.

I just wonder how Windows 7 starter plays with Ontario? It sounds like this little APU would make a nice multi-media machine (due to the graphics) but then look at what MS has stipulated:

http://windowsteamblog.com/Windows/...05/29/let-s-talk-about-windows-7-starter.aspx.

Or will MS change some of their restrictions with Google Chrome OS and/or Meego Netbook OS putting pressure on them in 2011? It would be great to be able to connect external DVD drive and 1080p TV/monitor to one of these little netbooks. Last time I checked the three applications restriction was removed but I don't know about the others.

 

[JFAMD]

JFAMD,

Can I ask you how many processes AMD uses for its chips?

Are we looking at just "high power" and "low power" in 2010 or do such things as "medium power" exist?

For example, are the current 2010 mainstream mobile chips "high power" designs binned for lower voltages or are they specially designed for "medium power"? How about the recently launched "Black Edition" laptop cpus?

I don't know what we do on the client side, I only know servers. We do 2 families, the 6000 which have SE (high performance), standard and HE (low power).

On the 4000 side we do standard, HE and EE (ultra low power).

They are all from the same dies, it is just binning, but on the client side you will see the bobcat core (which is optimized for low power) along with what they are using in Llano and then Bulldozer on the top. So there are things that make bobcat different from bulldozer. All of the people that say bobcat is just a few stripped down bulldozer cores don't understand it, it was a completely different design.

 

[IntelUser2000]

I don't understand why anyone would be impressed by low power chips. Aren't low power chips just the same as high power chips but they're severely underclocked and undervolted?

You can't compare a full fledged high performance chip that has been undervolted to the point of stability issues and will still consume over 20W to a chip that has been designed to be operated at far low power with a die size that is 1/5.

Bobcat will consume <20W at higher frequencies with die size that is under 100mm2 while including a decent on-die GPU.

I'd like to see how much the modern chips can be undervolted especially considering minimum voltages needed for proper functioning isn't too far away from nominal lower power states.

 

[Zstream]

Searching through some benchmarks for comparison I gotta say the bobcat is not looking so good anymore.

For comparison I'm using the i3-330UM and the i5-520UM, both rated at 18W TDP for CPU+GPU, exactly the same as the AMD Ontario. Furthermore I'm using the AMD Athlon II X2 250u mentioned in the leak as being slightly faster than the Ontario

PassMark score:

Intel core i5-520UM: 1605
Intel core i3-330UM: 1295
AMD Athlon II X2 250u: 994

http://www.cpubenchmark.net/cpu_list.phphttp://www.cpubenchmark.net/cpu_lookup.php?cpu=AMD+Athlon+II+X2+250u

Those benchmarks uploads are often quite wrong. I have bene using a celeronx2 culv and the scores are no where near what they state. The same with a few I5's we tested with.