AMD CPU speculation... and expert conjecture

[juanrga]

Huumm. This looks all suspicious. Then a simple search reveals that this technical paper is being cited in news sites that claim that Intel will compete with ARM on mobiles. Huuum more suspicion. Then we take a look to the conference website and we found Intel name among 'sponsors'

http://www.carch.ac.cn/~hpca19

Then we search the main author of the article and... surprise surprise he has an Intel folk as collaborator and several Intel students

http://research.cs.wisc.edu/vertical/wiki/index.php/Main/People

You're confusing an IEEE conference sponsor with the actual people who wrote that particular paper, which largely have nothing to do with the conference other than submitting their paper for review and presenting it. The conference doesn't pay people or provide funding to write papers.

There is a huge difference. A conference sponsor means they get their names on the SWAG (stuff we all get, t-shirts, coffee mugs, and crap like that) and signs and other hand outs that are given to the people who PAY to go to the conference.

There is an Intel person listed as a collaborator for the research group, but there is also a Broadcomm person. Broadcomm uses ARM cores in their products not x86 cores.

"Support for this research was provided by NSF grants CCF-0845751, CCF-0917238, and CNS-0917213,
and the Cisco Systems Distinguished Graduate Fellowship."

Cisco uses X86 and ARM products so your Intel bias assumption doesn't have much merit.

It's amazing how quick you are to cut down the paper but take marketing slides as gospel.

What is really amazing is how you have snipped relevant parts of my analysis of their paper, such as their scaling methodology and power consumption measurement favouring Intel.

I wrote 'sponsors' not sponsors.

As mentioned before, there are three (3) Intel persons in his research group (one collaborator plus two students), not only one (1).

No sure why you mention Broadcom and Cisco. Do they have a historical record of dishonest competition/cheating? Also, do you believe that because there is people from other companies the possibility of aggressive 'sponsoring' is eliminated? AMD, Nvidia, and VIA were members of SYSMARK consortium and still Intel cheated the SYSMARK benchmarks for artificially increasing the score of its products.

If that article that you mentioned was right and Intel (x86) is able to compete with ARM face to face, then Intel had not been caught using its ICC compiler to cheat the Antutu benchmarks this year.

The cheated benchmark gave ARM processors a mere 50% upgrade in performance whereas Intel chips were seeing increases of up to 292%.

Pay attention. The ICC biased CPU dispatcher gives Intel chips about a 15% advantage over the AMD chips. When fighting ARM, Intel had to give its chips up to ~10x more advantage.

http://www.androidauthority.com/analyst-says-intel-beating-arm-in-benchmarks-was-a-sham-243526/

http://www.eetimes.com/author.asp?section_id=36&doc_id=1318894

http://www.slashgear.com/intel-atom-z2580-antutu-benchmark-falls-20-following-revision-13290307/

http://seekingalpha.com/article/1550372-intel-and-antutu-mea-culpa

Why did they choose Intel but not AMD chips? If they had tested Bulldozer FX instead of an SB i7, they had found poor efficiency and even poor performance in several tests. Why did they not test a modern ARM chip?

When it comes to conference papers they're often submitted a year in advance of the conference. And the work probably started a year or more before that, so you're looking at a 2+ year lag from start to presentation (Feb 2013).

So the question why they didn't test an A15 is because they simply weren't out yet. Apple was the first to ship a phone with one and that wasn't until September 2012, which is only 5 months before the paper was presented.

They're not doing speculation here. They have real world hardware they had to get in the lab, code to write, testing, getting results, taking measurements, getting feedback from advisers, peer reviews, etc.

They used what they had access to. The i7 Sandy is old as well. It came out in January 2011, which is around the time A9s were introduced to the market. I.e the first Tegra 2 phone in January 2011.

http://www.networkworld.com/news/2011/010511-nvidia-shows-off-lg-optimus.html

I have submitted conference and contest papers a pair of months in advance. However, there are other objections to your hypothesis. The first, that they cite references published so late as 2013. The second, that they mention the A15 and even give some numbers for it. The third, that they claim that they evaluated but rejected more modern SoCs such as Tegra 3 before starting the lab work.

Now explain us: how did they evaluate a 2012 product if their lab work was ready in early 2011 according to you?

The facts are:
- They rejected better ARM designs with improved performance and efficiency.
- They didn't test Bulldozer FX or similar ones but the best possible x86 chips.
- The scaling methodology and power measurement favoured Intel.
- Their compiler choices favoured Intel.
- «they believe the results would look much the same — in terms of the relationship between power and performance» when looking at newer ARM designs, including the 64-bit ARMv8 architecture.
- They have 3 Intel people in their small research group, but nobody from ARM or AMD.

All together implies bias pro Intel.

Don't be fooled my friend. That is not for CPUs. I was discussing only CPUs.

That link is for Intel MIC project. The Xeon Phi coprocessor is somewhat based in x86. But it doesn't run the same x86 binaries that you can run in an i7-3770k for instance. For using the Phi you have to recompile the code at least and if you want to use all its potential you have to rewrite the code.

No one is making CPU only HPC from here on out. Even IBM systems are incorporating NVidia GPUs. BlueGene/Q is the only one that could be considered CPU only on the top500 list but it's floating point accelerator is essentially a GPU.

Your beloved project denver (Tegra 6) isn't getting it's massive performance gains from the ARM cores. The 100x performance increase is largely due to the Maxwell GPU cores that support CUDA, instead of their ancient GeForce ULP cores.

Recompiling for HPC environments is common place. They're all unique with various bandwidth, storage, cpu variations. Dig on the Phi all you want but it currently powers the #1 supercomputer in the world.

The same will have to be done for a supercomputer powered by Tegra 6. It will mostly be CUDA code with some ARM code for moving data around.

If you pay attention to the part of my message that _you snipped_. I already said you that CPU only HPCs are being abandoned. I was the first one in this forum who mentioned (it was many pages ago) how IBM joined to Nvidia to develop heterogeneous POWER8+CUDA supercomputers. You don't need to repeat what I already know.

I also said you that

ARM+CUDA or ARM+GCN will be the interesting combos.

What apparently you don't understand is that up to a 40% of the power consumption of heterogeneous supercomputers comes from the CPUs alone. Therefore it doesn't make many sense to improve the efficiency of the GPGPU by an order of magnitude without improving the efficiency of the CPUs first. That is why there are projects to replace x86 CPUs by ARM CPUs.

Nobody is saying you that Denver will be 100x faster than A9. But the A15 core is about a 70% faster than the previous A9 core. Nvidia claims that Denver will be faster than the last A57. Therefore, Denver being ~3x faster than A9 is a reasonable estimation.

My point was that once you recompile then the benefit of compatibility with legacy x86 binaries is lost. Sure that the Phi powers the #1 in top500, but not in green500, because that #1 is obtained by a brute force approach:

Looking at the latest data, what’s particularly astounding about Tianhe-2 is simply how large it is. Placing on the Top500 list requires both efficiency and brute force, and in the case of Tianhe-2 there’s an unprecedented amount of brute force in play. The official power consumption rating for Tianhe-2 is 17.8 megawatts, more than double Titan’s 8.2MW.

Take the old Titan, double it, and you recover the #1 (approx.). But as explained to you before, this pure brute approach doesn't work for exascale supercomputing. You cannot take a x86-based supercomputer and scale it 1000x times because you cannot generate the immense power needed for that beast. That is why HPC experts are developing an ARM supercomputer for exascale.

Also you seem to not understand that the performance of the Xeon Phi has very little to do with x86, but is directly related to an _extension_ of the x86 ISA which is exclusive for Phi. Follow my advice, don't be fooled by Intel marketing ;-)

I doubt any supercomputer will be powered by Tegra 6. Who said you that?

Nvidia has already ported CUDA to ARM for instance. CUDA will be not used in servers.

Or course CUDA will be used in NVidia servers. That's how you program/execute Kepler/Maxwell cores.

https://developer.nvidia.com/content/cuda-arm-platforms-now-available

Read the link that you give about CUDA on ARM.

How many times the word "server" appears? Zero (0). How many times the word "supercomputer" and derivative (supercomputers, supercomputing, HPC) appears? Seven (7).

 

[8350rocks]

Huumm. This looks all suspicious. Then a simple search reveals that this technical paper is being cited in news sites that claim that Intel will compete with ARM on mobiles. Huuum more suspicion. Then we take a look to the conference website and we found Intel name among 'sponsors'

http://www.carch.ac.cn/~hpca19

Then we search the main author of the article and... surprise surprise he has an Intel folk as collaborator and several Intel students

http://research.cs.wisc.edu/vertical/wiki/index.php/Main/People

You're confusing an IEEE conference sponsor with the actual people who wrote that particular paper, which largely have nothing to do with the conference other than submitting their paper for review and presenting it. The conference doesn't pay people or provide funding to write papers.

There is a huge difference. A conference sponsor means they get their names on the SWAG (stuff we all get, t-shirts, coffee mugs, and crap like that) and signs and other hand outs that are given to the people who PAY to go to the conference.

There is an Intel person listed as a collaborator for the research group, but there is also a Broadcomm person. Broadcomm uses ARM cores in their products not x86 cores.

"Support for this research was provided by NSF grants CCF-0845751, CCF-0917238, and CNS-0917213,
and the Cisco Systems Distinguished Graduate Fellowship."

Cisco uses X86 and ARM products so your Intel bias assumption doesn't have much merit.

It's amazing how quick you are to cut down the paper but take marketing slides as gospel.

What is really amazing is how you have snipped relevant parts of my analysis of their paper, such as their scaling methodology and power consumption measurement favouring Intel.

I wrote 'sponsors' not sponsors.

As mentioned before, there are three (3) Intel persons in his research group (one collaborator plus two students), not only one (1).

No sure why you mention Broadcom and Cisco. Do they have a historical record of dishonest competition/cheating? Also, do you believe that because there is people from other companies the possibility of aggressive 'sponsoring' is eliminated? AMD, Nvidia, and VIA were members of SYSMARK consortium and still Intel cheated the SYSMARK benchmarks for artificially increasing the score of its products.

If that article that you mentioned was right and Intel (x86) is able to compete with ARM face to face, then Intel had not been caught using its ICC compiler to cheat the Antutu benchmarks this year.

The cheated benchmark gave ARM processors a mere 50% upgrade in performance whereas Intel chips were seeing increases of up to 292%.

Pay attention. The ICC biased CPU dispatcher gives Intel chips about a 15% advantage over the AMD chips. When fighting ARM, Intel had to give its chips up to ~10x more advantage.

http://www.androidauthority.com/analyst-says-intel-beating-arm-in-benchmarks-was-a-sham-243526/

http://www.eetimes.com/author.asp?section_id=36&doc_id=1318894

http://www.slashgear.com/intel-atom-z2580-antutu-benchmark-falls-20-following-revision-13290307/

http://seekingalpha.com/article/1550372-intel-and-antutu-mea-culpa

Why did they choose Intel but not AMD chips? If they had tested Bulldozer FX instead of an SB i7, they had found poor efficiency and even poor performance in several tests. Why did they not test a modern ARM chip?

When it comes to conference papers they're often submitted a year in advance of the conference. And the work probably started a year or more before that, so you're looking at a 2+ year lag from start to presentation (Feb 2013).

So the question why they didn't test an A15 is because they simply weren't out yet. Apple was the first to ship a phone with one and that wasn't until September 2012, which is only 5 months before the paper was presented.

They're not doing speculation here. They have real world hardware they had to get in the lab, code to write, testing, getting results, taking measurements, getting feedback from advisers, peer reviews, etc.

They used what they had access to. The i7 Sandy is old as well. It came out in January 2011, which is around the time A9s were introduced to the market. I.e the first Tegra 2 phone in January 2011.

http://www.networkworld.com/news/2011/010511-nvidia-shows-off-lg-optimus.html

I have submitted conference and contest papers a pair of months in advance. However, there are other objections to your hypothesis. The first, that they cite references published so late as 2013. The second, that they mention the A15 and even give some numbers for it. The third, that they claim that they evaluated but rejected more modern SoCs such as Tegra 3 before starting the lab work.

Now explain us: how did they evaluate a 2012 product if their lab work was ready in early 2011 according to you?

The facts are:
- They rejected better ARM designs with improved performance and efficiency.
- They didn't test Bulldozer FX or similar ones but the best possible x86 chips.
- The scaling methodology and power measurement favoured Intel.
- Their compiler choices favoured Intel.
- «they believe the results would look much the same — in terms of the relationship between power and performance» when looking at newer ARM designs, including the 64-bit ARMv8 architecture.
- They have 3 Intel people in their small research group, but nobody from ARM or AMD.

All together implies bias pro Intel.

Don't be fooled my friend. That is not for CPUs. I was discussing only CPUs.

That link is for Intel MIC project. The Xeon Phi coprocessor is somewhat based in x86. But it doesn't run the same x86 binaries that you can run in an i7-3770k for instance. For using the Phi you have to recompile the code at least and if you want to use all its potential you have to rewrite the code.

No one is making CPU only HPC from here on out. Even IBM systems are incorporating NVidia GPUs. BlueGene/Q is the only one that could be considered CPU only on the top500 list but it's floating point accelerator is essentially a GPU.

Your beloved project denver (Tegra 6) isn't getting it's massive performance gains from the ARM cores. The 100x performance increase is largely due to the Maxwell GPU cores that support CUDA, instead of their ancient GeForce ULP cores.

Recompiling for HPC environments is common place. They're all unique with various bandwidth, storage, cpu variations. Dig on the Phi all you want but it currently powers the #1 supercomputer in the world.

The same will have to be done for a supercomputer powered by Tegra 6. It will mostly be CUDA code with some ARM code for moving data around.

If you pay attention to the part of my message that _you snipped_. I already said you that CPU only HPCs are being abandoned. I was the first one in this forum who mentioned (it was many pages ago) how IBM joined to Nvidia to develop heterogeneous POWER8+CUDA supercomputers. You don't need to repeat what I already know.

I also said you that

ARM+CUDA or ARM+GCN will be the interesting combos.

What apparently you don't understand is that up to a 40% of the power consumption of heterogeneous supercomputers comes from the CPUs alone. Therefore it doesn't make many sense to improve the efficiency of the GPGPU by an order of magnitude without improving the efficiency of the CPUs first. That is why there are projects to replace x86 CPUs by ARM CPUs.

Nobody is saying you that Denver will be 100x faster than A9. But the A15 core is about a 70% faster than the previous A9 core. Nvidia claims that Denver will be faster than the last A57. Therefore, Denver being ~3x faster than A9 is a reasonable estimation.

My point was that once you recompile then the benefit of compatibility with legacy x86 binaries is lost. Sure that the Phi powers the #1 in top500, but not in green500, because that #1 is obtained by a brute force approach:

Looking at the latest data, what’s particularly astounding about Tianhe-2 is simply how large it is. Placing on the Top500 list requires both efficiency and brute force, and in the case of Tianhe-2 there’s an unprecedented amount of brute force in play. The official power consumption rating for Tianhe-2 is 17.8 megawatts, more than double Titan’s 8.2MW.

Take the old Titan, double it, and you recover the #1 (approx.). But as explained to you before, this pure brute approach doesn't work for exascale supercomputing. You cannot take a x86-based supercomputer and scale it 1000x times because you cannot generate the immense power needed for that beast. That is why HPC experts are developing an ARM supercomputer for exascale.

Also you seem to not understand that the performance of the Xeon Phi has very little to do with x86, but is directly related to an _extension_ of the x86 ISA which is exclusive for Phi. Follow my advice, don't be fooled by Intel marketing ;-)

I doubt any supercomputer will be powered by Tegra 6. Who said you that?

Nvidia has already ported CUDA to ARM for instance. CUDA will be not used in servers.

Or course CUDA will be used in NVidia servers. That's how you program/execute Kepler/Maxwell cores.

https://developer.nvidia.com/content/cuda-arm-platforms-now-available

Read the link that you give about CUDA on ARM.

How many times the word "server" appears? Zero (0). How many times the word "supercomputer" and derivative (supercomputers, supercomputing, HPC) appears? Seven (7).

Don't take off your tinfoil hat, they might zap your brain with a ray gun!

 

[The Q6660 Inside]

@juanrga, don't hype up ARM, you'll be disappointed very soon my friend, don't fall for the marketing slides and think realistically...

Now, to burn some time and to put the nail in the coffin, the end all be all power consumption (Full load, 4 hours a day, 3 years, "man you need a 3930K for these long workloads" scenario) formula is here!

4670\4770K Power consumption: http://imgur.com/3IIf21K
8350 Power consumption :http://imgur.com/j37Dx29
OCd 8350 with H80i\9370 with AMD FX CLC: http://imgur.com/GhpPByc

4670K (On sale) Final Bill: http://imgur.com/F9m72qW
4770K Final Bill: http://imgur.com/GC4rqOY
4670K (On sale) final bill with $80 CLC or comparable air cooler: http://imgur.com/nXlTmJ9
4770K Final Bill with $80 CLC or comparable air cooler: http://imgur.com/JQO951U
8350 Final Bill: http://imgur.com/hPH95kg
OCd 8350 with CLC/9370 With FX AIO final bill: http://imgur.com/Y8AvLUX
9370 final bill if you have an old H70 lying around and sold the FX AIO CLC : http://imgur.com/BCGtQny

As we can see, the only time Intel truly wins in this extreme case scenario is with the 4670K (in this case, barely beating out the 8350 in final cost by a whopping $7, should the 4670K have been priced @$220 and not $240...

/Intelneverrealisticallypaysitselfoff

 

[GOM3RPLY3R]

@juanrga, don't hype up ARM, you'll be disappointed very soon my friend, don't fall for the marketing slides and think realistically...

Now, to burn some time and to put the nail in the coffin, the end all be all power consumption (Full load, 4 hours a day, 3 years, "man you need a 3930K for these long workloads" scenario) formula is here!

4670\4770K Power consumption: http://imgur.com/3IIf21K
8350 Power consumption :http://imgur.com/j37Dx29
OCd 8350 with H80i\9370 with AMD FX CLC: http://imgur.com/GhpPByc

4670K (On sale) Final Bill: http://imgur.com/F9m72qW
4770K Final Bill: http://imgur.com/GC4rqOY
4670K (On sale) final bill with $80 CLC or comparable air cooler: http://imgur.com/nXlTmJ9
4770K Final Bill with $80 CLC or comparable air cooler: http://imgur.com/JQO951U
8350 Final Bill: http://imgur.com/hPH95kg
OCd 8350 with CLC/9370 With FX AIO final bill: http://imgur.com/Y8AvLUX
9370 final bill if you have an old H70 lying around and sold the FX AIO CLC : http://imgur.com/BCGtQny

As we can see, the only time Intel truly wins in this extreme case scenario is with the 4670K (in this case, barely beating out the 8350 in final cost by a whopping $7, should the 4670K have been priced @$220 and not $240...

/Intelneverrealisticallypaysitselfoff

Just to throw it out there, what about the 3930k? Honestly imo for price to performance its the best Intel processor (not including the Xeons). Sure it's $570 (4930k @ $580), however, it's literally a Extreme series under clocked. For the same price of an extreme series, you could get that, a motherboard for it, and a h100i (or hyper evo 212 depending on what type of person you are), for about the same price, then overclock it to 4.2, of even 4.5 with really no problems. ^_^

 

[The Q6660 Inside]

@juanrga, don't hype up ARM, you'll be disappointed very soon my friend, don't fall for the marketing slides and think realistically...

Now, to burn some time and to put the nail in the coffin, the end all be all power consumption (Full load, 4 hours a day, 3 years, "man you need a 3930K for these long workloads" scenario) formula is here!

4670\4770K Power consumption: http://imgur.com/3IIf21K
8350 Power consumption :http://imgur.com/j37Dx29
OCd 8350 with H80i\9370 with AMD FX CLC: http://imgur.com/GhpPByc

4670K (On sale) Final Bill: http://imgur.com/F9m72qW
4770K Final Bill: http://imgur.com/GC4rqOY
4670K (On sale) final bill with $80 CLC or comparable air cooler: http://imgur.com/nXlTmJ9
4770K Final Bill with $80 CLC or comparable air cooler: http://imgur.com/JQO951U
8350 Final Bill: http://imgur.com/hPH95kg
OCd 8350 with CLC/9370 With FX AIO final bill: http://imgur.com/Y8AvLUX
9370 final bill if you have an old H70 lying around and sold the FX AIO CLC : http://imgur.com/BCGtQny

As we can see, the only time Intel truly wins in this extreme case scenario is with the 4670K (in this case, barely beating out the 8350 in final cost by a whopping $7, should the 4670K have been priced @$220 and not $240...

/Intelneverrealisticallypaysitselfoff

Just to throw it out there, what about the 3930k? Honestly imo for price to performance its the best Intel processor (not including the Xeons). Sure it's $570 (4930k @ $580), however, it's literally a Extreme series under clocked. For the same price of an extreme series, you could get that, a motherboard for it, and a h100i (or hyper evo 212 depending on what type of person you are), for about the same price, then overclock it to 4.2, of even 4.5 with really no problems. ^_^

I was referring to the full load for 4 hours scenario I calculated, if you are that workload-intensive, you might wanna invest in a 3930K The 3930K is the GTX 780 of CPUs, not the best value, but certainly not the worst, just the victim of the price creep syndrome ...

 

[GOM3RPLY3R]

@juanrga, don't hype up ARM, you'll be disappointed very soon my friend, don't fall for the marketing slides and think realistically...

Now, to burn some time and to put the nail in the coffin, the end all be all power consumption (Full load, 4 hours a day, 3 years, "man you need a 3930K for these long workloads" scenario) formula is here!

4670\4770K Power consumption: http://imgur.com/3IIf21K
8350 Power consumption :http://imgur.com/j37Dx29
OCd 8350 with H80i\9370 with AMD FX CLC: http://imgur.com/GhpPByc

4670K (On sale) Final Bill: http://imgur.com/F9m72qW
4770K Final Bill: http://imgur.com/GC4rqOY
4670K (On sale) final bill with $80 CLC or comparable air cooler: http://imgur.com/nXlTmJ9
4770K Final Bill with $80 CLC or comparable air cooler: http://imgur.com/JQO951U
8350 Final Bill: http://imgur.com/hPH95kg
OCd 8350 with CLC/9370 With FX AIO final bill: http://imgur.com/Y8AvLUX
9370 final bill if you have an old H70 lying around and sold the FX AIO CLC : http://imgur.com/BCGtQny

As we can see, the only time Intel truly wins in this extreme case scenario is with the 4670K (in this case, barely beating out the 8350 in final cost by a whopping $7, should the 4670K have been priced @$220 and not $240...

/Intelneverrealisticallypaysitselfoff

Just to throw it out there, what about the 3930k? Honestly imo for price to performance its the best Intel processor (not including the Xeons). Sure it's $570 (4930k @ $580), however, it's literally a Extreme series under clocked. For the same price of an extreme series, you could get that, a motherboard for it, and a h100i (or hyper evo 212 depending on what type of person you are), for about the same price, then overclock it to 4.2, of even 4.5 with really no problems. ^_^

I was referring to the full load for 4 hours scenario I calculated, if you are that workload-intensive, you might wanna invest in a 3930K The 3930K is the GTX 780 of CPUs, not the best value, but certainly not the worst, just the victim of the price creep syndrome ...

Yeah. I want to record and stream 1080p gaming. I just tried it with my 4.5 Ghz 3570k, but i couldn't even do 720p without dropping at least 15 frames on DayZ.

I wanted to get it too. But it's something I'll have to wait until next summer for. However, by Christmas I may get another 770. :3

And I just don't get what Intel is doing. If they dropped their prices by 15-20%, they would be getting so many more sales it's not even funny, while still being more expensive than AMD.

 

[The Q6660 Inside]

@juanrga, don't hype up ARM, you'll be disappointed very soon my friend, don't fall for the marketing slides and think realistically...

Now, to burn some time and to put the nail in the coffin, the end all be all power consumption (Full load, 4 hours a day, 3 years, "man you need a 3930K for these long workloads" scenario) formula is here!

4670\4770K Power consumption: http://imgur.com/3IIf21K
8350 Power consumption :http://imgur.com/j37Dx29
OCd 8350 with H80i\9370 with AMD FX CLC: http://imgur.com/GhpPByc

4670K (On sale) Final Bill: http://imgur.com/F9m72qW
4770K Final Bill: http://imgur.com/GC4rqOY
4670K (On sale) final bill with $80 CLC or comparable air cooler: http://imgur.com/nXlTmJ9
4770K Final Bill with $80 CLC or comparable air cooler: http://imgur.com/JQO951U
8350 Final Bill: http://imgur.com/hPH95kg
OCd 8350 with CLC/9370 With FX AIO final bill: http://imgur.com/Y8AvLUX
9370 final bill if you have an old H70 lying around and sold the FX AIO CLC : http://imgur.com/BCGtQny

As we can see, the only time Intel truly wins in this extreme case scenario is with the 4670K (in this case, barely beating out the 8350 in final cost by a whopping $7, should the 4670K have been priced @$220 and not $240...

/Intelneverrealisticallypaysitselfoff

Just to throw it out there, what about the 3930k? Honestly imo for price to performance its the best Intel processor (not including the Xeons). Sure it's $570 (4930k @ $580), however, it's literally a Extreme series under clocked. For the same price of an extreme series, you could get that, a motherboard for it, and a h100i (or hyper evo 212 depending on what type of person you are), for about the same price, then overclock it to 4.2, of even 4.5 with really no problems. ^_^

I was referring to the full load for 4 hours scenario I calculated, if you are that workload-intensive, you might wanna invest in a 3930K The 3930K is the GTX 780 of CPUs, not the best value, but certainly not the worst, just the victim of the price creep syndrome ...

Yeah. I want to record and stream 1080p gaming. I just tried it with my 4.5 Ghz 3570k, but i couldn't even do 720p without dropping at least 15 frames on DayZ.

I wanted to get it too. But it's something I'll have to wait until next summer for. However, by Christmas I may get another 770. :3

And I just don't get what Intel is doing. If they dropped their prices by 15-20%, they would be getting so many more sales it's not even funny, while still being more expensive than AMD.

Haswell has been price creeping up, but it is all part of the grand scheme, they can overcharge and make a big profit out of it.

Yknow, I wanted to go Dual-Crossfire Eyefinity Edition 5870s back in mid 2010. I bought one, a 1080p monitor, and a secondhand 920 with a brand new Sabertooth X58 in anticipation for this crazy setup. I never was able to get another 5870, or 2 extra 1080p monitors due to monetary problems and I eventually sold off my only 1080p monitor, got a 1680x1050 Dell IPS, and gave my 5870 to a friend for $200, in exchange for his lowly 4870 ...

 

[GOM3RPLY3R]

@juanrga, don't hype up ARM, you'll be disappointed very soon my friend, don't fall for the marketing slides and think realistically...

Now, to burn some time and to put the nail in the coffin, the end all be all power consumption (Full load, 4 hours a day, 3 years, "man you need a 3930K for these long workloads" scenario) formula is here!

4670\4770K Power consumption: http://imgur.com/3IIf21K
8350 Power consumption :http://imgur.com/j37Dx29
OCd 8350 with H80i\9370 with AMD FX CLC: http://imgur.com/GhpPByc

4670K (On sale) Final Bill: http://imgur.com/F9m72qW
4770K Final Bill: http://imgur.com/GC4rqOY
4670K (On sale) final bill with $80 CLC or comparable air cooler: http://imgur.com/nXlTmJ9
4770K Final Bill with $80 CLC or comparable air cooler: http://imgur.com/JQO951U
8350 Final Bill: http://imgur.com/hPH95kg
OCd 8350 with CLC/9370 With FX AIO final bill: http://imgur.com/Y8AvLUX
9370 final bill if you have an old H70 lying around and sold the FX AIO CLC : http://imgur.com/BCGtQny

As we can see, the only time Intel truly wins in this extreme case scenario is with the 4670K (in this case, barely beating out the 8350 in final cost by a whopping $7, should the 4670K have been priced @$220 and not $240...

/Intelneverrealisticallypaysitselfoff

Just to throw it out there, what about the 3930k? Honestly imo for price to performance its the best Intel processor (not including the Xeons). Sure it's $570 (4930k @ $580), however, it's literally a Extreme series under clocked. For the same price of an extreme series, you could get that, a motherboard for it, and a h100i (or hyper evo 212 depending on what type of person you are), for about the same price, then overclock it to 4.2, of even 4.5 with really no problems. ^_^

I was referring to the full load for 4 hours scenario I calculated, if you are that workload-intensive, you might wanna invest in a 3930K The 3930K is the GTX 780 of CPUs, not the best value, but certainly not the worst, just the victim of the price creep syndrome ...

Yeah. I want to record and stream 1080p gaming. I just tried it with my 4.5 Ghz 3570k, but i couldn't even do 720p without dropping at least 15 frames on DayZ.

I wanted to get it too. But it's something I'll have to wait until next summer for. However, by Christmas I may get another 770. :3

And I just don't get what Intel is doing. If they dropped their prices by 15-20%, they would be getting so many more sales it's not even funny, while still being more expensive than AMD.

Haswell has been price creeping up, but it is all part of the grand scheme, they can overcharge and make a big profit out of it.

Yknow, I wanted to go Dual-Crossfire Eyefinity Edition 5870s back in mid 2010, I bought one and a secondhand 920 with a brand new Sabertooth X58 in anticipation for this crazy setup. I never was able to get another 5870, or 2 extra 1080p monitors due to monetary problems and I eventually sold off my only 1080p monitor, got a 1680x1050 Dell IPS, and gave my 5870 to a friend for $200, in exchange for his lowly 4870 ...

Lol, if you had it going, why didn't you just keep it? My friend gets loads of stuff off craigslist and stuff and hes only 15! He got a $2500 worth ASUS Gaming laptop with a GTX 680 M, i7-3630QM, and everything (sadly) after he overclocked his i7 to 4.2 Ghz, stays cooler than my current setup. Which by the way, is defeating logic.

But yeah, hes getting a server computer, and all this other crap. He probably has $8000 worth of computer stuff that he got on his own. Did I mention to get the ASUS laptop, he traded a crappy Macbook that was only worth $900? It's all in where you look!

As a side note: Me and him had a little talk and I thought it was interesting. If someone made a modified version of windows, or just a OS that could run everything in 256-bit (or 384-bit), we could use our GPU's to power everything, which would be stronger than Swifty's overclocked Xeon E5's combined! :3 I thought it was interesting.

 

[The Q6660 Inside]

@juanrga, don't hype up ARM, you'll be disappointed very soon my friend, don't fall for the marketing slides and think realistically...

Now, to burn some time and to put the nail in the coffin, the end all be all power consumption (Full load, 4 hours a day, 3 years, "man you need a 3930K for these long workloads" scenario) formula is here!

4670\4770K Power consumption: http://imgur.com/3IIf21K
8350 Power consumption :http://imgur.com/j37Dx29
OCd 8350 with H80i\9370 with AMD FX CLC: http://imgur.com/GhpPByc

4670K (On sale) Final Bill: http://imgur.com/F9m72qW
4770K Final Bill: http://imgur.com/GC4rqOY
4670K (On sale) final bill with $80 CLC or comparable air cooler: http://imgur.com/nXlTmJ9
4770K Final Bill with $80 CLC or comparable air cooler: http://imgur.com/JQO951U
8350 Final Bill: http://imgur.com/hPH95kg
OCd 8350 with CLC/9370 With FX AIO final bill: http://imgur.com/Y8AvLUX
9370 final bill if you have an old H70 lying around and sold the FX AIO CLC : http://imgur.com/BCGtQny

As we can see, the only time Intel truly wins in this extreme case scenario is with the 4670K (in this case, barely beating out the 8350 in final cost by a whopping $7, should the 4670K have been priced @$220 and not $240...

/Intelneverrealisticallypaysitselfoff

Just to throw it out there, what about the 3930k? Honestly imo for price to performance its the best Intel processor (not including the Xeons). Sure it's $570 (4930k @ $580), however, it's literally a Extreme series under clocked. For the same price of an extreme series, you could get that, a motherboard for it, and a h100i (or hyper evo 212 depending on what type of person you are), for about the same price, then overclock it to 4.2, of even 4.5 with really no problems. ^_^

I was referring to the full load for 4 hours scenario I calculated, if you are that workload-intensive, you might wanna invest in a 3930K The 3930K is the GTX 780 of CPUs, not the best value, but certainly not the worst, just the victim of the price creep syndrome ...

Yeah. I want to record and stream 1080p gaming. I just tried it with my 4.5 Ghz 3570k, but i couldn't even do 720p without dropping at least 15 frames on DayZ.

I wanted to get it too. But it's something I'll have to wait until next summer for. However, by Christmas I may get another 770. :3

And I just don't get what Intel is doing. If they dropped their prices by 15-20%, they would be getting so many more sales it's not even funny, while still being more expensive than AMD.

Haswell has been price creeping up, but it is all part of the grand scheme, they can overcharge and make a big profit out of it.

Yknow, I wanted to go Dual-Crossfire Eyefinity Edition 5870s back in mid 2010, I bought one and a secondhand 920 with a brand new Sabertooth X58 in anticipation for this crazy setup. I never was able to get another 5870, or 2 extra 1080p monitors due to monetary problems and I eventually sold off my only 1080p monitor, got a 1680x1050 Dell IPS, and gave my 5870 to a friend for $200, in exchange for his lowly 4870 ...

Lol, if you had it going, why didn't you just keep it? My friend gets loads of stuff off craigslist and stuff and hes only 15! He got a $2500 worth ASUS Gaming laptop with a GTX 680 M, i7-3630QM, and everything (sadly) after he overclocked his i7 to 4.2 Ghz, stays cooler than my current setup. Which by the way, is defeating logic.

But yeah, hes getting a server computer, and all this other crap. He probably has $8000 worth of computer stuff that he got on his own. Did I mention to get the ASUS laptop, he traded a crappy Macbook that was only worth $900? It's all in where you look!

I kept the 920 and the Sabertooth X58, but the rest just had to go due to financial reasons, getting rid of both the 5870 and 1080p monitor was a quick ~$400.

 

[jdwii]

Man just to think it took the industry 10 years to start making big AAA titles demand a 64bit processor

 

[8350rocks]

@juanrga, don't hype up ARM, you'll be disappointed very soon my friend, don't fall for the marketing slides and think realistically...

Now, to burn some time and to put the nail in the coffin, the end all be all power consumption (Full load, 4 hours a day, 3 years, "man you need a 3930K for these long workloads" scenario) formula is here!

4670\4770K Power consumption: http://imgur.com/3IIf21K
8350 Power consumption :http://imgur.com/j37Dx29
OCd 8350 with H80i\9370 with AMD FX CLC: http://imgur.com/GhpPByc

4670K (On sale) Final Bill: http://imgur.com/F9m72qW
4770K Final Bill: http://imgur.com/GC4rqOY
4670K (On sale) final bill with $80 CLC or comparable air cooler: http://imgur.com/nXlTmJ9
4770K Final Bill with $80 CLC or comparable air cooler: http://imgur.com/JQO951U
8350 Final Bill: http://imgur.com/hPH95kg
OCd 8350 with CLC/9370 With FX AIO final bill: http://imgur.com/Y8AvLUX
9370 final bill if you have an old H70 lying around and sold the FX AIO CLC : http://imgur.com/BCGtQny

As we can see, the only time Intel truly wins in this extreme case scenario is with the 4670K (in this case, barely beating out the 8350 in final cost by a whopping $7, should the 4670K have been priced @$220 and not $240...

/Intelneverrealisticallypaysitselfoff

Just to throw it out there, what about the 3930k? Honestly imo for price to performance its the best Intel processor (not including the Xeons). Sure it's $570 (4930k @ $580), however, it's literally a Extreme series under clocked. For the same price of an extreme series, you could get that, a motherboard for it, and a h100i (or hyper evo 212 depending on what type of person you are), for about the same price, then overclock it to 4.2, of even 4.5 with really no problems. ^_^

I was referring to the full load for 4 hours scenario I calculated, if you are that workload-intensive, you might wanna invest in a 3930K The 3930K is the GTX 780 of CPUs, not the best value, but certainly not the worst, just the victim of the price creep syndrome ...

Yeah. I want to record and stream 1080p gaming. I just tried it with my 4.5 Ghz 3570k, but i couldn't even do 720p without dropping at least 15 frames on DayZ.

I wanted to get it too. But it's something I'll have to wait until next summer for. However, by Christmas I may get another 770. :3

And I just don't get what Intel is doing. If they dropped their prices by 15-20%, they would be getting so many more sales it's not even funny, while still being more expensive than AMD.

Haswell has been price creeping up, but it is all part of the grand scheme, they can overcharge and make a big profit out of it.

Yknow, I wanted to go Dual-Crossfire Eyefinity Edition 5870s back in mid 2010, I bought one and a secondhand 920 with a brand new Sabertooth X58 in anticipation for this crazy setup. I never was able to get another 5870, or 2 extra 1080p monitors due to monetary problems and I eventually sold off my only 1080p monitor, got a 1680x1050 Dell IPS, and gave my 5870 to a friend for $200, in exchange for his lowly 4870 ...

Lol, if you had it going, why didn't you just keep it? My friend gets loads of stuff off craigslist and stuff and hes only 15! He got a $2500 worth ASUS Gaming laptop with a GTX 680 M, i7-3630QM, and everything (sadly) after he overclocked his i7 to 4.2 Ghz, stays cooler than my current setup. Which by the way, is defeating logic.

But yeah, hes getting a server computer, and all this other crap. He probably has $8000 worth of computer stuff that he got on his own. Did I mention to get the ASUS laptop, he traded a crappy Macbook that was only worth $900? It's all in where you look!

As a side note: Me and him had a little talk and I thought it was interesting. If someone made a modified version of windows, or just a OS that could run everything in 256-bit (or 384-bit), we could use our GPU's to power everything, which would be stronger than Swifty's overclocked Xeon E5's combined! :3 I thought it was interesting.

Almost, the issue with 256 bit is that serial instructions would still be massively slow for GPUs to compute. That's why we don't run our OS on the GPU...because the CPU is far better at most of the things the OS needs to calculate.

Now when it comes to things like fluid dynamics, high level physics, and protein synthesis that can be done in massively parallel formats quite easily...the GPU wins hands down.

 

[palladin9479]

Man just to think it took the industry 10 years to start making big AAA titles demand a 64bit processor

Dude I'm so happy they finally did it. 32-bit NT has entirely too many restrictions, we've needed a true 64-bit cut over for awhile now.

Almost, the issue with 256 bit is that serial instructions would still be massively slow for GPUs to compute. That's why we don't run our OS on the GPU...because the CPU is far better at most of the things the OS needs to calculate.

Now when it comes to things like fluid dynamics, high level physics, and protein synthesis that can be done in massively parallel formats quite easily...the GPU wins hands down.

That's scalar vs vector processing for you. The vast majority of code produced will be scalar but what little an be vectorized will benefit massively from doing so. We're talking 300~400% for the low hanging fruit and 1000~5000%+ for the truly insane.

 

[griptwister]

I saw a 3930K on ebay the other day, and I almost won the battle on it. It was bought for only $60. I'm quite sad I wasn't there to make the final bid ((

On another note, has anyone seen the specs for CoD Dogs??? Er... I mean Call of Duty Ghosts? They're friggin heavy! 50Gb of HDD space is required! 3-4 core CPU and a mid-range GPU for minimum requirements! I smell HD+ res textures!!!

@The Q6660 Inside, 1080P is over rated anyways. Just get a monitor with a high pixel density or go 1440P I say. Also, those 2560x1080 monitors look like a fun resolution to play at.

 

[jdwii]

I saw a 3930K on ebay the other day, and I almost won the battle on it. It was bought for only $60. I'm quite sad I wasn't there to make the final bid ((

On another note, has anyone seen the specs for CoD Dogs??? Er... I mean Call of Duty Ghosts? They're friggin heavy! 50Gb of HDD space is required! 3-4 core CPU and a mid-range GPU for minimum requirements! I smell HD+ res textures!!!

@The Q6660 Inside, 1080P is over rated anyways. Just get a monitor with a high pixel density or go 1440P I say. Also, those 2560x1080 monitors look like a fun resolution to play at.

Well looking at the graphics i would of never guessed BF4 looks much better and the graphics don't demand as much and can we say bigger maps as well!

 

[The Q6660 Inside]

I saw a 3930K on ebay the other day, and I almost won the battle on it. It was bought for only $60. I'm quite sad I wasn't there to make the final bid ((

On another note, has anyone seen the specs for CoD Dogs??? Er... I mean Call of Duty Ghosts? They're friggin heavy! 50Gb of HDD space is required! 3-4 core CPU and a mid-range GPU for minimum requirements! I smell HD+ res textures!!!

@The Q6660 Inside, 1080P is over rated anyways. Just get a monitor with a high pixel density or go 1440P I say. Also, those 2560x1080 monitors look like a fun resolution to play at.

I already got mysef a VH238H with the ASUS DCUIIOC GTX 760 and 500R case, I won't be going for any crazy eyefinity or surround setups this time around, they just are not worth it. I will get a 2560x1080 monitor someday, someday indeed... Probably when the R9 290X falls into the $400 range

 

[lithium6]

I can't remember if this has already been noticed, but it seems that FM2+ processors will have max. 95W TDP:

http://www.asrock.com/microsite/FM2+/FM2-en.html

 

[gamerk316]

Man just to think it took the industry 10 years to start making big AAA titles demand a 64bit processor

Remember that a LOT of the work had been pushed to the GPU, and since the GPU exists outside the CPU's Address Space, you could make due with a 2GB Address Space limit.

Allowing a larger Address Space via Win64 simply means you are going to do a LOT less moving of data around, as you can allow the working set to grow huge, rather then having to keep it under 2GB. Nevermind the positive benefits in regards of paging...I don't see Texture sizes growing too much too fast due to the GPU, though you will see them grow over time as GPU power increases.

 

[gamerk316]

I saw a 3930K on ebay the other day, and I almost won the battle on it. It was bought for only $60. I'm quite sad I wasn't there to make the final bid ((

On another note, has anyone seen the specs for CoD Dogs??? Er... I mean Call of Duty Ghosts? They're friggin heavy! 50Gb of HDD space is required! 3-4 core CPU and a mid-range GPU for minimum requirements! I smell HD+ res textures!!!

@The Q6660 Inside, 1080P is over rated anyways. Just get a monitor with a high pixel density or go 1440P I say. Also, those 2560x1080 monitors look like a fun resolution to play at.

OR very unoptimized. Note that only DX11 GPU's are supported, hence the high GPU requirements. Also note the recommended C2D CPU. So MASSIVE GPU bottleneck, which is typical of CoD.

 

[juanrga]

@juanrga, don't hype up ARM, you'll be disappointed very soon my friend, don't fall for the marketing slides and think realistically...

What hype?

Now, to burn some time and to put the nail in the coffin, the end all be all power consumption (Full load, 4 hours a day, 3 years, "man you need a 3930K for these long workloads" scenario) formula is here!

4670\4770K Power consumption: http://imgur.com/3IIf21K
8350 Power consumption :http://imgur.com/j37Dx29
OCd 8350 with H80i\9370 with AMD FX CLC: http://imgur.com/GhpPByc

4670K (On sale) Final Bill: http://imgur.com/F9m72qW
4770K Final Bill: http://imgur.com/GC4rqOY
4670K (On sale) final bill with $80 CLC or comparable air cooler: http://imgur.com/nXlTmJ9
4770K Final Bill with $80 CLC or comparable air cooler: http://imgur.com/JQO951U
8350 Final Bill: http://imgur.com/hPH95kg
OCd 8350 with CLC/9370 With FX AIO final bill: http://imgur.com/Y8AvLUX
9370 final bill if you have an old H70 lying around and sold the FX AIO CLC : http://imgur.com/BCGtQny

As we can see, the only time Intel truly wins in this extreme case scenario is with the 4670K (in this case, barely beating out the 8350 in final cost by a whopping $7, should the 4670K have been priced @$220 and not $240...

/Intelneverrealisticallypaysitselfoff

Efficiency is fundamental for phones, tablets, servers and supercomputers, but it is not for traditional desktops.

 

[juanrga]

As a side note: Me and him had a little talk and I thought it was interesting. If someone made a modified version of windows, or just a OS that could run everything in 256-bit (or 384-bit), we could use our GPU's to power everything, which would be stronger than Swifty's overclocked Xeon E5's combined! :3 I thought it was interesting.

GPUs are optimized for throughput. CPUs are optimized for latency. Using a GPU for running CPU tasks is very inefficient.

Heterogeneous computing joins both worlds and provides a system that is strong at both latency and throughput.

 

[The Q6660 Inside]

@juanrga, don't hype up ARM, you'll be disappointed very soon my friend, don't fall for the marketing slides and think realistically...

What hype?

Now, to burn some time and to put the nail in the coffin, the end all be all power consumption (Full load, 4 hours a day, 3 years, "man you need a 3930K for these long workloads" scenario) formula is here!

4670\4770K Power consumption: http://imgur.com/3IIf21K
8350 Power consumption :http://imgur.com/j37Dx29
OCd 8350 with H80i\9370 with AMD FX CLC: http://imgur.com/GhpPByc

4670K (On sale) Final Bill: http://imgur.com/F9m72qW
4770K Final Bill: http://imgur.com/GC4rqOY
4670K (On sale) final bill with $80 CLC or comparable air cooler: http://imgur.com/nXlTmJ9
4770K Final Bill with $80 CLC or comparable air cooler: http://imgur.com/JQO951U
8350 Final Bill: http://imgur.com/hPH95kg
OCd 8350 with CLC/9370 With FX AIO final bill: http://imgur.com/Y8AvLUX
9370 final bill if you have an old H70 lying around and sold the FX AIO CLC : http://imgur.com/BCGtQny

As we can see, the only time Intel truly wins in this extreme case scenario is with the 4670K (in this case, barely beating out the 8350 in final cost by a whopping $7, should the 4670K have been priced @$220 and not $240...

/Intelneverrealisticallypaysitselfoff

Efficiency is fundamental for phones, tablets, servers and supercomputers, but it is not for traditional desktops.

I was there to put the nail in the coffin of the "ERMEHGAWD VISHERA WILL LIEK MAYK YOAR POWHURR BILLS 9001 TIAMS MOAR SPENSIVE" crap, desktop power consumption is pointless. Don't assume major players are going to be replacing x86 with ARM anytime soon..

 

[esrever]

I saw a 3930K on ebay the other day, and I almost won the battle on it. It was bought for only $60. I'm quite sad I wasn't there to make the final bid ((

On another note, has anyone seen the specs for CoD Dogs??? Er... I mean Call of Duty Ghosts? They're friggin heavy! 50Gb of HDD space is required! 3-4 core CPU and a mid-range GPU for minimum requirements! I smell HD+ res textures!!!

@The Q6660 Inside, 1080P is over rated anyways. Just get a monitor with a high pixel density or go 1440P I say. Also, those 2560x1080 monitors look like a fun resolution to play at.

It will either be a really shitty PC port like all COD or they are just bullshitting so they don't look so bad next to bf4.

 

[8350rocks]

I saw a 3930K on ebay the other day, and I almost won the battle on it. It was bought for only $60. I'm quite sad I wasn't there to make the final bid ((

On another note, has anyone seen the specs for CoD Dogs??? Er... I mean Call of Duty Ghosts? They're friggin heavy! 50Gb of HDD space is required! 3-4 core CPU and a mid-range GPU for minimum requirements! I smell HD+ res textures!!!

@The Q6660 Inside, 1080P is over rated anyways. Just get a monitor with a high pixel density or go 1440P I say. Also, those 2560x1080 monitors look like a fun resolution to play at.

It will either be a really shitty PC port like all COD or they are just bullshitting so they don't look so bad next to bf4.

Emphasis mine...

 

[Avro Arrow]

The FX 9370 seems pretty good now, you know, with an Asetek LCS and pre-ocd to 4.7Ghz and Far Cry 3, it can hold quite the grudge against a 4770K with a sh*t stock cooler...
http://www.newegg.com/Product/Product.aspx?Item=N82E16819113352

Meanwhile.., AMD is secretly killing the tin FX 83XX boxes

I know eh? I was like "Hey, why is my FX-8350 in a cardboard box?" and I was told "The tins are gone". It's ok, it was just a gimmick to begin with (albeit a really nice one...lol).

On another note, how did the "Steamroller" thread get turned into "ARM vs x86"???

 

[Cazalan]

I wrote 'sponsors' not sponsors.

As mentioned before, there are three (3) Intel persons in his research group (one collaborator plus two students), not only one (1).

You're really grasping at straws here. The paper was written by university students, not Vertical Research group. They just credit Vertical Research group for helping review the paper.

No sure why you mention Broadcom and Cisco. Do they have a historical record of dishonest competition/cheating? Also, do you believe that because there is people from other companies the possibility of aggressive 'sponsoring' is eliminated? AMD, Nvidia, and VIA were members of SYSMARK consortium and still Intel cheated the SYSMARK benchmarks for artificially increasing the score of its products.

Your reason for assuming the research group is biased is because there are some Intel people in it. Well there's also a Broadcom and Google person, and the project was actually funded by Cisco. So there are plenty of ARM companies represented in the research group.

And two of those Intel people are listed as Alumni, which are likely no longer involved at all anymore.

If that article that you mentioned was right and Intel (x86) is able to compete with ARM face to face, then Intel had not been caught using its ICC compiler to cheat the Antutu benchmarks this year.

The cheated benchmark gave ARM processors a mere 50% upgrade in performance whereas Intel chips were seeing increases of up to 292%.

Now you're just going off on an ICC tangent, when the paper used GCC for compiling. Is GCC now biased for Intel too?

Pay attention. The ICC biased CPU dispatcher gives Intel chips about a 15% advantage over the AMD chips. When fighting ARM, Intel had to give its chips up to ~10x more advantage.

http://www.androidauthority.com/analyst-says-intel-beating-arm-in-benchmarks-was-a-sham-243526/

http://www.eetimes.com/author.asp?section_id=36&doc_id=1318894

http://www.slashgear.com/intel-atom-z2580-antutu-benchmark-falls-20-following-revision-13290307/

http://seekingalpha.com/article/1550372-intel-and-antutu-mea-culpa

Yes I'm fully aware of the benchmark shenanigans.

Perhaps you missed how every Android maker also cheats on benchmarks? It's not an Intel only thing.

http://www.anandtech.com/show/7384/state-of-cheating-in-android-benchmarks

I have submitted conference and contest papers a pair of months in advance. However, there are other objections to your hypothesis. The first, that they cite references published so late as 2013. The second, that they mention the A15 and even give some numbers for it. The third, that they claim that they evaluated but rejected more modern SoCs such as Tegra 3 before starting the lab work.

I'm sure they had their reasons. Cost, availability, who knows. You're welcome to do some real research yourself rather than making spreadsheets.

The fact is they used parts that were comparable for the time frame. If they used an Ivy and just an old A8 you'd have a point, but they didn't. They used A8/A9, an old atom and an old i7 (Sandy).

Now explain us: how did they evaluate a 2012 product if their lab work was ready in early 2011 according to you?

The facts are:
- They rejected better ARM designs with improved performance and efficiency.

No. As I already said they were not readily available and on the market. The only A15 even out in that time frame was from Apple. Comparing Apple benchmarks to GCC benchmarks on the other platforms would have been apples & oranges at best.

They reported some released figures for A15 from other companies just as a point of reference.

- They didn't test Bulldozer FX or similar ones but the best possible x86 chips.

Why would they bother? It is Intel's ISA after all. They also didn't use VIA chips which have been really low power for a while now.

- The scaling methodology and power measurement favoured Intel.

That may be the only valid point you have here. The rest is conspiracy theory.

- Their compiler choices favoured Intel.

How does GCC favor Intel?

- «they believe the results would look much the same — in terms of the relationship between power and performance» when looking at newer ARM designs, including the 64-bit ARMv8 architecture.

The only reference to ARMv8 is that the 64-bit parts are still under development. You're grasping at straws here.

- They have 3 Intel people in their small research group, but nobody from ARM or AMD.

All together implies bias pro Intel.

I'll say it again. The paper was written by 3 students. ONE of which happens to be part of that Research group. It was not written or funded by that research group.

They only credit Veritcal research group for helping review the paper, among many others.

"We thank the anonymous reviewers, the Vertical group, and the PARSA group for comments. "

You also failed to see that one of the members works for Google, and another works for Broadcom. And the project was actually funded by Cisco (among others).

Google Android is what 99% powered by ARM cores? Broadcom makes ARM core parts. So exactly how would these astute people let Intel bias slide without question?

Take the old Titan, double it, and you recover the #1 (approx.). But as explained to you before, this pure brute approach doesn't work for exascale supercomputing. You cannot take a x86-based supercomputer and scale it 1000x times because you cannot generate the immense power needed for that beast. That is why HPC experts are developing an ARM supercomputer for exascale.

You keep saying this 1000x times number but you're overlooking that there is already a 50 PetaFLOP (Peak) supercomputer. That means to get to an ExaFLOP you only need to get 20x faster than what is out there today, not 1000x.

Suddenly that number doesn't look so big anymore.

Also you seem to not understand that the performance of the Xeon Phi has very little to do with x86, but is directly related to an _extension_ of the x86 ISA which is exclusive for Phi. Follow my advice, don't be fooled by Intel marketing ;-)

There's nothing to be fooled by. It's a parallel compute card. It's done differently than NVidia/AMD compute/graphics cards but that is how they chose to implement it. Intel is vertically integrated so they deemed that the best way to go.

And they're being aggressive with Phi by putting it on their 14nm node next year and leaving Haswell desktop at a 22nm refresh.

I doubt any supercomputer will be powered by Tegra 6. Who said you that?

You're telling me after all this time talking about Nvidia HPC and ARM supercomputers you suddenly forgot about Montblanc that was planning to use future Tegra mobile chips?

I see now they've decided to go with Samsung parts.

Read the link that you give about CUDA on ARM.

How many times the word "server" appears? Zero (0). How many times the word "supercomputer" and derivative (supercomputers, supercomputing, HPC) appears? Seven (7).

Now you're just splitting hairs. I'm sure they will sell smaller nodes that they'll classify as a "server".