Arm's Cortex-A76 Could Be The First True Challenger To x86 Chips On Laptops

[Lucian Armasu]

Arm announced a new IP family, including the high-performance Cortex-A76 CPU core, Mali-G76 GPU, and Mali-V76 VPU with 8k@60fps decoding support.

Arm's Cortex-A76 Could Be The First True Challenger To x86 Chips On Laptops : Read more

 

[fpga123]

The article doesn't even substantiate any comparative benchmark results. Click-baiting has become the norm, it seems.

 

[Sleepy_Hollowed]

That's significant, since on GeekBench, the iPhone 8 is severely close to an Alienware with a desktop 6700 non K CPU, which is insane, considering that Apple's throttles quite a bit.

I can only imagine this new ARM CPU with active cooling on a laptop.

 

[therealduckofdeath]

Benchmarking on mobile platforms, especially iOS, is so faked it's pointless. Compare with ported games like PUBG and you'll see that the synthetic benchmark numbers has nothing to do with reality. That game struggles to maintain 30 FPS on any ARM ecosystem even though it's running at ultra low settings with upscaled graphics.

 

[therealduckofdeath]

Fix your website, Tom's! (...for the billionth time!)
All the broken links caused by your forced redirections to the buggy .co.uk domain are annoying as hell. Stop deleting all complaints about this and just fix your site, please?
Also, fix your broken cookie disclaimer. That bug you're having on that part could actually cost you fines in the EU as I'm pretty sure it's not compliant.
Mobile platform functionality is also pretty terrible, which goes under the first complaint.

 

[computerguy72]

Heh Sleepy_Hollowed - "severely close" to an older 6700? wtf... It's not a good realworld test and even using geekbench it's laughably slower.

 

[bit_user]

The article doesn't even substantiate any comparative benchmark results. Click-baiting has become the norm, it seems.

Get a clue. For someone complaining about click-baiting, your media literacy seems shockingly poor.

They're reporting on ARM's announcement of their new IP. If there's any silicon implementing this design, it's probably only test chips in ARM's labs. Certainly no products yet contain it. Probably not until at least Q4 2018 or early 2019.

As you may know, ARM doesn't sell chips. They sell designs. Their customers license these designs for use in their own chips. Most of those customers then sell the chips to yet another set of companies, which are the device makers. This is why ARM needs to publicize its new designs well in advance of actual products, and why there's such a lag between the two.

Of course, there are slight variations on this model, such as Apple and Samsung (which license ARM IP and then use it in their own devices), but the main point is that ARM doesn't sell chips. So, there's nothing for independent journalists to test, and there won't be for a while.

 

[JonDol]

Even if the performances they are boasting are not faked this time, the adoption of the ARM Windozes will be slooooow and power users may even ignore them just as they did with the Chrome books and eePCs.... Just imagine the time it takes for all the tools whether free or commercial, we use on a daily basis to be available for ARM. I for one, use a few dozens daily. That performance boost alone is simply not enough to convice users to adopt it.

 

[AgentLozen]

Get a clue.

I've been checking Craigslist everyday for the past month but I've come up empty handed. Can anyone recommend a good clue?

 

[bit_user]

Get a clue.

I've been checking Craigslist everyday for the past month but I've come up empty handed. Can anyone recommend a good clue?

Edited. See above.

BTW, I think you're clueful enough.

 

[redgarl]

The day it will run Crysis, then we can talk I guess...

 

[justin.m.beauvais]

I doubt ARM has many, if any, hardware samples of their device as of yet. It is more likely that they are performing simulations of how the CPU will work, under the best possible circumstances.

Also, I have no doubt that ARM CPUs can compete with x86 in certain and specific tasks, but that is because ARM chips are pretty specifically designed for certain tasks. The real strength of x86 is how generalized it is. Where as CPUs like the ARM can perform some tasks pretty fast, x86 is faster on a much larger range of tasks. A good analogy is that x86 is the pickup truck with the big V8. Go anywhere, do anything, plenty fast. ARM is a little sports car with a 4 cyl turbo. Better on gas, pretty quick, but definitely not suitable for all situations.

 

[gggplaya]

The article doesn't even substantiate any comparative benchmark results. Click-baiting has become the norm, it seems.

Get a clue. For someone complaining about click-baiting, your media literacy seems shockingly poor.

They're reporting on ARM's announcement of their new IP. If there's any silicon implementing this design, it's probably only test chips in ARM's labs. Certainly no products yet contain it. Probably not until at least Q4 2018 or early 2019.

As you may know, ARM doesn't sell chips. They sell designs. Their customers license these designs for use in their own chips. Most of those customers then sell the chips to yet another set of companies, which are the device makers. This is why ARM needs to publicize its new designs well in advance of actual products, and why there's such a lag between the two.

Of course, there are slight variations on this model, such as Apple and Samsung (which license ARM IP and then use it in their own devices), but the main point is that ARM doesn't sell chips. So, there's nothing for independent journalists to test, and there won't be for a while.

No I agree with FPGA123. The author of this article was a bit clickbaity with his title and claims!!!! You can't just make claims without some sort of relevant comparison. He can make claims that ARM has a new huge generational leap in performance with the A76, but they can't claim it to be the "first true challenger to x86" without giving at least some relevant comparison metric to go off of. The only data ARM gave were comparison's to it's older designs. Since the relevant data wasn't there, and the author didn't take the time to extrapolate the data from ARM's claims of % of improvement of it's last generation, and using the benchmarks from the last generation to give some data points. Then they shouldn't make such a claim without something to back it up. PURE CLICKBAIT!!

 

[pug_s]

I haven't seen any products coming out with the Cortex A75 yet and they are announcing this? I would imagine that in order to get this kind of performance gains, it has to be like 7mm or even 5mm process.

 

[bit_user]

The author of this article was a bit clickbaity with his title and claims!!!! You can't just make claims without some sort of relevant comparison. He can make claims that ARM has a new huge generational leap in performance with the A76, but they can't claim it to be the "first true challenger to x86" without giving at least some relevant comparison metric to go off of. The only data ARM gave were comparison's to it's older designs. Since the relevant data wasn't there, and the author didn't take the time to extrapolate the data from ARM's claims of % of improvement of it's last generation, and using the benchmarks from the last generation to give some data points.

That's a fair point. Lucian isn't always very good at providing context.

Then they shouldn't make such a claim without something to back it up. PURE CLICKBAIT!!

That said, it's definitely not PURE CLICKBAIT. It's reporting a real announcement, which is actual news. Second, it conveyed the new information that ARM released. Third, if you do the work (which isn't exactly hard), you'll see the claims are justifiable. But the article did significantly reach with the title, and then neglected to connect the dots.

 

[southernshark]

I could see a specialized market for these, schools and commercial settings, where lap tops don't have to do anything too intensive and while at the same time need to stay charged all day.

 

[bit_user]

I haven't seen any products coming out with the Cortex A75

A modified version of it features in the Snapdragon 845.

yet and they are announcing this?

You know that Intel and AMD are working on multiple generations of CPUs at the same time, right? The difference is that we only get firm announcements from them when they basically have silicon shipping to customers, whereas ARM announces as each design is completed. So, it's as if the pipeline of one of the big chipmakers were bisected.

I would imagine that in order to get this kind of performance gains, it has to be like 7mm or even 5mm process.

The second slide in the article compares:

• 
  ■ A73 @ 16 nm
  ■ A75 @ 10 nm
  ■ A76 @ 7 nm

 

[gggplaya]

No I agree with FPGA123.

Really? Then you think Lucian should've somehow broken into ARM's labs and benchmarked it, himself? Or maybe not reported the announcement and then one day we start seeing reviews of A76-based chromebooks show up as if it came out of nowhere?

The author of this article was a bit clickbaity with his title and claims!!!! You can't just make claims without some sort of relevant comparison. He can make claims that ARM has a new huge generational leap in performance with the A76, but they can't claim it to be the "first true challenger to x86" without giving at least some relevant comparison metric to go off of. The only data ARM gave were comparison's to it's older designs. Since the relevant data wasn't there, and the author didn't take the time to extrapolate the data from ARM's claims of % of improvement of it's last generation, and using the benchmarks from the last generation to give some data points.

That's a fair point. Lucian isn't always very good at providing context.

Then they shouldn't make such a claim without something to back it up. PURE CLICKBAIT!!

That said, it's definitely not PURE CLICKBAIT. It's reporting a real announcement, which is actual news. Second, it conveyed the new information that ARM released. Third, if you do the work (which isn't exactly hard), you'll see the claims are justifiable. But the article did significantly reach with the title, and then neglected to connect the dots.

Why should "I" do the work??? The author is the one publishing an article with a very bold claim, even putting it in the title, yet provides no relevant chart or data to back it up. He says it several times throughout the article, but only shows performance differences between arm's older designs. Zero comparison's to x86. He could have showed a chart with a comparison between x86 and an older ARM design, like this quick google search popped up (https://www.phonearena.com/news/iPhone-X-beats-Macbook-Pro-in-benchmarks_id98035) a comparison between Apple's awesome custom arm design A11 chip vs the arm chip in the Galaxy S8 and a macbook pro with the Core i5 7360u. As you can see, if ARM's A76 design catches up to Apple's arm licensed custom A11 chip. It could be competitive with the Core i5 in multicore performance at the stated 35% improvement, but still lag behind in single core performance.

Again, this shouldn't be my job to do. It should be the author's job to put out a complete article. I cringe to agree with Trump, but call it "lazy journalism"? It's all clickbait and first to publish now?? Without actually doing any research or substantiating claims? That's what we're complaining about.

Haven't you ever written a paper in college?? When you make a bold game changing claim, you need to cite a source or back it up with some data. That should be journalism 101.

 

[bit_user]

The author of this article was a bit clickbaity with his title and claims!!!! You can't just make claims without some sort of relevant comparison. He can make claims that ARM has a new huge generational leap in performance with the A76, but they can't claim it to be the "first true challenger to x86" without giving at least some relevant comparison metric to go off of. The only data ARM gave were comparison's to it's older designs. Since the relevant data wasn't there, and the author didn't take the time to extrapolate the data from ARM's claims of % of improvement of it's last generation, and using the benchmarks from the last generation to give some data points.

That's a fair point. Lucian isn't always very good at providing context.

Then they shouldn't make such a claim without something to back it up. PURE CLICKBAIT!!

That said, it's definitely not PURE CLICKBAIT. It's reporting a real announcement, which is actual news. Second, it conveyed the new information that ARM released. Third, if you do the work (which isn't exactly hard), you'll see the claims are justifiable. But the article did significantly reach with the title, and then neglected to connect the dots.

Why should "I" do the work???

Not saying you should - I thought I agreed with you, on that point. I was just trying to say that the claim isn't completely unjustifiable.

But clickbait is where you take something unremarkable and try to paint it as news or something much more special than it is. I don't think that's what happened, here.

 

[gggplaya]

Why should "I" do the work???
Not saying you should - I thought I agreed with you, on that point. I was just trying to say that the claim isn't completely unjustifiable.

But clickbait is where you take something unremarkable and try to paint it as news or something much more special than it is. I don't think that's what happened, here.

That's not my definition of clickbait. Mine is a headline or thumbnail which shows one thing, with an expectation of that being in the article or video, but being mislead and receiving very little of it or none at all. In this case, the first half of the title is correct, that's not the point of contention. The second half of the title is where I felt mislead. I expected to read at least something substantiating or discussing the claim. But instead, I felt only the first half of the headline was discussed and the second half was only stated by the author and I felt used to entice readers. The legitimacy of the claim is irrelevant, most clickbait is usually factually correct.

 

[techy1966]

Personally I do not care how fast they get to Intel or AMD & how they tested this. Lets have a clock for clock compare and lets not test an i3 or Pentium dual core against an 8 core ARM CPU and then we can talk. Core for core a Ryzen 8/16 or Intel 8/16 setup will lay the smack down on these chips and lets not even get into how they have to emulate x86 instructions if you want to run every day Windows software made for x86. Lets see how they perform in some of the modern day AAA titles that lay waste to even the best CPU's or for that matter a AAA title from a few years back If they can beat a true x86 CPU then I will show great respect for ARM and what they can do.

I am not saying they are bad but they work best in mobile devices such as phones and tablets and low rent laptops like the Google Chrome books.

 

[bit_user]

The second half of the title is where I felt mislead. I expected to read at least something substantiating or discussing the claim. But instead, I felt only the first half of the headline was discussed and the second half was only stated by the author and I felt used to entice readers. The legitimacy of the claim is irrelevant, most clickbait is usually factually correct.

I'm sorry you felt cheated. I thought the article was worthwhile, though I agree he either shouldn't have used that title or could've added value by extrapolating how it might score on some benchmarks via x86.

Maybe you'll like this better:
https://www.anandtech.com/show/12785/arm-cortex-a76-cpu-unveiled-7nm-powerhouse

...which focuses mostly on comparing it to other ARM cores.

They do offer this:

One comparison that was made during the event is that in terms of area, three A76’s with larger caches would fit inside the size of a Skylake core – all while within 10% of the IPC of the Intel CPU, but obviously there’s also process node scaling considerations to take into account.

Relevant. Maybe Lucian saw but forgot to quote that bit.

I think the "process note scaling" clause means they're directly comparing the area of a "7 nm" A76 against a 14 nm Skylake, so I disregard the area aspect of the comparison. But the IPC is very interesting. I wish we knew how that was measured. I guess there's another unsubstantiated claim, for you.

 

[bit_user]

Personally I do not care how fast they get to Intel or AMD & how they tested this. Lets have a clock for clock compare and lets not test an i3 or Pentium dual core against an 8 core ARM CPU and then we can talk. Core for core a Ryzen 8/16 or Intel 8/16 setup will lay the smack down on these chips

Depends on what it's for. Intel/AMD can always win by burning more power. But if they're focused on laptops or tablets of a given size & weight and with a certain battery life, then the fair comparison would be with x86 that can deliver the same battery life and size/weight.

That would probably mean either comparing with Goldmont+ (or whatever comes after) or maybe some then-current equivalent of the core m series (with full cores and fast turbos, but only dual-core and achingly slow base clocks).

In other words, I think ARM is still playing the perf/W card. In a high-powered desktop (or gaming laptop), you're still going to want Intel or AMD.

 

[bit_user]

The real strength of x86 is how generalized it is. Where as CPUs like the ARM can perform some tasks pretty fast, x86 is faster on a much larger range of tasks. A good analogy is that x86 is the pickup truck with the big V8. Go anywhere, do anything, plenty fast. ARM is a little sports car with a 4 cyl turbo. Better on gas, pretty quick, but definitely not suitable for all situations.

Are you saying that ARM has yet to step completely out of its low-power/high-efficiency niche?

Or that there's something fundamental about the ARMv8-A ISA that fundamentally puts current x86-64 performance out of their reach? If so, what's lacking?

BTW, I think the A76's dual 128-bit vector pipelines are comparable to what's in Ryzen.

 

[madvfrh]

Acorn Archimedes returns. Nuff said.