anand's jaguar article

[gorobei]

http://www.anandtech.com/show/6976/...powering-xbox-one-playstation-4-kabini-temash

the article is generally favorable towards jaguar:

The Jaguar power story actually looks very good, it's just hampered by traditional PC legacy. None of the launch APUs here support the low power IOs necessary to drive platform power down even further. AMD is getting very close though. Jaguar's core power is easily sub-2W for lightweight tablet tasks, the rest of the platform (excluding display) drives it up to 4 - 7W. AMD definitely has the right building blocks to go after truly low power tablets in a major way, should it have the resources and bandwidth to do so.

In its cost and power band, Jaguar is presently without competition. Intel’s current 32nm Saltwell Atom core is outdated, and nothing from ARM is quick enough. It’s no wonder that both Microsoft and Sony elected to use Jaguar as the base for their next-generation console SoCs, there simply isn’t a better option today. As Intel transitions to its 22nm Silvermont architecture however Jaguar will finally get some competition. For the next few months though, AMD will enjoy a position it hasn’t had in years: a CPU performance advantage.

I can’t stress enough how important it is that AMD continues to focus on driving the single threaded performance of its cat-line of cores. Second chances are rare in this business, but that’s exactly what AMD has been offered with the rise of good enough computing. Jaguar vs. Atom is the best CPU story AMD has had in years. Regular updates to the architecture coupled with solid execution are necessary to ensure that history doesn’t repeat itself in a new segment of AMD’s business.

i've bolded some parts. Is win8 or winBlue the low power OS he's talking about or is it android?

the good enough computing part seems odd since the amd prototype laptop in the kabini review seems underwhelming in the games aspect. http://www.anandtech.com/show/6974/amd-kabini-review/2
admittedly its 1080p res but any future gen x86 tablet is going to be at least in the 1080 or higher range. am i missing something or is it going to rely on gcn and CMT optimizations? or is it just the 2W ipad-esque battery life that is the real attraction?

 

[galego]

http://www.anandtech.com/show/6976/...powering-xbox-one-playstation-4-kabini-temash

the article is generally favorable towards jaguar:


i've bolded some parts. Is win8 or winBlue the low power OS he's talking about or is it android?

the good enough computing part seems odd since the amd prototype laptop in the kabini review seems underwhelming in the games aspect. http://www.anandtech.com/show/6974/amd-kabini-review/2
admittedly its 1080p res but any future gen x86 tablet is going to be at least in the 1080 or higher range. am i missing something or is it going to rely on gcn and CMT optimizations? or is it just the 2W ipad-esque battery life that is the real attraction?

Sorry but says IO not OS.

One of the main advantage of jaguar cores is on their adaptability. AMD is using them also for their custom chip division. AMD can go from an ultra-low power chip for a kiosk to a gaming console CPU with 102.4 GFLOPs like that used in the PS4 (note that an i7-3770k has 112 GFLOPs).

 

[strata8]

One of the main advantage of jaguar cores is on their adaptability. AMD is using them also for their custom chip division. AMD can go from an ultra-low power chip for a kiosk to a gaming console CPU with 102.4 GFLOPs like that used in the PS4 (note that an i7-3770k has 112 GFLOPs).

Around 25-30W for 8 cores at 2Ghz if I'm estimating correctly. Very impressive.

 

[Enigmoid]

Competetive vs ARM? Even ivy bridge ULV is barely competitive vs arm in perf/watt (arm is much more efficient but can't reach the performance levels which is what saves ivy bridge).

Outperforms a dual core a-15 with a tdp of 3-4 watts vs a 15 watt tdp.

What about tegra 4 and snapdragon 800? Both with significantly lower power use.

In the graphics tests it looks like its a bit more than twice as fast as an adreno 320 (what about 330?).

Much better than A6x and adreno 320 until you look at the power use.

 

[strata8]

No one said the 15W was competitive with ARM. But the 3.9W certainly is.

Kraken: 9800ms
GL/DXBenchmark: 17fps

Of course when Snapdragon 800 and Tegra 4 release it won't be.

 

[NTMBK]

No one said the 15W was competitive with ARM. But the 3.9W certainly is.

Kraken: 9800ms
GL/DXBenchmark: 17fps

Of course when Snapdragon 800 and Tegra 4 release it won't be.

But what power usage do we expect Tegra 4 to be? There's a reason why Shield has a fan...

 

[zlatan]

But what power usage do we expect Tegra 4 to be? There's a reason why Shield has a fan...

Depends on the clocks, but the fastest Tegra 4 will consume 8 W. But this is normal even a Snapdragon 800 consume 7 W.

 

[joshhedge]

CPU with 102.4 GFLOPs like that used in the PS4 (note that an i7-3770k has 112 GFLOPs).

Don't compare the two, the i7 is in a vastly greater league.

 

[Exophase]

Drawing wide conclusions of performance relative to Cortex-A15 from Kraken alone is a travesty. Look at Geekbench of Exynos 5410 vs Bobcat, for instance (look at the single threaded tests, note that these are the same clock):

http://browser.primatelabs.com/geekbench2/compare/1950660/1159110

This is using the best scores I could find for both. The 64-bit Linux scores are a fair bit better for E-350 than the Windows ones.

 

[sontin]

Depends on the clocks, but the fastest Tegra 4 will consume 8 W. But this is normal even a Snapdragon 800 consume 7 W.

Tegra 4 has a TDP of 5W. 8W is nonsense. nVidia would not even come close to the 4-5h battery time of shield.

 

[insertcarehere]

No one said the 15W was competitive with ARM. But the 3.9W certainly is.

Kraken: 9800ms
GL/DXBenchmark: 17fps

Of course when Snapdragon 800 and Tegra 4 release it won't be.

If we look at the 3Dmark score and assume perfect downward scaling, the 3.9W part will be outperformed by the Snapdragon 600.

 

[strata8]

If we look at the 3Dmark score and assume perfect downward scaling, the 3.9W part will be outperformed by the Snapdragon 600.

In the Physics score, sure, but they're head to head in the graphics score and its only outperformed by 7% in the Ice Storm score.

Drawing wide conclusions of performance relative to Cortex-A15 from Kraken alone is a travesty. Look at Geekbench of Exynos 5410 vs Bobcat, for instance (look at the single threaded tests, note that these are the same clock):

http://browser.primatelabs.com/geekbench2/compare/1950660/1159110

This is using the best scores I could find for both. The 64-bit Linux scores are a fair bit better for E-350 than the Windows ones.

Hasn't it been established that Geekbench can't be used to compare across different architectures? The disparity in the Linux scores certainly shows that it's not even reliable across different OSs.

 

[Exophase]

Hasn't it been established that Geekbench can't be used to compare across different architectures? The disparity in the Linux scores certainly shows that it's not even reliable across different OSs.

Established by who, and why would it be okay for the same architecture but not different ones? In this case the OS (kernel anyway) and compiler are close enough, a lot closer than Windows vs Linux would be.

 

[SiliconWars]

AMD have got this wrong like I said earlier - Temash should have had it's own small die designed purely for tablets and with the excessive legacy and I/O removed. As Anand indicated it is ripe for this kind of mid-life refresh. I'd also like to see a single CU version as that would take it into 7" tablet range with some very good performance.

 

[galego]

Don't compare the two, the i7 is in a vastly greater league.

Less than 10 GFLOPs between both or Epic selecting an i7 in their comparison PC-PS4 say the contrary. Unless you refer to power consumption, then I agree, the i7 is much more power hungry.

 

[Phynaz]

Less than 10 GFLOPs between both or Epic selecting an i7 in their comparison PC-PS4 say the contrary. Unless you refer to power consumption, then I agree, the i7 is much more power hungry.

You'll find that most posters here aren't going to fall for a theoretical GFLOPs argument.

You aren't being novel trying this approach.

 

[Exophase]

We're comparing CPU power using GFLOPs now?

Really?

Besides, that comparison is totally wrong because you're using single-precision GFLOPs for PS4 and double-precision for the i7. These are the correct numbers:
SB/IB: 16 SP FLOPs/cycle (8 FADD + 8 FMUL), 8 DP FLOPs/cycle (4 FADD + 4 FMUL)
Jaguar: 8 SP FLOPs/cycle (4 FADD + 4 FMUL), 3 DP FLOPs/cycle (2 FADD + 1 FMUL)

i7-3770K: 3.5GHz x 4 cores = 224 SP FLOPS, 112 DP FLOPS
PS4: 1.6GHz x 8 cores = 102.4 SP FLOPS, 38.4 DP FLOPS

Double precision FLOPs will rarely matter for games.

 

[tokie]

What's most interesting is that the Xbox/PS4 will be relying heavily on multithreading to eek out decent CPU performance.

I can't imagine running something like SC2 on Jaguar, since it would crush it.

 

[LogOver]

What's most interesting is that the Xbox/PS4 will be relying heavily on multithreading to eek out decent CPU performance.

I can't imagine running something like SC2 on Jaguar, since it would crush it.

Or may be not. Big game studios don't like to spend money/time/effort on optimizations. I expect we will see beautiful graphics and great image quality (because of powerful GPUs) but weak physics and AI.

 

[Olikan]

Or may be not. Big game studios don't like to spend money/time/effort on optimizations. I expect we will see beautiful graphics and great image quality (because of powerful GPUs) but weak physics and AI.

what?
game engines that use defered shadding, is a clear use optimization for consoles.....and every damm company uses that

 

[Riek]

What's most interesting is that the Xbox/PS4 will be relying heavily on multithreading to eek out decent CPU performance.

I can't imagine running something like SC2 on Jaguar, since it would crush it.

with something like SC2, you mean something which is written with very poor threading/load balancing in mind i assume?

Why would a developer studio who develop software for xbox/ps4 and maybe pc write a game so it uses 1/8th of the available power?

 

[galego]

We're comparing CPU power using GFLOPs now?

Really?

Besides, that comparison is totally wrong because you're using single-precision GFLOPs for PS4 and double-precision for the i7. These are the correct numbers:
SB/IB: 16 SP FLOPs/cycle (8 FADD + 8 FMUL), 8 DP FLOPs/cycle (4 FADD + 4 FMUL)
Jaguar: 8 SP FLOPs/cycle (4 FADD + 4 FMUL), 3 DP FLOPs/cycle (2 FADD + 1 FMUL)

i7-3770K: 3.5GHz x 4 cores = 224 SP FLOPS, 112 DP FLOPS
PS4: 1.6GHz x 8 cores = 102.4 SP FLOPS, 38.4 DP FLOPS

Double precision FLOPs will rarely matter for games.

We don't have any benchmark or other info now. I assume that you did a typo and you mean GFLOPs instead of FLOPs.

The 102.4 GFLOPs are from the jaguar cores alone. There are 4 extra CUs for compute that provide 409.6 GFLOP if needed. Therefore we have a theoretical maximum of

PS4: 512 GFLOPs
i7-3770K: 224 GFLOPs

In case I did not mistake now

 

[monstercameron]

We don't have any benchmark or other info now. I assume that you did a typo and you mean GFLOPs instead of FLOPs.

The 102.4 GFLOPs are from the jaguar cores alone. There are 4 extra CUs for compute that provide 409.6 GFLOP if needed. Therefore we have a theoretical maximum of

PS4: 512 GFLOPs
i7-3770K: 224 GFLOPs

In case I did not mistake now

Then what about the total gflop number of the 3770k with CPU & GPU?

 

[ShintaiDK]

We don't have any benchmark or other info now. I assume that you did a typo and you mean GFLOPs instead of FLOPs.

The 102.4 GFLOPs are from the jaguar cores alone. There are 4 extra CUs for compute that provide 409.6 GFLOP if needed. Therefore we have a theoretical maximum of

PS4: 512 GFLOPs

Its a shame the PS4 is so weak then. The PS3 was marketed as 1Tflop.

 

[lefty2]

From what I see Kabini is designed to be dirt cheap to manufacture and ment for cheap netbooks, therefore it's competing against Celeron, not Core i3/i5.
The Pentium 2020M that they compair it with is a 35W part and costs $134.00. I don't know what the actual price of the A4 5000 is, but I assume it's going to be a lot less than that.