News Intel Unveils 10th-Gen Core Processors, 10nm Ice Lake, 18% IPC Improvement, Sunny Cove Cores, Gen11 Graphics, Thunderbolt 3

[PaulAlcorn]

Intel is finally shipping its 10nm Ice Lake processors, and they come with a whopping 18% IPC improvement.

Intel Unveils 10th-Gen Core Processors, 10nm Ice Lake, 18% IPC Improvement, Sunny Cove Cores, Gen11 Graphics, Thunderbolt 3 : Read more

 

[AlistairAB]

So Intel has 18 percent IPC over 4 years, AMD does 15 percent in one year.

 

[southernshark]

Talk about an Intel Fanboy.... This article was definitely written by one.

"Intel's next gen graphics are well known..." Haha... that's funny stuff. No one has even seen the next gen graphics yet.

And as for IceLake, unveiling it and shipping CPUs aren't the same thing. I haven't seen any press release about desk top CPUs shipping in 10nm this year.

 

[Deleted member 14196]

Yawn. They can keep their overpriced junk. I am so sick of these fanboys writing articles as well

 

[Delicieuxz]

The article says:

AMD claims that its new third-gen Ryzen products have exceeded Intel's Skylake single-core performance, long the hallmark of Intel's dominating performance, albeit by small 1 to 3% margins

Though, AMD's 1 - 3% performance single-threaded performance increase compares higher clock-speed Intel chips to lower clock-speed AMD chips. So, AMD's IPC could be more appreciably higher than Intel's.

 

[BaRoMeTrIc]

Talk about an Intel Fanboy.... This article was definitely written by one.

"Intel's next gen graphics are well known..." Haha... that's funny stuff. No one has even seen the next gen graphics yet.

And as for IceLake, unveiling it and shipping CPUs aren't the same thing. I haven't seen any press release about desk top CPUs shipping in 10nm this year.

Intel stated they won't ship 10nm desktop and server cpus until 2020

 

[BaRoMeTrIc]

The article says:



Though, AMD's 1 - 3% performance single-threaded performance increase compares higher clock-speed Intel chips to lower clock-speed AMD chips. So, AMD's IPC could be more appreciably higher than Intel's.

IPC doesn't change with higher frequencies. IPC and Frequency are 2 separate discussions

 

[Math Geek]

they had to do something to keep their fanboys busy until they can actually put out a new chip. some pretty ppt slides should keep them pretending for the next year

but really an 18% increase should just about make up for all the losses due to patches for all the vulnerabilities the chips have. so i'd expect them to stay on par with where they are now overall. of course they'll not apply those patches when testing to make it look better. but hey who's really paying attention anyway.........

 

[kinggremlin]

So Intel has 18 percent IPC over 4 years, AMD does 15 percent in one year.

It's easier to improve on a 6 minute mile than it is a 4 minute mile.

 

[ingtar33]

color me highly skeptical on those IPC claims.

Intel's own IPC hasn't budged a percentage since skylake, yet their ipc graph in this press release claimed steady improvement in IPC thrugh the various skylake refreshes. I think they're playing games with the meaning of IPC again, like how they did when they would release any one of their skylake refreshes with identical ipc and claim massive ipc improvements, which were made by higher core counts and clock speeds, proving forever intel's definition of ipc is different from everyone else's.

 

[jeremyj_83]

It's easier to improve on a 6 minute mile than it is a 4 minute mile.

When Zen 1 was launched it had an IPC of Broadwell, which was about 5% less than the Sky Lake architecture. The biggest reason for the wide performance gap between them was Intel's huge clock speed advantage in single threaded applications. Zen+ increased IPC by 3% so now AMD was within 2% clock for clock with Intel. With Zen 2 we are getting a 15% IPC boost so AMD will have a 10-12% performance advantage over Intel at the same clock speed, or about 1-3% in ST with both CPUs at max boost. Saying that AMD was doing a 6min miles isn't correct at all, it was more like a 4:15 mile that they improved to a 3:45 mile in 2 years, where as Intel was at a 4 mile and moved to a 3:42 mile in 4 years. Now don't forget that Intel isn't releasing Ice Lake to the desktop until sometime in 2020. When you look at the top end GHz it maxes out at 4.1GHz whereas the i7-8665U hits 4.8GHz which means Intel is having a very hard time increasing the clock speed on the CPU. That is why they aren't releasing it on the desktop since it won't be able to clock high enough right now to beat their current i9-9900K.

 

[jeremyj_83]

color me highly skeptical on those IPC claims.

Intel's own IPC hasn't budged a percentage since skylake, yet their ipc graph in this press release claimed steady improvement in IPC thrugh the various skylake refreshes. I think they're playing games with the meaning of IPC again, like how they did when they would release any one of their skylake refreshes with identical ipc and claim massive ipc improvements, which were made by higher core counts and clock speeds, proving forever intel's definition of ipc is different from everyone else's.

They are showing single threaded performance increase over the different releases to skylake. Each iteration had a higher boost clock so the ST performance would be higher. I agree that we will have to wait until products are available and then have benchmarks run with each CPU at the same clock to identify if the IPC really did increase by their claim.

 

[ElectrO_90]

So I didn't see in the article you will not need a new motherboard
Still the 10th series will suffer with the same architetcure bugs for the last 10+ years with so many vulnerbilities, swiss cheese has less holes.

 

[justin.m.beauvais]

"The Ice Lake lineup is destined for notebooks and other thing and light devices."

Notebooks are about to get more awesome, unfortunately "budget" will probably not be in the description of these new devices. Too bad Intel can't seem to get things right for the desktop. AMD will be proceeding unopposed for the near future in the desktop realm. The real game changer though will be AMD 7nm mobile devices, which are still a while off. My educated guess is that they will drop sometime next year. Hopefully Intel finds their footing and can crank these things out while they have the advantage, because a competitive mobile experience will really put a hurting on them.

 

[kinggremlin]

So I didn't see in the article you will not need a new motherboard
Still the 10th series will suffer with the same architetcure bugs for the last 10+ years with so many vulnerbilities, swiss cheese has less holes.

This is a mobile launch. When have you been able to drop a new CPU into your laptop?

 

[kinggremlin]

When Zen 1 was launched it had an IPC of Broadwell, which was about 5% less than the Sky Lake architecture. The biggest reason for the wide performance gap between them was Intel's huge clock speed advantage in single threaded applications. Zen+ increased IPC by 3% so now AMD was within 2% clock for clock with Intel. With Zen 2 we are getting a 15% IPC boost so AMD will have a 10-12% performance advantage over Intel at the same clock speed, or about 1-3% in ST with both CPUs at max boost. Saying that AMD was doing a 6min miles isn't correct at all, it was more like a 4:15 mile that they improved to a 3:45 mile in 2 years, where as Intel was at a 4 mile and moved to a 3:42 mile in 4 years. Now don't forget that Intel isn't releasing Ice Lake to the desktop until sometime in 2020. When you look at the top end GHz it maxes out at 4.1GHz whereas the i7-8665U hits 4.8GHz which means Intel is having a very hard time increasing the clock speed on the CPU. That is why they aren't releasing it on the desktop since it won't be able to clock high enough right now to beat their current i9-9900K.

Back in 2015, the Intel roadmap had 10nm Cannonlake launching in 2016. That would have put the architecture refresh in 2017. Obviously that didn't happen and now we are getting the 10nm migration and the architecture refresh at the same time in 2019. It isn't the fault of the architecture team that 10nm is 3 years late. Once 10nm was delayed, they couldn't just backport sunnycove to 14nm. It doesn't work that way.

 

[jeremyj_83]

Back in 2015, the Intel roadmap had 10nm Cannonlake launching in 2016. That would have put the architecture refresh in 2017. Obviously that didn't happen and now we are getting the 10nm migration and the architecture refresh at the same time in 2019. It isn't the fault of the architecture team that 10nm is 3 years late. Once 10nm was delayed, they couldn't just backport sunnycove to 14nm. It doesn't work that way.

I know you cannot back-port the changes into the old architecture, but that doesn't stop the fact that it still took 4 for the new architecture to come out. Don't forget in 2015 Intel was still saying tick-tock for their cadence, but within a year changed it to Process–architecture-optimization model. That changed their roadmap and roadmaps change all the time since the roadmap is nothing more than basically a wish list. What matters is available product and it took 4 years to get the IPC increase.

 

[kinggremlin]

I know you cannot back-port the changes into the old architecture, but that doesn't stop the fact that it still took 4 for the new architecture to come out. Don't forget in 2015 Intel was still saying tick-tock for their cadence, but within a year changed it to Process–architecture-optimization model. That changed their roadmap and roadmaps change all the time since the roadmap is nothing more than basically a wish list. What matters is available product and it took 4 years to get the IPC increase.

You're still not getting it. The delays are because of problems with the process side, not the architecture side. Yes, the tick tock schedule got screwed up with 14mm. Because of the huge delay for 10nm, we are getting the tick and the tock at the same time, as the architecture has likely been ready for years waiting for 10nm. With Intel claiming they are on schedule with 7nm in 2021, it's unlikely we will see an optimization step with 10nm, and we will see an accelerated new architecture after Sunnycove as you can be assured the architecture team hasn't been sitting around doing nothing these last couple of years.

 

[jeremyj_83]

You're still not getting it. The delays are because of problems with the process side, not the architecture side. Yes, the tick tock schedule got screwed up with 14mm. Because of the huge delay for 10nm, we are getting the tick and the tock at the same time, as the architecture has likely been ready for years waiting for 10nm. With Intel claiming they are on schedule with 7nm in 2021, it's unlikely we will see an optimization step with 10nm, and we will see an accelerated new architecture after Sunnycove as you can be assured the architecture team hasn't been sitting around doing nothing these last couple of years.

I totally understand where the problems arose. You don't get it that it having something "ready" means nothing if you cannot manufacture the item. They might have had the basic design figured out in time, but we don't know that for sure. If you cannot produce samples of the chip you cannot validate it for use, which means the architecture wasn't ready.

 

[jimmysmitty]

So Intel has 18 percent IPC over 4 years, AMD does 15 percent in one year.

Except that 15% was catching up. AMD had a massive IPC deficit pre Ryzen so making massive gains is not that hard to do.

The article says:



Though, AMD's 1 - 3% performance single-threaded performance increase compares higher clock-speed Intel chips to lower clock-speed AMD chips. So, AMD's IPC could be more appreciably higher than Intel's.

IPC has nothing to do with clock speed differences. It is literally Instructions Per Clock meaning that at the same clock speed Intel is faster than AMD. The extra clock speed just adds to the advantage.

they had to do something to keep their fanboys busy until they can actually put out a new chip. some pretty ppt slides should keep them pretending for the next year

but really an 18% increase should just about make up for all the losses due to patches for all the vulnerabilities the chips have. so i'd expect them to stay on par with where they are now overall. of course they'll not apply those patches when testing to make it look better. but hey who's really paying attention anyway.........

Sunnycove is supposed to have a lot of the fixed baked into hardware which will negate the need for the patches and the performance loss associated with the patches.

If you noticed, Intel tends to compare their CPUs against their older CPUs and not AMDs. In fact their only mention is that their new iGPU will take the APU performance crown from AMD. I wont hold my breath but the news we have had has me hopeful that we wil have some real APU competition and maybe a thrid dGPU competor to help keep the other two in check. I mean its not like AMD has challenged nVidia in the past few years, the price of the RTX series is proof of that.

When Zen 1 was launched it had an IPC of Broadwell, which was about 5% less than the Sky Lake architecture. The biggest reason for the wide performance gap between them was Intel's huge clock speed advantage in single threaded applications. Zen+ increased IPC by 3% so now AMD was within 2% clock for clock with Intel. With Zen 2 we are getting a 15% IPC boost so AMD will have a 10-12% performance advantage over Intel at the same clock speed, or about 1-3% in ST with both CPUs at max boost. Saying that AMD was doing a 6min miles isn't correct at all, it was more like a 4:15 mile that they improved to a 3:45 mile in 2 years, where as Intel was at a 4 mile and moved to a 3:42 mile in 4 years. Now don't forget that Intel isn't releasing Ice Lake to the desktop until sometime in 2020. When you look at the top end GHz it maxes out at 4.1GHz whereas the i7-8665U hits 4.8GHz which means Intel is having a very hard time increasing the clock speed on the CPU. That is why they aren't releasing it on the desktop since it won't be able to clock high enough right now to beat their current i9-9900K.

These are only up to 28W TDP processors meant for mobile applications. They don;t need the highest possible clock speed, they need power efficiency. People want longer battery life and more features.

They are showing single threaded performance increase over the different releases to skylake. Each iteration had a higher boost clock so the ST performance would be higher. I agree that we will have to wait until products are available and then have benchmarks run with each CPU at the same clock to identify if the IPC really did increase by their claim.

Boost has nothing to do with IPC. IPC is a solid clock speed not a boost clock speed.

So I didn't see in the article you will not need a new motherboard
Still the 10th series will suffer with the same architetcure bugs for the last 10+ years with so many vulnerbilities, swiss cheese has less holes.

Well they are mobile CPUs so there is no need to mention it.

That said, I don't doubt that any desktop chip with anything based on this uArch will need a new board. Will probably be a new socket.

As for the vulnerabilities, what makes you so sure? Do you think Intel wouldn't bake fixes for the known ones in? Do you really think they are willing to leave those in a new uArch and risk the same issues?

I don't see why people don't see this as a good thing. I am glad Intel is going to finally push out 10nm and hopefully they will be able to push to 7nm in 2021 as planned. I prefer to see both sides pushing it to their limit. Otherwise we will just end up with one or the other milking us for our money.

I also like the other improvements such as built in TB and the possibility of TB being integrated into USB 4. Thats fantastic news TBH as TB is betetr than the USB standard as it is.

 

[jeremyj_83]

These are only up to 28W TDP processors meant for mobile applications. They don;t need the highest possible clock speed, they need power efficiency. People want longer battery life and more features.

Yes those are up to 28W TDP CPUs, but that doesn't stop the fact that in 14nm Intel already has U series (15W) CPUs that can boost significantly higher than their 10nm cousins. That leads to the assumption that Intel cannot reach high enough clock speeds for the 10nm CPU to be faster than the 14nm CPU that has lower IPC. If Intel could clock Ice Lake high enough, you know they would release a desktop part for it right now to replace the 9900K, but since the speed isn't there it would make them look silly to release a 10700K/10900K that is slower than the 9700K/9900K.

Boost has nothing to do with IPC. IPC is a solid clock speed not a boost clock speed.

Where did I say anything about boost and IPC being the same thing?

What I responded to:
"color me highly skeptical on those IPC claims.

Intel's own IPC hasn't budged a percentage since skylake, yet their ipc graph in this press release claimed steady improvement in IPC thrugh the various skylake refreshes. I think they're playing games with the meaning of IPC again, like how they did when they would release any one of their skylake refreshes with identical ipc and claim massive ipc improvements, which were made by higher core counts and clock speeds, proving forever intel's definition of ipc is different from everyone else's."


My response:
"They are showing single threaded performance increase over the different releases to skylake. Each iteration had a higher boost clock so the ST performance would be higher. I agree that we will have to wait until products are available and then have benchmarks run with each CPU at the same clock to identify if the IPC really did increase by their claim."

The first sentence of my response was the person was talking about the graph that shows how ST performance increase with each refresh of Sky Lake. The second sentence is about testing Intel's claim of an 18% IPC increase over Sky Lake. The only way to do that is to have 2 CPUs tested at the same clock speed, if the Ice Lake CPU has 18% more performance then their claim of the higher IPC is valid, if it isn't within the margin of error across multiple tests then Intel's claim isn't valid.

Except that 15% was catching up. AMD had a massive IPC deficit pre Ryzen so making massive gains is not that hard to do.

Yes pre Ryzen AMD had a massive IPC deficit. However, with Zen 1 AMD was within 5-7% of the IPC of Intel and Zen+ within 2-5% of Intel's IPC. The 15% IPC boost that AMD is claiming, again will have to be tested against Gen 1 Zen with CPUs at the same clock, would put AMD 10-13% ahead of Intel's Sky Lake IPC.

 

[cryoburner]

color me highly skeptical on those IPC claims.

It's easy to increase IPC over Skylake when Skylake's IPC keeps going down every time a new vulnerability patch is released. : 3

https://www.tomshardware.com/reviews/intel-amd-storage-test-mds-zombieload-ridl-fallout,6146.html

Really though, if they are implementing fixes in hardware, it's very possible that they could account for a good chunk of those performance gains.

 

[bigdragon]

Intel has suffered too many vulnerabilities and major performance hits from patches. Boasting about IPC improvements, new features, and generational change concerns me because I expect these numbers are going to get reigned-in in 3 to 9 months when another Meltdown or MDS comes out.

Intel needs to step it up. Even their Meltdown and Spectre "fixed in-silicon" chips still have problems according to the MDS/Fallout/RIDL researchers. These sorts of puff presentations do little to alleviate concerns that Intel hardware is a bad investment right now.

 

[dudeman509]

Yes those are up to 28W TDP CPUs, but that doesn't stop the fact that in 14nm Intel already has U series (15W) CPUs that can boost significantly higher than their 10nm cousins. That leads to the assumption that Intel cannot reach high enough clock speeds for the 10nm CPU to be faster than the 14nm CPU that has lower IPC.

Looking back, they did this with the Skylake -> Kaby Lake chips across the board. New architecture, new process (Broadwell was the change to 14nm, but not many of those actually shipped before Skylake was out), low-ish to same clockspeeds as Broadwell.

Kaby Lake? Just bump up the clocks and sell it! People will snatch that up!

 

[setx]

The article make me feel that Intel is quite desperate.
Those 18% increase of IPC do sound like they are comparing software-mitigated current CPUs with hardware-fixed new ones.