AMD CPU speculation... and expert conjecture

[juanrga]

4. Yes the hardware is better than last generation, however, the architecture with no doubt will be lowered. My Core 2 Quad Q8300 and Radeon 5570 are about the same performance of these consoles.

Not even close. It is more like

PS4 ~ i7 + GTX 680/780 Windows PC

 

[juanrga]

@griptwister. The ~2x factor is not only from lightweight OS but from coding for a single spec. as well.

 

[noob2222]

Intel are ahead of any company on nm for cpu design. I think intel are a bargain they are selling amazing stuff, which will take amd another 4-5 years to get to haswell level and you get an igpu cutting edge technology.

Nope. Excavator is coming in 2015.

@juan

AGAIN LOL,

In multithreaded integer workloads, the Z3770 gets dangerously close to Ivy Bridge levels of performance. Again, we're overstating Bay Trail's performance here as the Z3770 has four cores while the Core i5-3317U only has two (but with Hyper Threading presenting another 2 virtual cores). I don't believe most tablet workloads are heavily threaded integer workloads, however the world is hardly single threaded anymore. The reality is that a quad-core Bay Trail should perform somewhere between 40% - 80% of a dual-core Ivy Bridge

wow, thats impressive you only took the first sentence and tried to defend yourself. 4 cores to equal 40-80% of a dual core Ivy bridge .... you fail miserably.

Thats funny, looking at that same test, AMD >>> Ivy bridge

you can continue to anxiously wait for your mediocre performance - non compliant products all you want.

I can also see you bought intel's new SDP gimmick for cpu power ... 2 watts SDP != 2W TDP

LOL. I avoided the part I was sure would confuse you... and your reply confirms that my suspicion was right.

What you continue without understanding is that they are testing a phone-level chip. Aka the chip is thermally constrained to fit inside a phone. That phone-level chip is offering a 40-80% of the performance of an i5/i7 laptop chip, but is consuming only 1W-2.5W.

Nobody mentioned SDPs or TDPs. You are again making up things in your brain. The phone chip was consuming a maximum of 2.5W during those tests (and ~700mW at iddle). I can sure you that the Ivy Bridge chips used in that comparison consume much much much more than 2.5W. The conclusion is evident the new Silvermont architecture is much much more efficient than Ivy Bridge.

If the phone-level chip was scaled up to something as 8W, it would be faster than the Ivy Bridge chips in raw performance, but guess what? Then it couldn't be used inside a phone.

The funny part is that the new Silvermont architecture by Intel has retained the crowd only during one week! Apple has just released its new phone chip and has humiliated Intel

http://www.anandtech.com/show/7335/the-iphone-5s-review/6

A dual-core ARM64 @ 1.3Ghz is offering about the same raw performance than a (Kabini) quad-core jaguar @ 1.5GHz. And don't miss that the Apple chip is thermally constrained to phone-level (single digit power consumption).

lets just check that shall we?

http://www.tomshardware.com/reviews/bay-trail-celeron-j1750-performance,3614-3.html

wow, what do you know ... a 10W bay trail ...

ya .. thats definately on par with Ivy Bridge ... 2.4ghz dual core to 2.6 dual core .. IVY is over 2x as fast. Must be that 200 mhz speed advantage. I bet If you clock that bay trail to 2.6 ghz it will wipe IVY bridge off the map.

LOL That was funny as well. You pretended to compare a ~2W chip to a ~20W chip and when I quadrupled the former chip to ~8W for comparing it with the ~20W in the same benchmark, showing you that the new architecture is more efficient, you reaction is... comparing now a 10W to a 55W chip! Funny, really funny. Let me guess If I now scale up the chip above the 10W you will change the other chip again. Will you chose a 130W maybe?

I wonder why you are ignoring the conclusions of the same article that you are using to obtain your figures. Take a look to the section "Bail Trail is faster and much more efficient":

When Intel introduced us to its Silvermont architecture, the company made grand claims of increased performance at a given thermal ceiling, or similar performance at reduced power.

[...]

The company wasn't exaggerating when it suggested that the Silvermont architecture could as much as double performance at a given power limit. Snappy little passively-cooled platforms are almost certainly on their way toward the end of the year. Intel tells us that its partners are already working on fully integrated desktops and all-in-one designs, as well as channel-oriented motherboards with soldered-down CPUs.

hey, don't blame me, your the one who said "RAW PERFORMANCE" not a hafijur claim of OMG PPW PPW.

but lets examine that other aspect since Toms didn't test it. 3317U, 1.7ghz base, 2.6ghz turbo dual core with HT, 17W

G1610 2.6ghz dual core no turbo, no HT, 55W

Obviously the single thread tests are going to go to the 3317U since its turbo = g1610 clock.

so lets figure some : minimum speed for the 3317 =65% of the g1610, HT is going to add in 20-30%, we will be conservative and say 20%

17W to 55W is 30% of the power draw as well.

so rough guesstimate of the 3317U would be at worst 6:10 without HT, 4:53 with 20% ht.

Ivy 1.7ghz 3317U shuld still be faster in Raw performance than the 10W 2.4 ghz bay trail, even without taking in HT boosts.

Now figure in the 3317U only rated at 1/3 the power of the G1610 ... and considering the G1610 was only 16% less efficient as tested ... 3317U would probably smoke the bay trail in PPW. Makes one wonder why it wasn't used in the review ...

[...] = But we still see that the entry-level J1750 finishes our suite of tests significantly faster than Atom D2700.

so the new atom beats the old atom ... wow ... skipped right over that comment didn't ya.

anyway keep trying to defend yourself by attempting to change what you said, its quite comical.

 

[juanrga]

Just because it says, "OH BOY THIS HAS A 7850 AND A 8 CORE CPU IN IT!!" A 7850 will totally fit in a tiny Console case. These "7850's" will definitely be less than a low profile 7850. Also take into consideration Power Consumption and Heat. The performance will literally drop, okay not a 5570, but something like a 6650 or 6750 should do the trick, which is no where close to "A average gaming PC."

Nope. It is not "this" CPU and "that" GPU. The PS4 has a custom HSA APU with hUMA. Custom means that it is improved hardware available only for it. E.g. the iGPU has about 4x the compute queues of a 7990 (if my memory doesn't fail), there is an extra bus for accessing data...

Even if a PC had exactly that custom hardware, the PS4 would be still faster:

http://www.eurogamer.net/articles/digitalfoundry-inside-metro-last-light

Oles Shishkovstov: No, you just cannot compare consoles to PC directly. Consoles could do at least 2x what a comparable PC can due to the fixed platform and low-level access to hardware.

This is not very different from car vs motorbike comparison. A 100HP motorbike is much faster than a 100HP car.

4. Yes the hardware is better than last generation, however, the architecture with no doubt will be lowered. My Core 2 Quad Q8300 and Radeon 5570 are about the same performance of these consoles.

Not even close. It is more like

PS4 ~ i7 + GTX 680/780 Windows PC

Nope an i3 3220 is faster and its like an inbetween 7850 7870 gpu performance level but supposedly on 8000 series architecture. The only good thing about the consoles is hyperthreaded dual cores are here to stay for hardcore gaming.

You have no idea of what you are talking, but well... I am not saying anything new.

 

[de5_Roy]

I had an i5 3317u in my previous laptop it turbo'd to 2.4ghz with 2 cores and total power consumption with screen on was 26-28w with 10-12w idle and it was 2.2x faster then my previous 2ghz core 2 duo. baytrail is amazing performance per watt as they dobled previous atom performance and improved gpu performance by 7x while taking around 75% less electricity to complete the tomshardware benchmark suite. It was at 2.6ghz for single thread stuff only which was rare.

*facepalm* not this 'ohh gawd my intel soooo gooood!!' again...
before you say bay trail is better than previous atom you have to take into account how bad the previous atoms were. since those set the bar very low and intel didn't tune it until now, bay trail looks comparatively better. but it's not really absolutely better. it's the same as saying steamroller/vishera is better than zambezi without realizing where zambezi really set the standard.

edit2: if you're gonna compare, do it after a53/a57 and kabini/temash devices come out. baytrail will have a really, really hard time.

 

[juanrga]

Intel are ahead of any company on nm for cpu design. I think intel are a bargain they are selling amazing stuff, which will take amd another 4-5 years to get to haswell level and you get an igpu cutting edge technology.

Nope. Excavator is coming in 2015.

@juan

AGAIN LOL,

In multithreaded integer workloads, the Z3770 gets dangerously close to Ivy Bridge levels of performance. Again, we're overstating Bay Trail's performance here as the Z3770 has four cores while the Core i5-3317U only has two (but with Hyper Threading presenting another 2 virtual cores). I don't believe most tablet workloads are heavily threaded integer workloads, however the world is hardly single threaded anymore. The reality is that a quad-core Bay Trail should perform somewhere between 40% - 80% of a dual-core Ivy Bridge

wow, thats impressive you only took the first sentence and tried to defend yourself. 4 cores to equal 40-80% of a dual core Ivy bridge .... you fail miserably.

Thats funny, looking at that same test, AMD >>> Ivy bridge

you can continue to anxiously wait for your mediocre performance - non compliant products all you want.

I can also see you bought intel's new SDP gimmick for cpu power ... 2 watts SDP != 2W TDP

LOL. I avoided the part I was sure would confuse you... and your reply confirms that my suspicion was right.

What you continue without understanding is that they are testing a phone-level chip. Aka the chip is thermally constrained to fit inside a phone. That phone-level chip is offering a 40-80% of the performance of an i5/i7 laptop chip, but is consuming only 1W-2.5W.

Nobody mentioned SDPs or TDPs. You are again making up things in your brain. The phone chip was consuming a maximum of 2.5W during those tests (and ~700mW at iddle). I can sure you that the Ivy Bridge chips used in that comparison consume much much much more than 2.5W. The conclusion is evident the new Silvermont architecture is much much more efficient than Ivy Bridge.

If the phone-level chip was scaled up to something as 8W, it would be faster than the Ivy Bridge chips in raw performance, but guess what? Then it couldn't be used inside a phone.

The funny part is that the new Silvermont architecture by Intel has retained the crowd only during one week! Apple has just released its new phone chip and has humiliated Intel

http://www.anandtech.com/show/7335/the-iphone-5s-review/6

A dual-core ARM64 @ 1.3Ghz is offering about the same raw performance than a (Kabini) quad-core jaguar @ 1.5GHz. And don't miss that the Apple chip is thermally constrained to phone-level (single digit power consumption).

lets just check that shall we?

http://www.tomshardware.com/reviews/bay-trail-celeron-j1750-performance,3614-3.html

wow, what do you know ... a 10W bay trail ...

ya .. thats definately on par with Ivy Bridge ... 2.4ghz dual core to 2.6 dual core .. IVY is over 2x as fast. Must be that 200 mhz speed advantage. I bet If you clock that bay trail to 2.6 ghz it will wipe IVY bridge off the map.

LOL That was funny as well. You pretended to compare a ~2W chip to a ~20W chip and when I quadrupled the former chip to ~8W for comparing it with the ~20W in the same benchmark, showing you that the new architecture is more efficient, you reaction is... comparing now a 10W to a 55W chip! Funny, really funny. Let me guess If I now scale up the chip above the 10W you will change the other chip again. Will you chose a 130W maybe?

I wonder why you are ignoring the conclusions of the same article that you are using to obtain your figures. Take a look to the section "Bail Trail is faster and much more efficient":

When Intel introduced us to its Silvermont architecture, the company made grand claims of increased performance at a given thermal ceiling, or similar performance at reduced power.

[...]

The company wasn't exaggerating when it suggested that the Silvermont architecture could as much as double performance at a given power limit. Snappy little passively-cooled platforms are almost certainly on their way toward the end of the year. Intel tells us that its partners are already working on fully integrated desktops and all-in-one designs, as well as channel-oriented motherboards with soldered-down CPUs.

hey, don't blame me, your the one who said "RAW PERFORMANCE" not a hafijur claim of OMG PPW PPW.

<snip hapidupi-style nonsense>

Raw performance when comparing the like with the like. Not comparing a phone level chip with ~10x thermal rated chip as you pretend. No pretending, as you did, that future ARM desktops by Apple will include the ARM chip of the iphone. No pretending, as you did, that AMD Seattle chip is a low performance chip...

Apparently you didn't still pay enough attention to this:

When Intel introduced us to its Silvermont architecture, the company made grand claims of increased performance at a given thermal ceiling, or similar performance at reduced power.

The original point here was that A57-based Seattle will be up to 2x faster than Steamroller-based Berlin. but power consumption will not be 2x. This explains why AMD is not releasing 8-core Steamroller Opterons. As a consequence, there is no 8-core Steamroller FX.

And the chips with custom ARM cores will be _much_ faster. I estimate Nvidia ARM64 cores will be about a 50% faster than the A57 2GHz used on Seattle. Then, I expect Keller to release a superb custom ARM64 core for Seattle successor.

 

[juanrga]

Apple abandoned PowerPC because of *money*... intel just put some bags of money on the table (sweat sweat deals) , and because Intel was not up to upgrade PowerPC at the rate Apple wanted... after all PPC was the only *desktop* chip of IBM ad only for Apple(Cell was different), and so the reason here for denial or feet dragging was also "money".

Apple switch doesn't have to do with ISAs or performance or features.

I said him exactly that, except I add a link to an insider explaining the IBM-Apple negotiations.

ARM is different... the Aarch32(v7) was more CISC than RISC... the *new* Aarch64 is now clearly RISC, and better than x86 in all aspects, except vectors and some small side issues(like HTM), which could be added very easily in next revisions. OTOH x86 is crowded with Instructions that nobody uses and some kind of redundant instructions, is a mess of an ISA.

RISC (reduced instruction set computer) is better than CISC (complex instruction set computer)... it waste less time and power scanning instruction for length, since the instruction length in RISC is fixed, while in CISC it can vary a lot (in x86 this is creepy, can go from 1byte (8bit) to 15 byte(120bit))... vector/scalar 128/256 bit are logic instructions in x86, in ARM are architectural(kind off) but only to 128bit...

RISC facilitates enormously decoding, and in this ARM is very good also, Qualcomm Krait has 4 full decoders for an embedded chip (x86 is stuck with 2 for Jaguar/Atom like because of power constrains), while AMD/Intel DT has in reality 3 per core, because the "complex decoder" is a hog.

Where CISC (x86) could have an advantage is at fetch, a same cache line could have more instructions in it, and memory reordering is also good in x86 with good Load to Store possibilities. This had some advantage in code density specially when DRAM was at a ultra price premium(who needs more than 640k... remember lol), and x86 got a good push when they started to change the *internal instruction format* to fixed length like RISC, matter of fact in AMD was called RISC86 (predecode-decode to RISC for x86).

x86 could have other little advantage compared to some RISC uarchs with strict memory ordering like MIPS, by using speculation methods due to its good load to store orientation and disambiguation possibilities... but this never stuck really with x86 to make a difference, good speculation is just too complicated, and the low power era was starting to boom (speculation and low power don't go well in the same sentence). This is reverted in favor of ARM because its a "relax memory model" (like HSA/HSAIL lol) and is already low power, it can have all (some) of the speculation features that x86 never really had, yet remain low power (watch server evolution).

Yes, CISC had advantage in the past when memory was very expensive. But not today.

Also, I want mention that the complexity of x86 decoders was the true reason why the original bulldozer design used a shared decoder. Implementing a second decoder complicates the module design, but this limitation is being now corrected with Steamroller. It took AMD two generations. LOL

The RISC approach not only provides simpler implementation at the chip level, but also favours compiler optimizations, which improves the efficiency of the hardware.

The least that RISC vs CISC has to do is with the *number of instruction*, though RISC tend to have less (good for decode)... CISC is not *Complete* its *Complex*, matter of fact in x86 is way bloated and akward complex R.I.P. lol

Aaah! Completely right. The first C in CISC is for "complex" not for "complete". I missed this important correction in my reply to him.

 

[juanrga]

http://www.pcper.com/news/General-Tech/Bay-Trail-M-Powered-Intel-NUC-Coming-Q1-2014-140

baytrail not only is much cheaper then jaguar but makes jaguar look like 5 year old tech. baytrail is competing against arm and providing significantly better performance at even lower power then arm. jaguar competitor is haswell 15w ulvs and we all know how much haswell trounces amd jaguar cpus.

No. First, Atom Bailtrail crow lasted only a week, before Apple humiliated Intel with the new Apple A7. The Intel new design is only interesting when compared with ancient 32bit ARM designs.

Second, Bailtrail will be not competing with old jaguar but with new jaguar successor.

Third, the 5 years old is only in your imagination. In fact the Intel design copies ideas from the design of jaguar.

Fourth, I expect Steamroller T series to compete with lower power haswells. Surely Intel will win in CPU raw performance, but AMD will win in GPU raw performance. I would call that a tie.

 

[juanrga]

During last five years Intel is going to beat ARM, still it is humiliated each time it tries. I predict that by 2017 you will here saying us how "Intel is going to..."

"Trounced" is what you believe is happening when a quad-core @1.5Ghz (Turbo 2.4GHz) is a bit faster than a thermal constrained dual-core @1.3Ghz in some few benchmarks? LOL

Why don't try to read what I said about Steamroller T series, instead making up things in you brain again?

 

[GOM3RPLY3R]

Just because it says, "OH BOY THIS HAS A 7850 AND A 8 CORE CPU IN IT!!" A 7850 will totally fit in a tiny Console case. These "7850's" will definitely be less than a low profile 7850. Also take into consideration Power Consumption and Heat. The performance will literally drop, okay not a 5570, but something like a 6650 or 6750 should do the trick, which is no where close to "A average gaming PC."

Nope. It is not "this" CPU and "that" GPU. The PS4 has a custom HSA APU with hUMA. Custom means that it is improved hardware available only for it. E.g. the iGPU has about 4x the compute queues of a 7990 (if my memory doesn't fail), there is an extra bus for accessing data...

Even if a PC had exactly that custom hardware, the PS4 would be still faster:

http://www.eurogamer.net/articles/digitalfoundry-inside-metro-last-light

Oles Shishkovstov: No, you just cannot compare consoles to PC directly. Consoles could do at least 2x what a comparable PC can due to the fixed platform and low-level access to hardware.

This is not very different from car vs motorbike comparison. A 100HP motorbike is much faster than a 100HP car.

4. Yes the hardware is better than last generation, however, the architecture with no doubt will be lowered. My Core 2 Quad Q8300 and Radeon 5570 are about the same performance of these consoles.

Not even close. It is more like

PS4 ~ i7 + GTX 680/780 Windows PC

Nope an i3 3220 is faster and its like an inbetween 7850 7870 gpu performance level but supposedly on 8000 series architecture. The only good thing about the consoles is hyperthreaded dual cores are here to stay for hardcore gaming.

You have no idea of what you are talking, but well... I am not saying anything new.

I don't know what fancy land you are in, but if your logic is correct, then almost every PC gamer would have a PS4 right now. -_-

In a comparison, realistically, its basically a Core 2 Quad and ~ a 6570-6670. Comparing that hardware to last generation, it's a huge jump. Realistically, if the Consoles had THAT much power, first of all, AMD would set a price MUCH higher than anything, and the console itself in turn would cost more than almost any High End PC out there.

This logic is just bad. For consoles, which are premade, and almost non-customize-able, to overpower Rigs that cost $2000 or more, and for such a low price, just does not make sense.

 

[hcl123]

Intel wasn't competing against arm before but each year they are taking less electricity for more performance and intel have now thought why not make easy work of ARM. Do you really think ARM have a chance to deliver as good cpu at same power envelope as intel. ARM are now scared big time as intel are going to take a lot of there marketshare as its cheap and much better performance at similar power.

Where is steamroller t series is it made up?

You should do consulting and advice Apple , Samsung etc...

You know what, at least when everybody would like a good joke they will listen to you (and you get payed, if thats what is not happening already ) lol...

[EDIT: SORRY FOR THE TRIPLICATED POST, if any body can do anything about it (Palladin ?) i appreciate the delete of 2 redundant posts.. i've been having troubles with 505 errors, supposedly its a server side error... everything is working and worked just fine excepy tomshardware.co.uk ...it seems ok now, perhaps someone didn't like what i've wrote lol (attack on tomshardware.co.uk ? ... bad server upgrade ? ... my bad ? ... can't tell ) ]

 

[Cazalan]

Excellent work of selective quoting and whole misunderstanding. You are quoting from slide 15 on "multicore performance". The phrase immediately above says "ARM multicores as efficient as Intel at the same frequency". They are comparing the old phone-level Tegra 2 and Tegra 3 chips to an i7. The Tegra chips are thermally constrained to phone-level power consumption. The i7 is not thermally constrained.

Thank you. I can read through 20+ pages of marketing BS and find the key slides. This group was paid millions of dollars to come up with a ARM cluster. Of course it will have its benefits and trade offs. Their own research showed that because Intel can run faster, it also had the highest performance/watt. That's rather key don't you think?

That's like saying a Porsche sucks if you limit it to 500rpm. Man that engine just can't put out any HP at 500rpm.

Yes they were older Tegra parts but they were compared to older i7 (Sandy) parts as well. By the 2015 time frame which they are going to production, Intel will be 4 generations from there (Ivy, Haswell, Broadwell, Skylake) and be at 10nm.

You can't just say ARM will be faster in the future as if Intel is sitting still. They're putting 10B+ into fabs and R&D a year.

I'm not dissing ARM. I've worked with ARM chips on an almost daily basis for years. Almost every project I work on has an ARM chip in it. They're everywhere. ARM is key to the market to keep Intel on their toes. It's in all of our best interest for the competition to remain fierce.

But the future is Quark! (kidding!)

 

[8350rocks]

Just because it says, "OH BOY THIS HAS A 7850 AND A 8 CORE CPU IN IT!!" A 7850 will totally fit in a tiny Console case. These "7850's" will definitely be less than a low profile 7850. Also take into consideration Power Consumption and Heat. The performance will literally drop, okay not a 5570, but something like a 6650 or 6750 should do the trick, which is no where close to "A average gaming PC."

Nope. It is not "this" CPU and "that" GPU. The PS4 has a custom HSA APU with hUMA. Custom means that it is improved hardware available only for it. E.g. the iGPU has about 4x the compute queues of a 7990 (if my memory doesn't fail), there is an extra bus for accessing data...

Even if a PC had exactly that custom hardware, the PS4 would be still faster:

http://www.eurogamer.net/articles/digitalfoundry-inside-metro-last-light

Oles Shishkovstov: No, you just cannot compare consoles to PC directly. Consoles could do at least 2x what a comparable PC can due to the fixed platform and low-level access to hardware.

This is not very different from car vs motorbike comparison. A 100HP motorbike is much faster than a 100HP car.

4. Yes the hardware is better than last generation, however, the architecture with no doubt will be lowered. My Core 2 Quad Q8300 and Radeon 5570 are about the same performance of these consoles.

Not even close. It is more like

PS4 ~ i7 + GTX 680/780 Windows PC

Nope an i3 3220 is faster and its like an inbetween 7850 7870 gpu performance level but supposedly on 8000 series architecture. The only good thing about the consoles is hyperthreaded dual cores are here to stay for hardcore gaming.

You have no idea of what you are talking, but well... I am not saying anything new.

I don't know what fancy land you are in, but if your logic is correct, then almost every PC gamer would have a PS4 right now. -_-

In a comparison, realistically, its basically a Core 2 Quad and ~ a 6570-6670. Comparing that hardware to last generation, it's a huge jump. Realistically, if the Consoles had THAT much power, first of all, AMD would set a price MUCH higher than anything, and the console itself in turn would cost more than almost any High End PC out there.

This logic is just bad. For consoles, which are premade, and almost non-customize-able, to overpower Rigs that cost $2000 or more, and for such a low price, just does not make sense.

No, you're wrong...the HD 6670 is on VLIW4 architecture, and the HD 7XXX series is on GCN 1.0, and the consoles are on GCN 1.1 (HD 9XXX series coming soon).

Additionally, the count of the number of processors in the PS4 exceeds the count in a HD 7850 by a small margin, but is also more advanced technology. So, it really is essentially a HD ~9790/9830 (if such a thing were to exist in the next batch of cards)

As far as the cores go, you're not talking about a FX 8350 downclocked to ~2.2 GHz. You're talking about 8 core CPU with 8 FPUs and 3 ALUs per core (this, is more ALUs per core than FX series).

I think you don't understand this entirely...

A better comparison would be to take 2x i5-3350p and downclock to 2.2 GHz and allow developers to code to bare metal.

That should give you some idea of what you're looking at in the consoles...

 

[hcl123]

Excellent work of selective quoting and whole misunderstanding. You are quoting from slide 15 on "multicore performance". The phrase immediately above says "ARM multicores as efficient as Intel at the same frequency". They are comparing the old phone-level Tegra 2 and Tegra 3 chips to an i7. The Tegra chips are thermally constrained to phone-level power consumption. The i7 is not thermally constrained.

Thank you. I can read through 20+ pages of marketing BS and find the key slides. This group was paid millions of dollars to come up with a ARM cluster. Of course it will have its benefits and trade offs. Their own research showed that because Intel can run faster, it also had the highest performance/watt. That's rather key don't you think?

That's like saying a Porsche sucks if you limit it to 500rpm. Man that engine just can't put out any HP at 500rpm.

Yes they were older Tegra parts but they were compared to older i7 (Sandy) parts as well. By the 2015 time frame which they are going to production, Intel will be 4 generations from there (Ivy, Haswell, Broadwell, Skylake) and be at 10nm.

You can't just say ARM will be faster in the future as if Intel is sitting still. They're putting 10B+ into fabs and R&D a year.

I'm not dissing ARM. I've worked with ARM chips on an almost daily basis for years. Almost every project I work on has an ARM chip in it. They're everywhere. ARM is key to the market to keep Intel on their toes. It's in all of our best interest for the competition to remain fierce.

But the future is Quark! (kidding!)

Baanh! ... comparing an A9 with an i7 is no comparative at all... totally different objectives and chip uarch and else characteristics.

An A9 will never make it to big clusters.period

OTOH, the A57 (ARMv8) is about 2x the perf/watt compared to the A9, and this on the ARM reference designs as is, and old 40nm process.

So if the A9 is already comparable to the i7, the A57 could be 100% (2x) or more perf/watt than i7(edt). I think Intel knows this, that is why the recent uarches, and future ones to, are more about power than anything else.(WHICH IS A SHAME, x86 HAS BETTER CHANCES AT REAL *HIGH PERFORMANCE*, WHERE IT HAS THE ADVANTAGE OF DECADES OF AHEAD DEVELOPMENTS ... yet ARM v8 could be better for this to lol... )

ARM strooke an winner, in an unbiased view everybody know that, the A57 is already the same *PERFORMANCE* if not more, AND HAS THE POTENTICIAL to be much more performance for the ~same power, than equivalent x86 Jaguars/ silvermonts et all... and it has a clear way better perf/watt than those.... way better, to the point it doesn't need exotic fab processes like 14nm Finfet, to be competitive an WIN, OTOH intel needs this more than anything else, but then losing hard on the (per/watt)/$, unless they subsidize it, and how long they can hemorrhage money and then cook the books or compensate with other cash caws, is to be seem ( theoretically long time, just as "server" IA-64 Itanium and like ( the Itanics)... but not forever... )

[UPDATE: MAJOR PROBLEM of intel being a clear law of steep diminishing returns established with smaller fab nodes... clock/speed wont improve much unless big breakthroughs, matter of fact they are already flat for "bulk" no matter finfet, perf/watt will get less and less... to the point going smaller only with exponential costly fab processes(exotic techs), and this only for making things smaller, the rest is gone (which with 3D stacking will make going smaller also a lose lose preposition) ]

Yes x86 on low power, or better in perf/watt which serves server to, seems like another IA-64 *Itanic* in the making ( AMD seems to see the light, Intel will fight tooth and nail for very long until the very bitter end-> like with Itanics).

 

[hcl123]

Yes Intel needs dearly to pull their head out of the arse... want low power or server oriented per/watt go ARM... after all they already have a license... and give a good high perofrmance x86 chip, the Nehalem/SBN/IB/Hasfail/Boadfail like hell .. and etc.. are more than caput designs, were good on their time, but botched bentmarks can't fool everybody forever, and the professional world knows this... those intel uarches are not that good for multi-threading, compute, and mixing light and swift Integer crunching on the same pipeline execution clusters with way too large vectors (now 256 bit, future 512 bit) (edt) is horribly inefficient

Hasfail in action... 5 to 7% better with FMA code big LOL
http://www.realworldtech.com/forum/?threadid=135386&curpostid=135386

That is hwy they didn't implement FMA4, or any FMA in SBN, just too bad uarch for this... the same why HTM (hardware transactional memory) while scheduled perhaps wont appear not even in HSW-E (only lock elision), and the same with SPECULATIVE MULTI-THREADING(spMT) and or "data speculation" execution, which IP has being(and still is a bit) developed since long, intel uarch simply not good this, AMD simply has a golden opportunity with their BD uarch...

All that is where x86 could win clearly for ARM( way too many years ahead in development)... even games could have lock elision if not HTM, since they are massively parallel, and even take advantage of data speculation...

Its sad many users evangelize a brand on botched benchmarks, and repeat single-thread mantras, instead of demanding for more, much more... is nice for the IDM because they can cut R&D money by half... and the stupid morons can yet pay double for 10 or 20% more performance, all that is needed is a good spin... all this is why ARM can yet win it all...

 

[noob2222]

^^ what about software compatibility and high end developes such as adobe (photoshop) autocad, maxxon, ect. running on ARM?

In order for ARM to just break into the desktop market, they need something to convince people that their current software will work instead of trying to get people to upgrade to ARM them go buy all new software.

Cost of upgrading a pc: cost of the pc

Cost of migrating from pc to ARM DT: cost of the pc +cost of the NEW software.

Until we can directly compare ARM to x86, its all just marketing hype. For some reason, a few people here are overjoyed by the hype and expect ARM to walk away with the market next month.

The current x86 android OS is still buggy and ARM doesn't run well on Windows RT. There is no direct comparison just like there is no direct transition.

Am I dissing ARM? not really. Im hoping my next phone won't crash all the time and that TOMS will quit screwing with the forums making it "mobile friendly".

I can't even post on my phone anymore, constant "server timeout", yet this is the cpu thats supposedly going to take over. The more I use my phone, the less I want to see it on the DT.

Would I try a DT ARM computer? maybe if it was cheap, but im not going to just rush out and buy the newest thing just because its ARM based.

 

[juanrga]

Intel wasn't competing against arm before but each year they are taking less electricity for more performance and intel have now thought why not make easy work of ARM. Do you really think ARM have a chance to deliver as good cpu at same power envelope as intel. ARM are now scared big time as intel are going to take a lot of there marketshare as its cheap and much better performance at similar power.

Where is steamroller t series is it made up?

Intel has been competing against ARM for years. I have been reading the intel is going to kill ARM fallacy since 2009 or so.

Intel is cheap now? Wait, this deserves an adequate reply


I think that you missed the facts of the last week when Apple released its new chip. Even Intel fans are admitting the facts:

Dude, even I will admit that it seems Intel needs to admit defeat to ARM. 1.3GHz dual core A7 with Smartphone class TDP is beating 4 core 1.4-2.4GHz Bay Trail with Tablet class TDP. And that's "28nm" vs 22nm.

The IPC of the A7 is on par with Ivy Bridge parts! Clock that sucker to 3GHz and it'll have a no worry replacement for the Macbook Air line and threaten even with x86 emulation.

I cannot parse your question about steamroller.

Just because it says, "OH BOY THIS HAS A 7850 AND A 8 CORE CPU IN IT!!" A 7850 will totally fit in a tiny Console case. These "7850's" will definitely be less than a low profile 7850. Also take into consideration Power Consumption and Heat. The performance will literally drop, okay not a 5570, but something like a 6650 or 6750 should do the trick, which is no where close to "A average gaming PC."

Nope. It is not "this" CPU and "that" GPU. The PS4 has a custom HSA APU with hUMA. Custom means that it is improved hardware available only for it. E.g. the iGPU has about 4x the compute queues of a 7990 (if my memory doesn't fail), there is an extra bus for accessing data...

Even if a PC had exactly that custom hardware, the PS4 would be still faster:

http://www.eurogamer.net/articles/digitalfoundry-inside-metro-last-light

Oles Shishkovstov: No, you just cannot compare consoles to PC directly. Consoles could do at least 2x what a comparable PC can due to the fixed platform and low-level access to hardware.

This is not very different from car vs motorbike comparison. A 100HP motorbike is much faster than a 100HP car.

4. Yes the hardware is better than last generation, however, the architecture with no doubt will be lowered. My Core 2 Quad Q8300 and Radeon 5570 are about the same performance of these consoles.

Not even close. It is more like

PS4 ~ i7 + GTX 680/780 Windows PC

Nope an i3 3220 is faster and its like an inbetween 7850 7870 gpu performance level but supposedly on 8000 series architecture. The only good thing about the consoles is hyperthreaded dual cores are here to stay for hardcore gaming.

You have no idea of what you are talking, but well... I am not saying anything new.

I don't know what fancy land you are in, but if your logic is correct, then almost every PC gamer would have a PS4 right now. -_-

In a comparison, realistically, its basically a Core 2 Quad and ~ a 6570-6670. Comparing that hardware to last generation, it's a huge jump. Realistically, if the Consoles had THAT much power, first of all, AMD would set a price MUCH higher than anything, and the console itself in turn would cost more than almost any High End PC out there.

This logic is just bad. For consoles, which are premade, and almost non-customize-able, to overpower Rigs that cost $2000 or more, and for such a low price, just does not make sense.

Therefore, not only you continue ignoring that the hardware in the PS4/XboxOne is custom (improved) and cannot be purchased in PC stores, but you also ignore this:

Oles Shishkovstov: No, you just cannot compare consoles to PC directly. Consoles could do at least 2x what a comparable PC can due to the fixed platform and low-level access to hardware.

LOL

Excellent work of selective quoting and whole misunderstanding. You are quoting from slide 15 on "multicore performance". The phrase immediately above says "ARM multicores as efficient as Intel at the same frequency". They are comparing the old phone-level Tegra 2 and Tegra 3 chips to an i7. The Tegra chips are thermally constrained to phone-level power consumption. The i7 is not thermally constrained.

Thank you. I can read through 20+ pages of marketing BS and find the key slides. This group was paid millions of dollars to come up with a ARM cluster. Of course it will have its benefits and trade offs. Their own research showed that because Intel can run faster, it also had the highest performance/watt. That's rather key don't you think?

That's like saying a Porsche sucks if you limit it to 500rpm. Man that engine just can't put out any HP at 500rpm.

Yes they were older Tegra parts but they were compared to older i7 (Sandy) parts as well. By the 2015 time frame which they are going to production, Intel will be 4 generations from there (Ivy, Haswell, Broadwell, Skylake) and be at 10nm.

You can't just say ARM will be faster in the future as if Intel is sitting still. They're putting 10B+ into fabs and R&D a year.

I'm not dissing ARM. I've worked with ARM chips on an almost daily basis for years. Almost every project I work on has an ARM chip in it. They're everywhere. ARM is key to the market to keep Intel on their toes. It's in all of our best interest for the competition to remain fierce.

But the future is Quark! (kidding!)

Therefore you decided to ignore again the quote that says that Intel is less efficient. Funny.

Certain problems require exascale level compute. The problem is that current petascale supercomputers cannot be scaled up 1000x due to the immense power consumption. Therefore research groups concluded that ARM was the kind of efficient architecture needed to build supercomputers at that scale.

The research group has shown that ancient Tegra chips are more efficient than the i7 in single core perf. and had the same efficiency that the i7 in multicore at the same freq.

Ivy Bridge is about a 10-15% better than Sandy and Haswell is (being generous here) about a 10% better than Ivy. I don't know how Broadwell and Skylake will perform, but I doubt that will introduce more than a 20%.

However, the current Tegra 4 is 10x/5x better than Tegra 2/3 they tested. And Parker will be about 100x better and comes in 2015, probably in 14nm. Intel Broadwell also comes in 14nm but it is not more for desktops

http://www.fudzilla.com/home/item/32524-broadwell-won%E2%80%99t-make-it-to-desktop

For the sake of comparison 100x is a 9900% >>> 20%

 

[juanrga]

Until we can directly compare ARM to x86, its all just marketing hype. For some reason, a few people here are overjoyed by the hype and expect ARM to walk away with the market next month.

Funny, because only a pair of posts ago we were discussing a comparison of ARM vs x86 using different software and benchmarks.

 

[juanrga]

hcl123 yet intel are ahead of everyone by miles

This is only in your imagination.

 

[de5_Roy]

^^ what about software compatibility <snipped>

that's a good point imo. arm needs a good, flagship os. according to my (rather newbie) observation, arm's market-dominating consumer os is android and it's already at windows level for malware writers. what i am trying to say is that android is already at the level microsoft was with windows and i don't mean that as a compliment. that is a good way to spin against arm. android's fragmentation also gets in the way. not to mention it's pretty much google's glorified data mining tool (microsoft is no better).

microsoft can run windows on arm (wart) but their core business is still based on x86, which they won't risk right now. apple has ios but they're too closed off. ironically, if anyone can offer a viable desktop os for arm, apple would be the best candidate. the ones that left are firefox os and tizen/meego/whatever-that-keeps-dying-and-hardware-companies-keep-renaming-and-then-abandoning os. supporting apple for arm would be the same as supporting intel.

i don't see current PC (x86) software suddenly being taken over by their arm counterparts. public perception of arm is still angry birds i.e. synonymous with casual and consumer devices. arm's business model also prevents anyone(of it's partners) to stand out in a significant way.

so as good as arm hardware, uarch and software is, it'll take a Lot more than that to pry pcs off of general public. technology and business are different. too bad (for arm) that even current low end pcs (that's dual core cpus, you haters ) and laptops are capable for general populace (i.e. pcs they already own) and they will be quite unwilling to part with them. instead arm will stay as complementary portable device for the forseeable future.

however. if arm armada can keep people interested in tablets a little while longer, like 2-4 years while arm 64bit and process tech advance, tablets will have the best chance to replace all pcs.

 

[Cazalan]

As far as the cores go, you're not talking about a FX 8350 downclocked to ~2.2 GHz. You're talking about 8 core CPU with 8 FPUs and 3 ALUs per core (this, is more ALUs per core than FX series).

Jaguar only has 2 ALU per core. The older K10 Athlons had 3.

 

[juanrga]

Nope. Intel is releasing a previously unplanned Haswell refresh for desktops because current desktop Haswell has been a complete fail. Intel is moving to mobile, because desktop is dying. Therefore Broadwell comes only for mobile. This is the same reason why AMD is not releasing 8-core FX Steamroller and moving to APUs (FXs don't make sense for mobile).

 

[8350rocks]

http://www.fudzilla.com/home/item/32524-broadwell-won%E2%80%99t-make-it-to-desktop

This basically explains why intel are going to flex there muscle and release finally an 8 core haswell cpu as basically they are focusing on laptops and releasing only a haswell refresh for desktops. This is the advantage of being so far aheaa brings though. I reckon 6 core cpus will become more mainstream in the £300 range with haswell and that should beat any amd cpu even a 220w tdp 8 core 5ghz steamroller if they made one. So basically intel are focusing on laptops while the demand for new desktops are low because cpus are already fast enough from sandy bridge onwards and people mainly just upgrade the graphics card.

Hasfail-E won't be near what you think it is.

It won't scale much better than the current 6 core CPUs except in integer calculation scenarios, it will also be about $1500 for an 8 core CPU equivalent of the 3970x and the die size will be massive...if they manage to get it to work, it will likely be at a TDP around 140W+ considering Hasfail has higher TDP than IB, and the X79 platform are still TDP of ~120W with 4930k.

Perf/Watt goes out the window on E series...because it's the same thing for Intel as the 5 GHz 9590 is for AMD. It's a "we don't care" statement...though Intel makes it massively more expensive.

 

[de5_Roy]

http://www.fudzilla.com/home/item/32524-broadwell-won%E2%80%99t-make-it-to-desktop

Spoiler

This basically explains why intel are going to flex there muscle and release finally an 8 core haswell cpu as basically they are focusing on laptops and releasing only a haswell refresh for desktops. This is the advantage of being so far aheaa brings though. I reckon 6 core cpus will become more mainstream in the £300 range with haswell and that should beat any amd cpu even a 220w tdp 8 core 5ghz steamroller if they made one. So basically intel are focusing on laptops while the demand for new desktops are low because cpus are already fast enough from sandy bridge onwards and people mainly just upgrade the graphics card.

you don't follow pcs news much, do you? intel already has 6,8 and 10 core cpus. 8c and 10c ones are xeons - big cpus unlike mainstream desktop ones.
according to older cpu roadmaps, we're supposed to have mainstream 6c core i5 and 8c core i7 a while ago, but that didn't happen. instead we got overpriced ivb and haswell cpus. in case you don't know, current intel cpus are overpriced a lot.
broadwell not being on desktop is not a good news for customers.
there's no muscle flexing or anything, intel can't get people to buy their stuff, so they're extorting more than usual.

 

[8350rocks]

^^ what about software compatibility <snipped>

that's a good point imo. arm needs a good, flagship os. according to my (rather newbie) observation, arm's market-dominating consumer os is android and it's already at windows level for malware writers. what i am trying to say is that android is already at the level microsoft was with windows and i don't mean that as a compliment. that is a good way to spin against arm. android's fragmentation also gets in the way. not to mention it's pretty much google's glorified data mining tool (microsoft is no better).

microsoft can run windows on arm (wart) but their core business is still based on x86, which they won't risk right now. apple has ios but they're too closed off. ironically, if anyone can offer a viable desktop os for arm, apple would be the best candidate. the ones that left are firefox os and tizen/meego/whatever-that-keeps-dying-and-hardware-companies-keep-renaming-and-then-abandoning os. supporting apple for arm would be the same as supporting intel.

i don't see current PC (x86) software suddenly being taken over by their arm counterparts. public perception of arm is still angry birds i.e. synonymous with casual and consumer devices. arm's business model also prevents anyone(of it's partners) to stand out in a significant way.

so as good as arm hardware, uarch and software is, it'll take a Lot more than that to pry pcs off of general public. technology and business are different. too bad (for arm) that even current low end pcs (that's dual core cpus, you haters ) and laptops are capable for general populace (i.e. pcs they already own) and they will be quite unwilling to part with them. instead arm will stay as complementary portable device for the forseeable future.

however. if arm armada can keep people interested in tablets a little while longer, like 2-4 years while arm 64bit and process tech advance, tablets will have the best chance to replace all pcs.

Actually Ubuntu is about as close to a flagship OS as ARM will get for a while...(at least in DT end)