Intel kills 10nm ?? oO

[EsaT]

Interesting fine print on their performance slides for that. Is Ice lake still affected by the security issues, and patches?



This one is even better!

Guess they run Coffee Lake with HyperThreading disabled and all holes patched to get performance difference...

Anyway after so many previous rebrandings of Skylake, why should we anymore believe Ice Lake to actually be any major re-design?
They didn't even change naming away from lakes.

one of the problems Intel's been having is controlling the thermal differentials between cobalt and the surrounding silicon, because they expand and contract at such different rates it causes the cobalt interconnects to crack.

Intel version of "bumpgate"...
Just this time baking it in oven won't help.

 

[4K8KW10]

Last I heard nVidia was still evaluating things like supply and unit cost and TSMC was far from a done deal. I don't have any newer information at the moment.

NVIDIA’s Ampere GPU Launching In 2020, Will Be Based On Samsung’s 7nm EUV Process
https://wccftech.com/nvidias-ampere-gpu-launching-in-2020-will-be-based-on-samsungs-7nm-euv-process/

 

[Freddie1980]

Wccftech.............. it's the equivalent of Tthe Sun for daily tech news.

 

[4K8KW10]

Wccftech.............. it's the equivalent of Tthe Sun for daily tech news.

Seriously?

Nvidia to adopt Samsung 7nm EUV process in next-gen GPU production https://www.digitimes.com/news/a20190605PD200.html

 

[Rroff]

Seriously?

Nvidia to adopt Samsung 7nm EUV process in next-gen GPU production https://www.digitimes.com/news/a20190605PD200.html

They've already changed their tune 3 times over nVidia 7nm and I know for a fact that atleast two of those times the information wasn't correct.

 

[4K8KW10]

They've already changed their tune 3 times over nVidia 7nm and I know for a fact that atleast two of those times the information wasn't correct.

Now I know you are not being serious which is nothing wrong, TBH.

 

[Grim5]

New Leaks out of China

http://tieba.baidu.com/p/6164223711

tl:dr If correct (take lots of salt) it shows that Intel's upcoming 10nm 4c/8t Sunnycove chips has made big IPC improvements.

The i7-1065G7 ES chip clocked to all core 3.7ghz has the same single core score as an all core 5.3ghz 9700k

 

[drunkenmaster]

https://hothardware.com/reviews/intel-core-i9-9980xe-review

Hard to find information but based on general clock speeds of the chip, I'm under the impression AVX512 makes a big difference in the CPu-Z benchmark. IF that clockspeed shown in real (hard to tell with boost clocks) then AVX512 then that 3Ghz using AVX512 is around where a 9900k is (ballpark) using much higher clocks.

People going gaga over the score, though in fairness most Intel guys are saying it's almost certainly fake, seem to be forgetting that Icelake is supposed to have AVX512 in the consumer chips.

AVX512 should me a few benchmarks it's hugely faster in, most more real world usage sees near no usage at home.

 

[Deleted member 209350]

New Leaks out of China

http://tieba.baidu.com/p/6164223711

tl:dr If correct (take lots of salt) it shows that Intel's upcoming 10nm 4c/8t Sunnycove chips has made big IPC improvements.

The i7-1065G7 ES chip clocked to all core 3.7ghz has the same single core score as an all core 5.3ghz 9700k

Seems a bit fishy that they were making such little IPC gains from gen to gen, and now all of a sudden there's such a massive jump?

 

[orbitalwalsh]

Seems a bit fishy that they were making such little IPC gains from gen to gen, and now all of a sudden there's such a massive jump?

have SPENT a longggggggggggggggggggggggggggg time trying to get 10nm right. and 7nm node was always meant to target intels 10nm node since naming/sizing of nodes got a little bit out of tune with the rise of ARM chips and trying to make them 'smaller'

their 10nm development was separate to everything else. AMD had flow between node shrinks and die designs. believe intel stated this would change now with their designs going ahead. as everyone knows, competition finally pushing Intel in the right direction !

i hope 10nm performance is TRUE.. it needs to be ! need to keep AMD on their toes just as much! they could rely on node shrinks to much etc though Zen2 luckily wasn't what ryzen 1+ was

 

[Hawk Tuah]

I don't get how Intel are having such problems with 10nm when Qualcomm, HiSilicon and Mediatek all seem to be managing 7nm without issue.

Is there something in the X86 architecture which makes this harder?

 

[Minstadave]

I don't get how Intel are having such problems with 10nm when Qualcomm, HiSilicon and Mediatek all seem to be managing 7nm without issue.

Is there something in the X86 architecture which makes this harder?

The clock speeds they're hitting are all sub 2.5ghz aren't they? And they're all small die size, ARM based architectures.

They're all likely to be very efficiency focused whereas Intel need performance.

 

[r22snapper]

Seems a bit fishy that they were making such little IPC gains from gen to gen, and now all of a sudden there's such a massive jump?

Reminds me of all the "leaks" before Skylake that were showing 15-20% IPC increase...

 

[Deleted member 209350]

have SPENT a longggggggggggggggggggggggggggg time trying to get 10nm right. and 7nm node was always meant to target intels 10nm node since naming/sizing of nodes got a little bit out of tune with the rise of ARM chips and trying to make them 'smaller'

their 10nm development was separate to everything else. AMD had flow between node shrinks and die designs. believe intel stated this would change now with their designs going ahead. as everyone knows, competition finally pushing Intel in the right direction !

i hope 10nm performance is TRUE.. it needs to be ! need to keep AMD on their toes just as much! they could rely on node shrinks to much etc though Zen2 luckily wasn't what ryzen 1+ was

So do you think its a fake leak then? I mean a 3.7ghz quad core outperforming pretty much everything? It almost seems too good to be true

 

[4K8KW10]

So do you think its a fake leak then? I mean a 3.7ghz quad core outperforming pretty much everything? It almost seems too good to be true

But the quad-core is outperforming the others only in a single-threaded application, in multi-threaded applications it stands no chance because of not sufficient amount of cores/threads.

 

[orbitalwalsh]

So do you think its a fake leak then? I mean a 3.7ghz quad core outperforming pretty much everything? It almost seems too good to be true

well optimised apps maybe . gains are to high but they should higher then what intel got from zen to zen2 ! since its been in the making that long! I dont hold out much for speed, not on their first 10nm tap out . hoping it would be faster ! but like zen3, expect the second or refined version on the 10nm node to perform better

But the quad-core is outperforming the others only in a single-threaded application, in multi-threaded applications it stands no chance because of not sufficient amount of cores/threads.

not always , if the gains are so good across single core- they could claw performance last against larger core counts. I.e it would now take a 4 core to match an 8 core instead of being a 6 core unit with lots of speed .

wouldn't surprise me if 10nm was designed all those years ago at 4 core with no intention of having more cores!!!! tell intel came along ! lol why we are getting 10 cores on 14nm+++ and 4 on 10nm

 

[Deleted member 209350]

well optimised apps maybe . gains are to high but they should higher then what intel got from zen to zen2 ! since its been in the making that long! I dont hold out much for speed, not on their first 10nm tap out . hoping it would be faster ! but like zen3, expect the second or refined version on the 10nm node to perform better




not always , if the gains are so good across single core- they could claw performance last against larger core counts. I.e it would now take a 4 core to match an 8 core instead of being a 6 core unit with lots of speed .

wouldn't surprise me if 10nm was designed all those years ago at 4 core with no intention of having more cores!!!! tell intel came along ! lol why we are getting 10 cores on 14nm+++ and 4 on 10nm

Makes sense I suppose. But I still somehow doubt that this is the level of performance intel is going to release. If it is true however, then AMD are back to being screwed

 

[orbitalwalsh]

Makes sense I suppose. But I still somehow doubt that this is the level of performance intel is going to release. If it is true however, then AMD are back to being screwed

then they'll hopefully attack again - we'll finally be in a loop. Currently now AMD are ABLE to increase prices! they'll need another reason to drop them or lower next product

 

[HeX]

Seems a bit fishy that they were making such little IPC gains from gen to gen, and now all of a sudden there's such a massive jump?

43% IPC jump is akin to AMD FX --> ZEN, Intel have either been sitting on a revolutionary new CPU architecture for years while they milked the market, or it's a load of rubbish!

 

[hominid]

When ever, if it actually does, Intel releases on 10nm is almost guaranteed to have lower clocks than 14nm(~4Gz at best), so the large, for Intel, IPC improvements will be needed to attempt to offset that clock speed regression.

I'm no expert on Intel processors so I don't much about AVX, although I do know it's more for business(servers) for now. I know AVX512 hasn't been rushed into *Lake as I've read about it since ~2017 being included in *Lake, but there is still very little use for it for normal users.

AVX will have a big impact on benchmarks, especially the ones Intel will highlight, but once all the mitigations are implemented, are they all?, for the average user I think there could be a performance regression.. It's probably why, also bad yields and process issues, we haven't seen them yet, their only known 10nm release, a dual core with graphics disabled, had yield rates in the high single digits - so WAY off being viable.