Question Intel Q3: Ouch

[jpiniero]

https://d1io3yog0oux5.cloudfront.net/_c675d259e2193ee11667e182e2fab5ca/intel/news/2020-10-22_Intel_Reports_Third_Quarter_2020_Financial_1424.pdf

Desktop volume down 16%, Notebook up 16% but ASP down 7%.

Q4 forecast is bad too, they are expecting 14% decline in revenue compared to Q4 last year.

 

[Exist50]

Shouldn't this go in the Lakes/Rapids thread?

 

[cortexa99]

I won't bother why only 8 core and power consumption but everything looks quite pessimistic, especially 'double digit IPC increase' statement sounds pretty unconfident......it doesn't include the number behind decimal point right??

 

[ondma]

I won't bother why only 8 core and power consumption but everything looks quite pessimistic, especially 'double digit IPC increase' statement sounds pretty unconfident......it doesn't include the number behind decimal point right??

What did you want, the IPC increase to 5 decimal places? Obviously it will vary by application as well.

 

[aigomorla]

What did you want, the IPC increase to 5 decimal places? Obviously it will vary by application as well.

well to intel double digit means 10.... they want it to sound more, but in reality its 10... maybe 12 at best.
We probably wont see another Conroe/Kentsfield for a long time if ever from intel which offered a TRIPLE digit increase if overclocking was factored in.

 

[ondma]

well to intel double digit means 10.... they want it to sound more, but in reality its 10... maybe 12 at best.
We probably wont see another Conroe/Kentsfield for a long time if ever from intel which offered a TRIPLE digit increase if overclocking was factored in.

Conroe was a pretty much one of a kind event. AMD made a big jump (40-50% ???) from Bulldozer to Zen (starting from a low bar with Vishera) but was still behind the 2 or 3 year old Skylake architecture. Subsequent AMD steps have been what around 15%, maybe 20 at most?

 

[chrisjames61]

Conroe was a pretty much one of a kind event. AMD made a big jump (40-50% ???) from Bulldozer to Zen (starting from a low bar with Vishera) but was still behind the 2 or 3 year old Skylake architecture. Subsequent AMD steps have been what around 15%, maybe 20 at most?

I think it was 52% jump vs Piledriver.

 

[rainy]

I think it was 52% jump vs Piledriver.

It was indeed 52 percent, however vs Excavator.

 

[Thunder 57]

Conroe was a pretty much one of a kind event. AMD made a big jump (40-50% ???) from Bulldozer to Zen (starting from a low bar with Vishera) but was still behind the 2 or 3 year old Skylake architecture. Subsequent AMD steps have been what around 15%, maybe 20 at most?

I wonder what was going on in Intel's Marketing Department at that time. They were used to pumping GHz. All of a sudden low clocked Core 2 Duo's beat the crap out of P4 derivatives. clocked way faster. Clearly they saw it coming as the switched to model numbers shortly after Prescott I believe.

It was indeed 52 percent, however vs Excavator.

That is what I remember.

 

[blckgrffn]

It was indeed 52 percent, however vs Excavator.

Yeah but wasn’t excavator essentially an APU only CPU? I mean, I have one to this day but it seemed niche even then. I used/sold a lot more bulldozer and steamroller (mainly, iirc) CPUs back in the day. Even got a check for buying them just this year 😂

I am trying to say in general terms that hat whole generation stunk so much and was so limited in scope I want to know the % IPC uplift vs Thuban more than vs Excavator

 

[TheGiant]

well to intel double digit means 10....

you mean up to 10

 

[DrMrLordX]

250W PL2 is brutal...

It's the same as the 10900k. Intel buyers should be used to it by now.

Shouldn't this go in the Lakes/Rapids thread?

You're right, actually.

 

[Zepp]

Yeah but wasn’t excavator essentially an APU only CPU? I mean, I have one to this day but it seemed niche even then. I used/sold a lot more bulldozer and steamroller (mainly, iirc) CPUs back in the day. Even got a check for buying them just this year 😂

I am trying to say in general terms that hat whole generation stunk so much and was so limited in scope I want to know the % IPC uplift vs Thuban more than vs Excavator

Thuban was on average in between piledriver and steamroller. excavator was around 10% or so uplift from steamroller IIRC.

 

[Spartak]

AMD 2015: 28nm Excavator X4 970 GB5 sc 538 / mc 1834
AMD 2020: 7(10)nm Zen3 5950X GB5 sc 1663 / mc 15860
improvement in 5 years sc 310% mc 865%

Intel 2015: 14nm Skylake 6700K GB5 sc 1150 / mc 4485
Intel 2020: 14nm Skylake 10900K GB5 sc 1412 / mc 11096
improvement in 5 years sc 23% mc 247%

Brutal contrast.

 

[dullard]

AMD 2015: 28nm Excavator X4 970 GB5 sc 538 / mc 1834
AMD 2020: 7(10)nm Zen3 5950X GB5 sc 1663 / mc 15860
improvement in 5 years sc 310% mc 865%

Intel 2015: 14nm Skylake 6700K GB5 sc 1150 / mc 4485
Intel 2020: 14nm Skylake 10900K GB5 sc 1412 / mc 11096
improvement in 5 years sc 23% mc 247%

Brutal contrast.

Your point is correct, TSMC has done a great job--way better than Intel lately. But your examples are awful. X4 970 was not in 2015, it was 2017. Also the 5950x is what, at least 10x the price that the X4 970 was. So are you going to add a ~1000% increase in price column for AMD and 44% increase in price for Intel?

 

[Gideon]

Your point is correct, TSMC has done a great job--way better than Intel lately. But your examples are awful. X4 970 was not in 2015, it was 2017. Also the 5950x is what, at least 10x the price that the X4 970 was. So are you going to add a ~1000% increase in price column for AMD and 44% increase in price for Intel?

Agreed that the examples are needlessly cherrypicked to give AMD an unfair advantage (that is totally unnecessary as they'll win by a lot regardless).

A more fairer example would have been FX-9590 that scored 671 ST and 3545 MT. Still an insane uplift but "just" 248% in ST and 445% in MT.

What he is correct about though is that X970 (with the similar A12-9800) was the first generation top-of-the-line AM4 socket CPU. It's crazy how well that socket has scaled.

 

[ondma]

AMD 2015: 28nm Excavator X4 970 GB5 sc 538 / mc 1834
AMD 2020: 7(10)nm Zen3 5950X GB5 sc 1663 / mc 15860
improvement in 5 years sc 310% mc 865%

Intel 2015: 14nm Skylake 6700K GB5 sc 1150 / mc 4485
Intel 2020: 14nm Skylake 10900K GB5 sc 1412 / mc 11096
improvement in 5 years sc 23% mc 247%

Brutal contrast.

Easy to show a big gain when you start from far behind with a low score. Even after all that, AMD is only ~15% ahead in single thread.

 

[UsandThem]

Even after all that, AMD is only ~15% ahead in single thread.

Even if they were only ahead with 1% in single thread performance that would be a big deal for AMD. Why? Because it was really was the last thing Intel always advertised (along with 720P and 1080P gaming performance). From basically not having a product that could compete with Intel's CPUs for the last decade, to overtaking them in both MT and ST performance in 3 years. I'd call that pretty impressive.

 

[Fanatical Meat]

Your point is correct, TSMC has done a great job--way better than Intel lately. But your examples are awful. X4 970 was not in 2015, it was 2017. Also the 5950x is what, at least 10x the price that the X4 970 was. So are you going to add a ~1000% increase in price column for AMD and 44% increase in price for Intel?

I want to add that the previous comparison is showing how little intel has improved and also showing how crappy AMD was pre Ryzen.
Not a fan boy just saying those chips were dogs. Particularly Bulldozer.

 

[aigomorla]

Easy to show a big gain when you start from far behind with a low score. Even after all that, AMD is only ~15% ahead in single thread.

15% behind on a higher nm node is not that bad.

However they are losing the market like a land slide, its building up momentum.
But at the same note, AMD is no longer a budget build anymore.
The price on a B550 motherboard is absolutely ridiculous when one compares it to a X570.
Not to mention they have the highest priced enthusiast tier processor on record so far.
This is something intel has never done, although making us upgrade boards each gen, but a H series board was always budget orientated vs a Z series.

 

[scannall]

The price on a B550 motherboard is absolutely ridiculous when one compares it to a X570.

The thing is though, PCIe4 requires more board layers, and more expensive signaling components to work. More expensive to make than the B450's for example. So they will cost more. No way around it yet. Though when Intel starts pushing PCIe4 then maybe the component cost will go down. The number of layers won't though.

 

[uzzi38]

But at the same note, AMD is no longer a budget build anymore.
The price on a B550 motherboard is absolutely ridiculous when one compares it to a X570.

I would agree if it weren't for the fact that A520 still exists and is perfectly adequate for budget builds with a Ryzen 5 or 7. Depending in the SKU - I'd stick to the 65W ones.

You only lose core overclocking as far as performance features go. You still keep memory overclocking which is where the biggest performance improvement are going to come from.

 

[TheELF]

The motivation to use a more expensive node is to make a superior (or even competitive) product so you can maintain market share without drastic price cuts. Yes, Intel is making a lot of money right now using 14 nm, but it cant continue indefinitely.

But you think that TSMC can go on indefinitely with making smaller nodes thinking that the next node will be the one that will make intel start losing money?
Yes, Intel is making a lot of money right now using 14 nm and they will continue to use 14nm until 14nm isn't making them a lot of money anymore.
If at that point, when intel is losing money, they can't produce 10nm (or whatever) then they will be in trouble

I won't bother why only 8 core and power consumption but everything looks quite pessimistic, especially 'double digit IPC increase' statement sounds pretty unconfident......it doesn't include the number behind decimal point right??

Because intel is still only competing against themselves,RL will be based on sunny cove and sunny has 25% more compute units plus whatever else is increased(cache, frontside, ram freq. etc) and intel already claimed 18% higher IPC so the new 8core CPU will have the same multithreaded scores as the 10900k with a much higher single score which will make them boatloads of money, next gen will have a 10core CPU and HTT enabled on the other CPUs which will make them boatloads of money once again.

Examining Intel's Ice Lake Processors: Taking a Bite of the Sunny Cove Microarchitecture

www.anandtech.com

 

[DrMrLordX]

But you think that TSMC can go on indefinitely with making smaller nodes thinking that the next node will be the one that will make intel start losing money?

TSMC doesn't care if Intel "starts losing money". They have enough customers to buy every wafer they can produce regardless. They have a lot of customers, and that list of customers is growing. Including Intel. I wasn't going to post this because I think the article rephrases a bunch of stuff we know already, but here we go!

Intel contemplates outsourcing advanced production, upending Oregon’s central role

The company says a decision is likely in January.

www.oregonlive.com

And the idea that Intel is only "competing against themselves" is lolololol no

 

[maddie]

TSMC doesn't care if Intel "starts losing money". They have enough customers to buy every wafer they can produce regardless. They have a lot of customers, and that list of customers is growing. Including Intel. I wasn't going to post this because I think the article rephrases a bunch of stuff we know already, but here we go!

Intel contemplates outsourcing advanced production, upending Oregon’s central role

The company says a decision is likely in January.

www.oregonlive.com

And the idea that Intel is only "competing against themselves" is lolololol no

Yep. All TSMC wants, as all companies, is to grow. With a limited TAM (meaning not infinite), by definition if one party increases revenue share, the other loses. It's a secondary effect, not a primary goal.