Intel: 8th-Gen Processors Are 30% Faster Than 7th-Gen

[Guest]

Intel announced during its Computex 2017 keynote that the eighth-generation Coffee Lake processors would feature 30% more performance than the seventh-generation Kaby Lake processors.

Intel: 8th-Gen Processors Are 30% Faster Than 7th-Gen : Read more

 

[iPanda]

wait, maybe i'm just tired... but shouldn't it have scaled a bit better than that with the faster clocks, memory, and more cores/threads?

 

[lanyard_loop]

15W TDP ? Can this number be correct? More like 150W TDP I think.

 

[legokangpalla]

Is it Tick-tock-tock or Tick-tock-tock-tock?
Broadwell was Tick, so Skylake-Kaby-Coffee is....

 

[bigpinkdragon286]

They are comparing a 2C/4T @ 3.5 GHz to a 4C/8T that can turbo up to 4 GHz. Of course performance is going to scale with computing resources. What is the actual IPC improvement, or is there any?

 

[mac_angel]

double the cores, double the threads, faster RAM, and they are bragging about a 30% increase?

 

[InvalidError]

[quotemsg=19755685,0,938891]What is the actual IPC improvement, or is there any?[/quotemsg]
I'd guess the same IPC per core gain as usual, which would be 5-8%.

 

[TheKappaccino]

@IPANDA Well the 7700HQ (the current base quad core and a 45W chip) scores about 1.55x higher than the 7500U they are comparing to. Taking that into account, as well as a possible decrease in base clock to help manage the increase in cores, a 30% increase in performance is reasonable. Especially if they don't want the new quad core U-series chips performing too well and cannibalizing the more expensive HQ models.

 

[TEAMSWITCHER]

"double the cores, double the threads, faster RAM, and they are bragging about a 30% increase?"

Do you think that an 18-Core i9 or 16-Core ThreadRipper will run at 4.5 GHz? Clock speeds will drop to counter power consumption.

 

[Graham_23]

15W TDP is correct for the "U" models. Those are the ultra-low power processors.

 

[anbello262]

This will only apply to low power cpus, which would actually mean a 30% efficiency improvement at most. Unlocked desktop processors with no power limit will most likely be quite similar to what we already have...

 

[none12345]

Wait a 4c/8t cpu clocked higher is only 30% faster then a 2c/4t cpu? Thats pathetic if true, it should be >100% faster.

Im assuming they mean 30% faster/watt. Which is something i guess, but thats not what is implied.

 

[none12345]

"I'd guess the same IPC per core gain as usual, which would be 5-8%."

It was 0% for kaby lake. In normal work loads clock for clock it was the same. There are some small improvements for media encoding, and a slight clock speed increase, but IPC in normal workloads, its the same as skylake.

I hope they do better then the last 'normal'

 

[IInuyasha74]

According to Intel's new update scheme, this should use essentially the same architecture as Kaby Lake, but with a move to a new manufacturing process. I don't think Intel has confirmed it yet, but that would mean a move to 10nm, which would make sense. A die shrink helps to improve efficiency more than anything else, and that would make it easier to push double the cores into the same TDP. It would also explain the high TDP, as that 4GHz mark will likely be for just a single core. The performance increase likely come from a mixture of slight IPC improvement, higher clock speeds, and possibly higher core count. I wouldn't be surprised to see a more mainstream 6-core chip from Intel.

 

[dudmont]

Garbage report. Show of hands, who here thinks the 4c/8t unlocked K at the top of the stack is going be 30% faster than the 7700k?

 

[DerekA_C]

[quotemsg=19755782,0,182337]"double the cores, double the threads, faster RAM, and they are bragging about a 30% increase?"

Do you think that an 18-Core i9 or 16-Core ThreadRipper will run at 4.5 GHz? Clock speeds will drop to counter power consumption.[/quotemsg]

I love the competition but, Intel is still going to lag behind in the end because they aren't honest, notice no talk of actual IPC gain but when Ryzen came they talked a lot about it and when it came to finished product AMD gained 60% IPC performance within a generation that's pretty impressive. If they do even half that with Kyzen they will beat Intel. ThreadRipper 16 is the base model for home studio pro gamer streaming buyers which will be more affordable with really decent performance. That said only a matter of time before AMD will release 32 core for Home business and 64 core for servers.
They are about to match Intel at 7nm node within around same time. AMD is for sure catching up, but in a hurry thank you Dr. Lisa Su, smart engineering woman PERIOD. They own the console market which is a much bigger base then PC gaming. Sure it's a smaller check but it's consistent with console upgrades coming every 1.5 years. However with the major push for that many more cores will benefit everyone making programmers to code for more cores making better software and games. Along with better cross porting games within all OS's.

 

[razor512]

One thing that needs to be explored is the full range of power specifications. One trick that intel likes to pull with their lower powered CPUs, are alternate TDP ranges where a 15 watt TDP is set as a default but they may add an alternative power limit that the CPU can use if the thermal headroom is available.

For example, on some of the really low powered Intel CPUs targeted at tablets and ultrabooks, you may see a CPU with a factory default package power limit of 11 watts, and a and an alternative power limit that is used if thermal headroom is available for 15 watts. And tablet makers looking to get away with very little cooling, will set a power limit of 6 watts or less, thus causing the CPU to not even reach its base speed if the integrated GPU enters its 3D acceleration mode, since the package will TDP throttle.

Intel will then push out newer generations where they up the boost clock, and then up the alternate power limits, and then claim massive performance improvements, but when implemented in the real world, those benefits disappear because people realize that they only last for about 30 seconds before the system throttles.

 

[InvalidError]

[quotemsg=19756056,0,2428111]They own the console market which is a much bigger base then PC gaming. Sure it's a smaller check but it's consistent with console upgrades coming every 1.5 years.[/quotemsg]
There is no guarantee that the next generation consoles will continue using AMD's chips. Microsoft and Sony could decide to follow Nintendo's lead and go with Nvidia's next time they decide to break backward compatibility.

 

[acme64]

are the desktop cpu's coming out this year as well?

 

[LORD_ORION]

What a surprise... they can increase "speed" by 30% per generation if they really wanted to.... but usually no reason unless AMD is on their heels.

 

[Serban13]

This is all "estimates" and all. No real test. The end product will be 5-6% faster overall. Intel knows it's still got the upper hand and they are playing it.

 

[dalauder]

[quotemsg=19755782,0,182337]"double the cores, double the threads, faster RAM, and they are bragging about a 30% increase?"

Do you think that an 18-Core i9 or 16-Core ThreadRipper will run at 4.5 GHz? Clock speeds will drop to counter power consumption.[/quotemsg]No, the 8th gen appears to offer a 500MHz clock increase (not that the base clock was divulged).

 

[dalauder]

[quotemsg=19755686,0,169108]double the cores, double the threads, faster RAM, and they are bragging about a 30% increase?[/quotemsg]That's exactly my thought. This benchmark, that supposedly runs well on Intels should utilize the cores better than that...unless there's an issue with the thermal ceiling? Maybe the chip realistically can only boost 1 core to 4.0GHz and so it is ACTUALLY SLOWER per core on multi-threaded workloads (which it makes up by doubling the core count)? There's a lot of strange and unimpressive massaging of numbers Intel is doing with that press release. The complete lack of mention of IPC is telling though.

Obviously, AMD won't have another 50% IPC improvement next gen, but with a 10% gain and a bit better programming optimizations by other parties, they should catch Intel on all fronts next generation.

On the other hand, 4.0GHz/3.5GHz should only give a 14% gain. If it yields double that, it's either faster IPC by an unusally good jump (coincidentally, 14%) or a strange utilization of extra cores. Either way, they've piqued my interest.

 

[ET3D]

[quotemsg=19756319,0,404510unless there's an issue with the thermal ceiling? Maybe the chip realistically can only boost 1 core to 4.0GHz and so it is ACTUALLY SLOWER per core on multi-threaded workloads (which it makes up by doubling the core count)?[/quotemsg]

That would be my guess. Single core performance is probably up by 14% (clock difference) and multi-core by 30% due to more cores but limited thermal headroom. Still a pretty decent gain for users.

 

[DerekA_C]

[quotemsg=19756132,0,125865][quotemsg=19756056,0,2428111]They own the console market which is a much bigger base then PC gaming. Sure it's a smaller check but it's consistent with console upgrades coming every 1.5 years.[/quotemsg]
There is no guarantee that the next generation consoles will continue using AMD's chips. Microsoft and Sony could decide to follow Nintendo's lead and go with Nvidia's next time they decide to break backward compatibility.[/quotemsg]

I would say Microsoft has invested pretty heavily with AMD from software(dx12) to hardware CPU/GPU/APU I don't see them ever leaving specially if they match or outperform other consoles and Project Scorpio a VR console slated for November or December. Nintendo kept chumping on everything and it bit them in the butt. Loss of income due to limitations of there last several systems including the Switch, it is a niche and it won't last for hardcore gamers just doesn't have the titles and the controllers suck compared to the others and it isn't really VR capable and that is the future again Nintendo lagging behind by going with the cheapest parts possible. So they are trying to merge their console and portables to an all in one to say they have the largest market share which in reality they don't.