Predict: The Next Generation Console Tech

[liolio]

I'm really iffy about MS going with a cell successor, it doesn't match their goals in the bigger plan. They've just release I don't remember which extension to I don't remember which programming language that allow to adress both CPUs cores and GPUs cores at the some time. THey will continue to push direct compute too. I can't see pushing a third type of hardware resources into the equation.

On the other hand I would not completely surprised if the next cell (an IBM only product) could be completely different beast than the first one (ie no SPUs), I remember some IBM engineer stating that the "heart" of the Cell was to continue to live in IBM roadmap. From IBM pov I'm sure the heart of Cell is the PPU, Powerpc A2 is for me a successor to xenon philosophy. I would not be too surprised if the is a new "cell" for IBM that it looks more like an A2 augmented by potent SIMDs than a Cell.

 

[Blazkowicz]

Some custom ARM Cortex A15(Vita = A9) and powervr 6(Vita-> SGX 543 MP4+) for ps4 annyone(and some SPUs for BC)?

I dunno, how fast is an A9 exactly, and how fast will be an A15?
if it's only playing catchup to decade old CPUs such as the pentium 4 and athlon 64, or PPU / xenon, it's not too sexy.

nvidia Denver would presumably be a more performance oriented ARM CPU core.

 

[ninzel]

What's their record like with predictions?

 

[Squilliam]

Will the cortex A18 or whatever be coming out before the next generation consoles hit?

 

[Heinrich4]

I dunno, how fast is an A9 exactly, and how fast will be an A15?
if it's only playing catchup to decade old CPUs such as the pentium 4 and athlon 64, or PPU / xenon, it's not too sexy.

nvidia Denver would presumably be a more performance oriented ARM CPU core.

I researched about A15* in sites and according to the manufacturer under the same clock reaches about 40% more performance than the A9.

The overall performance compared to other cpus still do not know for sure, only that information ** indicates that on 8 cores (14000 mips * 4) at 2GHz will probably reach at least 8 watts (2 core early 1900mW) performance comparable to the Intell Core 2 Extreme Quad QX9770.

* http://en.wikipedia.org/wiki/ARM_Cortex-A15_MPCore

** http://en.wikipedia.org/wiki/Instructions_per_second



Perhaps the universe without restrictions and limitations of portable with 100/150 Watts closed box (gpu + cpu, drive, HDD, PCB etc.) allow customization clocked at range 3/3.5GHz probably reach some good performance.

I think an ARM cpu could fit very recent statements from Sony(if) they really want to make life easier for developers and growing market of portable, perhaps using an architecture that is dominant in this segment (and very well documented and developed) ARM is the answer to their aspirations (friendly, relatively inexpensive and low TDP).

 

[homerdog]

The overall performance compared to other cpus still do not know for sure, only that information ** indicates that on 8 cores (14000 mips * 4) at 2GHz will probably reach at least 8 watts (2 core early 1900mW) performance comparable to the Intell Core 2 Extreme Quad QX9770.

Is that a typo or do you really suppose an 8 watt chip can perform like a QX9770? Because there is no way.

 

[Heinrich4]

Is that a typo or do you really suppose an 8 watt chip can perform like a QX9770? Because there is no way.

As i say we dont know,but only compare at mips side see the link.

See... " The overall performance compared to other cpus still do not know for sure, only that information (MIPs)**

 

[homerdog]

Well if you don't know for sure, let me be the one to tell you that no 8W chip will be on par with the QX9770 any time soon.

 

[Blazkowicz]

a close current thing would be the pentium G620, about half the performance of a QX9770 with (per hardware.fr) a max power use of 24 watts.

MIPS are meaningless, as it's a raw number of instructions regardless of the useful amount of work done. RISC processors get a high MIPS number but they use many compact, single instructions whereas with an x86 instruction set, they are more complex and get more work done.

so the chip would have 3x or 4x better performance/watt than the G620, rather than 6x.

 

[hoho]

http://www.brightsideofnews.com/news/2011/5/19/the-coming-war-arm-versus-x86.aspx?pageid=0 <- a benchmark of A8 vs mobile Barton vs VIA nano vs Atom N450. A8 isn't that bad considering it has roughly 8x lower peak memory bandwidth and probably far higher memory latency.

 

[homerdog]

a close current thing would be the pentium G620, about half the performance of a QX9770 with (per hardware.fr) a max power use of 24 watts.

The TDP of the G620 is 65W

 

[Heinrich4]

Well if you don't know for sure, let me be the one to tell you that no 8W chip will be on par with the QX9770 any time soon.

Sorry typo always happen to me (english is my third language)



Maybe only just numbers and estimates(about MIPs A15), and you're totally right, but maybe we should look at these data below (correct me if im wrong please):


Here talks about A9 with half performance Xeon 5450 and estimates A15 Cortex:

http://www.itproportal.com/2010/09/13/arm-cortex-a15-eagle-performance-benchmark-estimate/


See OMAP 4430 (dual core A9 Cortex 1GHz) reachs =~5700/5800 coremark
http://www.coremark.org/benchmark/index.php

See Xeon X5450 and others here
http://www.cpubenchmark.net/cpu_lookup.php?cpu=Intel+Xeon+X5450+@+3.00GHz

Specs about Xeon X5450
http://h18000.www1.hp.com/products/quickspecs/12849_na/12849_na.pdf

About A15 exclusively:
"2.5GHz reachs over 20000 Dmips barrier"
http://www.itproportal.com/2011/03/14/exclusive-arm-cortex-a15-40-cent-faster-cortex-a9/

ARM A9 "Calxeda" for servers:

http://www.itproportal.com/2011/03/14/calxeda-reveals-more-details-about-arm-based-server-chip/


So if ARM A9 dual core 1GHz with less than .5/.6 watt in some situations can match half performance Xeon X5450 with 120 watts at 3GHz,maybe is possible A15 8 core 2 GHz reach performance Intell Core 2 Extreme Quad QX9770 and who knows how this new architecture (A15) coming from manufacture dominating the portable market (growing fast and with large research and development) with low TDP can achieve efficiency compared to other.. even being a huge intell( they are not directed to the mobile market..yet)*.

Intell takes ARM for mobile market...
* http://www.pcworld.com/businesscent...takes_on_arm_for_mobile_processor_market.html

 

[pjbliverpool]

Here talks about A9 with half performance Xeon 5450 and estimates A15 Cortex:

http://www.itproportal.com/2010/09/13/arm-cortex-a15-eagle-performance-benchmark-estimate/

According to that, clock for clock and core for core an A15 is no faster than an A8.

i.e. "a Cortex A8 will be 5x less powerful than a dual core Cortex A15"

So 2.5x faster per core with the A15 running at 2.5Ghz and the A8 running at 1Ghz.

And according to the benchmarks in this link... http://www.brightsideofnews.com/news....aspx?pageid=0

...an A8 is roughly 1/3 to 1/2 as fast as an AthlonXP at the same clock speed.

Assuming an A64 is 30% faster than an AXP clock for clock and then Pentryn is 30% faster again, then a very rough calculation puts a single Penryn core at 3.5 - 5 times faster than an A15 core, clock for clock.

So accounting for the clock speed difference a 3Ghz Core2 Quad should be around 4 - 6 times faster than a quad A15 at 2.5Ghz.

 

[Blazkowicz]

The TDP of the G620 is 65W

wow
it's customary for low end desktop CPUs to use much less power than their stated TDP, but here Intel went truely over the board with it.
they want you to buy a core i3 2100T with a fat margin for them. or maybe a G620T, with a lower frequency so it looks less appealing to you.

 

[homerdog]

Yeah I guess so. But if you put that G620 in a console where developers could extract every ounce of performance out of it, it would probably get closer to that 65W TDP than it does in a PC.

So if ARM A9 dual core 1GHz with less than .5/.6 watt in some situations can match half performance Xeon X5450 with 120 watts at 3GHz,maybe is possible A15 8 core 2 GHz reach performance Intell Core 2 Extreme Quad QX9770 and who knows how this new architecture (A15) coming from manufacture dominating the portable market (growing fast and with large research and development) with low TDP can achieve efficiency compared to other.. even being a huge intell( they are not directed to the mobile market..yet)*.

Just a hunch, but I think pjbliverpool is closer to reality with his calculations than some esoteric MIPS benchmark. Intel processors are brutally fast in real-world applications.

 

[Blazkowicz]

consoles can run finely tuned software, but computers can run OCCT and furmark.

a precision I overlooked, figures are without using the IGP and the article says you'd have to add ten watts. it's from there.
http://www.hardware.fr/focus/43/pentium-g620-quelles-performances.html

 

[Heinrich4]

According to that, clock for clock and core for core an A15 is no faster than an A8.

i.e. "a Cortex A8 will be 5x less powerful than a dual core Cortex A15"

So 2.5x faster per core with the A15 running at 2.5Ghz and the A8 running at 1Ghz.

And according to the benchmarks in this link... http://www.brightsideofnews.com/news....aspx?pageid=0

...an A8 is roughly 1/3 to 1/2 as fast as an AthlonXP at the same clock speed.

Assuming an A64 is 30% faster than an AXP clock for clock and then Pentryn is 30% faster again, then a very rough calculation puts a single Penryn core at 3.5 - 5 times faster than an A15 core, clock for clock.

So accounting for the clock speed difference a 3Ghz Core2 Quad should be around 4 - 6 times faster than a quad A15 at 2.5Ghz.

Maybe you're right and now we're almost in the realm of speculation, these links* indicate that the A15 is higher than the A8, but still below core per core than the X86 can offer(if customised for console aplications reach 50% Core 2 Extreme QX9770 its an excelent result), but who knows with customizations and without limitation watts of portable and I said earlier, an A15 can provide considerable performance and very low consumption.

* http://arstechnica.com/business/news/2010/09/arms-eagle-has-landed-meet-the-a15.ars

http://www.xbitlabs.com/news/cpu/di...o_Compete_Against_x86_but_Not_Everywhere.html


And as posted earlier on the same link itproportal* talks about two core A9 has about half the estimated performance of the Xeon X5450 3 GHz, 120 watts. And we dont have any benchmarks A15,but is just numbers(anandtech benchs ipod2 indicates they are conservative in his numbers) if the manufacturer has admitted that about 40% more performance (same clock) than A9, A15 with 8 cores at 2/2.5GHz may reach about 2.8 times(just a guess here) the performance of the Xeon X5450. ( 2 core A9 = 50%[Xenon X5450] * 1.4 = 70% = 2 core A15 * 4 *=2.8) which can be considered a good performance at 8 watts (1.9 watts = A15 2 cores).

* http://www.itproportal.com/2010/09/13/arm-cortex-a15-eagle-performance-benchmark-estimate/

**Xeon X5450 benchmark.

http://www.cpubenchmark.net/cpu.php?cpu=Intel+Xeon+X5450+@+3.00GHz

Xeon X5450 specs

http://gadgets.softpedia.com/gadgets/Computer-Hardware/Intel-Xeon-X5450-processor-8487.html


We have to make some considerations for exercising a line of reasoning and I thank you for the patience and opportunity to exercise it with the arguments presented.

On site brightsideofnews they test only A8 which is inferior in performance to the A9*,which only now beginning to be employed and reaching the mainstream (ps vita, many cell phones, Iphone 2) * and the final conclusions on spoken clearly about it(conclusion part).

ARM cores are placed on die with GPU to a system on chip in a similar way to APU AMD Bobcats that has no FPU and they are working in synergy with GPUs in this regard.

And in conclusion, if the A8 cortex is superior to A9 cortex why used in Ipad2,PS Vita and and many cell phones that are being developed(why ARM dont use A8 cortex in your Calxeda server nodes?)?

It is likely that A15 with reasonable performance up (40% up according manufacturer) to to the A9 (see what is being created for the PSVita already in first generation of software!) is probably relative powerfull or at least a good cpu with sinergy with enough powerfull GPUs( 800 to 1000 strean processor AMD like ,384+ Nvidia Fermi like or even 24/32 cores powervr 6...) for next gen console.


* A8 vs A9 cortex
http://www.eeejournal.com/search/label/Benchmarks

List ARM processors:

http://en.wikipedia.org/wiki/List_of_ARM_microprocessor_cores

Another benchmark A8

http://aceshardware.freeforums.org/some-cortex-a8-benchmarks-vs-x86-t971.html

of topic....amazing performance for cell phones market:

A15 dual core 2.5GHz + Rogue PowerVR 6 = 210 Gflops,5Gpixel/13 gpixel,350 million polys/second
http://forum.beyond3d.com/showthread.php?t=59574

 

[Heinrich4]

Yeah I guess so. But if you put that G620 in a console where developers could extract every ounce of performance out of it, it would probably get closer to that 65W TDP than it does in a PC.




Just a hunch, but I think pjbliverpool is closer to reality with his calculations than some esoteric MIPS benchmark. Intel processors are brutally fast in real-world applications.

You're absolutely right ,is just a guess, but we dont have benchmarks A15 and we can only speculate, but there are some indications(40% more performance than A9 at same clock) that perhaps this cpu is an excellent choice for a game console that is on limitations 100/150 watt TDP (cpu, gpu, drive, HDD , PCBs etc.). Maybe with ARM cpu with good performance and very low wattage manufacturers they free to place with eDRAM, more powerful gpu.

 

[Megadrive1988]

I'm still expecting a tailor-made custom Kepler or Maxwell generation GPU in PS4. That alone will make PS4 game graphics so much better than those of the PS3. It really doesn't matter as much on the CPU side of things. Sure it would be nice to get 32-64 SPU chip in there, but a 16 SPU chip combined with a massively more powerful GPU will still insure PS4 graphics leap over 360/PS3.

 

[pjbliverpool]

And as posted earlier on the same link itproportal* talks about two core A9 has about half the estimated performance of the Xeon X5450 3 GHz, 120 watts. And we dont have any benchmarks A15,but is just numbers(anandtech benchs ipod2 indicates they are conservative in his numbers) if the manufacturer has admitted that about 40% more performance (same clock) than A9, A15 with 8 cores at 2/2.5GHz may reach about 2.8 times(just a guess here) the performance of the Xeon X5450. ( 2 core A9 = 50%[Xenon X5450] * 1.4 = 70% = 2 core A15 * 4 *=2.8) which can be considered a good performance at 8 watts (1.9 watts = A15 2 cores).

To be honest, common sense tells us we can ignore the claim of a dual A9 having half the performance of an X5450. It simply isn't possible given the size, power and architecture differences. For this to be true it would need more performance per clock/core than a Sandybridge at about 1/100th the power draw. Cleary a ludicrous expectation.

On site brightsideofnews they test only A8 which is inferior in performance to the A9*,which only now beginning to be employed and reaching the mainstream (ps vita, many cell phones, Iphone 2) * and the final conclusions on spoken clearly about it(conclusion part).

And in conclusion, if the A8 cortex is superior to A9 cortex why used in Ipad2,PS Vita and and many cell phones that are being developed(why ARM dont use A8 cortex in your Calxeda server nodes?)?

A8 obviously is slower than A9. But we are comparing core for core, clock for clock performance, not overall performance with the A9 using more cores at a higher clock speed.

That said, even if we take A15 to be twice as fast as A8 core for core and clock for clock, a quad version of it is still going to be 1/3 to 1/2 the performance of an X5450. Or an Octo core version, assuming perfect core scaling is going to be at 2/3rds to equal in performance to an X5450. But thats with all factors leaning in the A15's favour.

The 2.8x claim though is totally unrealistic.