'Blasphemy!' says Intel (kind of) as they attack AMD using Zen 2 cores in a 7000 series chip

[Agall]

31 minutes ago, porina said:

They sat at 14nm forever because they couldn't move forward. Everything was broken. Even 14nm was broken at launch. Look at Broadwell. It launched late with desktop getting two offerings just to say they released it at all. They're still on their recovery path right now, and the end of the road is in sight. The next two years will be very interesting.

 

About the only way out they could have had earlier would be shifting to TSMC.

 

Do you not count Intel 7? Or are you sticking to their old NAME of 10nm? Intel 7 still doesn't seem as polished as N7 but it's pretty close.

 

In just over a week we'll have on sale Meteor Lake cores on Intel 4. While the absence of desktop parts makes it less interesting to many around here, it will give us a good look at how their recovery path is going.

 

The next step for desktop could be on 20A process, or at worst some variation of N3. We could be looking at about 3 node jump on desktop so it better be good.

That naming convention seems inherently designed to obfuscate the lithography.

 

Until 'Intel 4' its not technically '7nm'.

 

What is Intel 7? The Raptor Lake architecture explained | Trusted Reviews

[starsmine]

8 minutes ago, Agall said:

That naming convention seems inherently designed to obfuscate the lithography.

 

Until 'Intel 4' its not technically '7nm'.

 

What is Intel 7? The Raptor Lake architecture explained | Trusted Reviews

The naming is supposed to communicate node improvement (vs the previous node from the same fab), honestly having comparable nodes between fabs the same number is better. The fact that intel 10nm is better then TSMC 7nm is not obvious, and confused nerds like us and consumers alike. renaming 10nm superfin+ (or 10nm+++ for my entertainment) to intel 7, and the old 7nm to intel 4 which best compares tech wise to TSMC quarter node of 4nm is better then having to know 10 is better than 7 (when the nodes worked).

intel also changing to amstrong when they move to gaafet is also convenient. 
The fact that Samsung 3nm and TSMC 3nm are going to be so fundamentally different makes communication hard, as TSMC does not move to gaafet until 2nm (and intel with 20A) 

[Agall]

6 minutes ago, starsmine said:

The naming is supposed to communicate node improvement (vs the previous node from the same fab), honestly having comparable nodes between fabs the same number is better. The fact that intel 10nm is better then TSMC 7nm is not obvious, and confused nerds like us and consumers alike. renaming it to intel 7, and the old 7nm to intel 4 which best compares tech wise to TSMC quarter node of 4nm is better then having to know 10 is better than 7 (when the nodes worked).

intel also changing to amstrong when they move to gaafet is also convenient. 
The fact that Samsung 3nm and TSMC 3nm are going to be so fundamentally different makes communication hard, as TSMC does not move to gaafet until 2nm (and intel with 20A) 

Also where my point stands, since the architectures aren't 1:1 comparable unless done on the same process. Where my "if Intel decided to scale the 14900k down to TSMC 5nm it would be a blood bath" comes in. If a ~10nm lithography is competitive with 5nm (6nm I/O die), Intel is definitely doing something right in the design and microarchitecture, my assumption being a limitation of MCM.

 

We would be able to see the actual architectural limitations/disadvantages to MCM, which we haven't seen a direct-ish comparison to since Ryzen 1st generation. A monolithic die should always be better, especially on a serial processor, and the fact that the 14900k beats out the 7950x is impressive (noting both CPUs will draw 250W under a full core load, but 7950x does beat it at 105W likely because of the smaller lithography). Even if we had that comparison, AMD would likely win out because of cost, since the MCM 5nm/6nm architecture will inevitably be cheaper to manufacture. Although Intel is supposedly moving to MCM as well, so we'll see.

 

3D v-cache is really the saving grace for AMD, since none of us would be talking about how power efficient and powerful the 5800x3D or 7800x3D is for gaming without it. We'd likely have written off AMD on the high end, since the 7700x vs 13600k at launch or worse, the 7700x vs 14600k is pretty cut and dry pointing towards Intel. AMD definitely gets a lot more sales than they should with the promise of future drop in AM5 CPU upgrades, which is unlikely to be even close to as good as AM4 was in that front.

 

Bear in mind, I'm a HUGE fan of AM4/AM5 overall and 3D v-cache, being a 5800x3D and 7950x3D owner and regularly do nuanced tests between Zen4 and Zen4 3D. That doesn't mean it doesn't have its faults, like how often we have to explain the thermodynamics of the chips with how hot they runs. Also personally owning both a 3600, 3950x, and 7600 system while subsequently only owning two LGA1700 systems. At work having a 12400, 7600, and 7950x build.

[starsmine]

16 minutes ago, Agall said:

Also where my point stands, since the architectures aren't 1:1 comparable unless done on the same process. Where my "if Intel decided to scale the 14900k down to TSMC 5nm it would be a blood bath" comes in. If a ~10nm lithography is competitive with 5nm (6nm I/O die), Intel is definitely doing something right in the design and microarchitecture, my assumption being a limitation of MCM.

 

We would be able to see the actual architectural limitations/disadvantages to MCM, which we haven't seen a direct-ish comparison to since Ryzen 1st generation. A monolithic die should always be better, especially on a serial processor, and the fact that the 14900k beats out the 7950x is impressive (noting both CPUs will draw 250W under a full core load, but 7950x does beat it at 105W likely because of the smaller lithography). Even if we had that comparison, AMD would likely win out because of cost, since the MCM 5nm/6nm architecture will inevitably be cheaper to manufacture. Although Intel is supposedly moving to MCM as well, so we'll see.

 

3D v-cache is really the saving grace for AMD, since none of us would be talking about how power efficient and powerful the 5800x3D or 7800x3D is for gaming without it. We'd likely have written off AMD on the high end, since the 7700x vs 13600k at launch or worse, the 7700x vs 14600k is pretty cut and dry pointing towards Intel. AMD definitely gets a lot more sales than they should with the promise of future drop in AM5 CPU upgrades, which is unlikely to be even close to as good as AM4 was in that front.

 

Bear in mind, I'm a HUGE fan of AM4/AM5 overall and 3D v-cache, being a 5800x3D and 7950x3D owner and regularly do nuanced tests between Zen4 and Zen4 3D. That doesn't mean it doesn't have its faults, like how often we have to explain the thermodynamics of the chips with how hot they runs. Also personally owning both a 3600, 3950x, and 7600 system while subsequently only owning two LGA1700 systems. At work having a 12400, 7600, and 7950x build.

Intel also regularly did node shrinks
(Westmere, Ivybridge, Broadwell), that was the whole tiktok thing. 
You don't get better IPC doing this, just better performance per watt (and also harder to cool cause denser cores) (and for intel, knowledge and experience with the node before doing a new arch with a refined version of the node)
So anyone arguing a node shrink would be a bloodbath is just not thinking things all the way through, regardless of how nodes compare. 

Node shrinks don't give free performance... well they do, but not IPC. A node shrink allows more transistors per mm, OR more clocks for same power, OR less power at the same clock. 

None of that is IPC other than more transistors per mm, which isn't a shrink, but a new architecture. Sure shrinking Raptor Lake to intel 4 would let Raptorlake hit 6.5ghz at the same 300 wattage if you are not thermal limited (which you would be, more than now) Or hit 6ghz at like 250W instead of 300W. But I doubt either of those would make anyone argue that intel is now having a bloodbath. (but this is why Zen 2 shrunk to 6nm is as competitive as it is)

[sazrocks]

 

The laptop that the AMD chip is in is $350:
https://store.acer.com/en-us/a315-24p-r2sc-us

 

But the laptop Intel used for its chip is $750???

https://www.lenovo.com/us/en/configurator/cto/index.html?bundleId=82YNCTO1WWUS1

 

What on earth is this comparison?

[filpo]

4 minutes ago, sazrocks said:

 

The laptop that the AMD chip is in is $350:
https://store.acer.com/en-us/a315-24p-r2sc-us

 

But the laptop Intel used for its chip is $750???

https://www.lenovo.com/us/en/configurator/cto/index.html?bundleId=82YNCTO1WWUS1

 

What on earth is this comparison?

Thanks for bringing this up, I'll put it in my post, nice spot

 

But it's horrible that they did this comparison with such widely different (in terms of price at least so most likely in other specs too) laptops

 

Also it looks like they benchmarked the AMD chip about 4 months before benchmarking the intel system so driver and chipset updates could have improved its score a bit (albeit not by much of course)

[Agall]

11 minutes ago, starsmine said:

Intel also regularly did node shrinks
(Westmere, Ivybridge, Broadwell), that was the whole tiktok thing. 
You don't get better IPC doing this, just better performance per watt (and also harder to cool cause denser cores) (and for intel, knowledge and experience with the node before doing a new arch with a refined version of the node)
So anyone arguing a node shrink would be a bloodbath is just not thinking things all the way through, regardless of how nodes compare. 

Node shrinks don't give free performance... well they do, but not IPC. A node shrink allows more transistors per mm, OR more clocks for same power, OR less power at the same clock. 

None of that is IPC other than more transistors per mm, which isn't a shrink, but a new architecture. Sure shrinking Raptor Lake to intel 4 would let Raptorlake hit 6.5ghz at the same 300 wattage if you are not thermal limited (which you would be, more than now) Or hit 6ghz at like 250W instead of 300W. But I doubt either of those would make anyone argue that intel is now having a bloodbath. (but this is why Zen 2 shrunk to 6nm is as competitive as it is)

It's not always the case, sure, I remember dismissing the 6700k because of how trivial the performance increase was over the 4790k, then all the subsequent generations till maybe 10th generation. That's demonstrated through benchmarks as well. Even if the change is small, its still an upgrade, and that's assuming Intel hasn't learned over the last 10 generations on shrinking nodes. Even a 10% IPC improvement would put Intel squarely in the lead, since its far closer than I expect given the technological gap between Intel and AMD. They've at least learned the hard way that scaling up a node isn't a good idea with 11th generation  

[porina]

1 hour ago, Agall said:

Until 'Intel 4' its not technically '7nm'.

From your link: 

Quote

The change from nanometer to this naming convention helps to communicate the fact that Intel’s 10-nanometer process actually offers the same transistor density as 7-nanometer chips made by TSMC.

If you accept N7 as is, why not Intel 7? That's all the rebranding is - to match the number game by TSMC and Samsung.

 

7 minutes ago, Agall said:

It's not always the case, sure, I remember dismissing the 6700k because of how trivial the performance increase was over the 4790k, then all the subsequent generations till maybe 10th generation.

For Prime95 I found Skylake offered +14% IPC over Haswell. Cinebench R15 saw >+11%. So if you had a top end Haswell there wasn't much point in replacing it, but it was still a measurable step forwards.

[Sauron]

Kind of rich coming from Intel considering they often do this themselves on lower end SKUs... and being old is not as bad as being shit, which is what Intel should be concerned about right now.

 

Kinda odd to target competitors directly and by name like this too, even if you leave the bad taste aside... it's kind of begging for a law suit. It probably wouldn't be that hard to convince a judge that the comparison is at the very least misleading. Usually companies doing this kind of advertisement at least leave the name out, even though it's usually obvious who they're talking about (doubly so in a 2 player market...).

[GOTSpectrum]

5 hours ago, porina said:

For Prime95 I found Skylake offered +14% IPC over Haswell. Cinebench R15 saw >+11%. So if you had a top end Haswell there wasn't much point in replacing it, but it was still a measurable step forwards.

I miss the days before everything was a lake... bring back differing uArch names! 

[Taf the Ghost]

I see someone in Intel Marketing had to come up with something. Fairly impressive how bad it is, but, well, they got ad copy out!

[05032-Mendicant-Bias]

14 hours ago, TVwazhere said:

This isnt the first time Intel has taken a smear campaign tactic to competition. Anyone remembered "glues together CPU dies"?

 

https://www.pcgamer.com/intel-slide-criticizes-amd-for-using-glued-together-dies-in-epyc-processors/

 

I find it really distasteful when manufacturers attack each other in marketing material. The best marketing is the one that compares a product with the older products from the same company, Apple for all their flaws are much better at this.

  

13 hours ago, Agall said:

Until 'Intel 4' its not technically '7nm'.

When Intel was ahead in manufacturing, TSMC and others started skipping numbers, then when intel was behind in manufacturing, intel started skipping numbers. The names of manufacturing processes have been pretty much meaningless for a while.

 

You can spill lots of links comparing names, at the end of the day, if you compare transistor densities, there isn't much of a gap. Intel is behind in power efficiency to TSMC.

 

Intel 10 compares with TSMC 7

Intel 7 compares with TSMC 5

Intel 4 is supposed to compare with TSMC 3

 

There is a lot of tuning done that makes the numbers hard to compare, like the number of dummy transistors there just to make the lithography work, or the tuning of the transistor sizes and thresholds even within the processes. E.g. Apple chose to tune the M chips for power efficiency even for desktop, so those humongous dies lags in performance compared to a Nvidia chip of similar size tuned for performance.

 

The LMC metric is supposed to make the comparison more apple to apple, but I see no signs of adoption yet.

[Lunar River]

Intel you make good stuff, you don't need to fling mud at your competitor, talk about what good stuff YOU do...this isn't a US election campaign run

[LAwLz]

On one hand, I have been saying AMD should stop their bullshit naming schemes forever. It really feels like they named things in a way to confuse people and it's far from the first time AMD has done this in recent times.

 

On the other hand, it is a very bad look in my eyes when companies try and throw mud at each other. I much prefer when a company talks about how good they are, instead of talking about how bad their competitor is.

Intel has a lot of good things going for them right now. In a lot of cases, I'd even say they are better buys than AMD (although it heavily depends on regional pricing). They should focus on that instead.

[RejZoR]

That's pretty cringe. What oozes confidence the most is when you don't mention competition AT ALL. You mention and compare your own stuff only to yourself and you do this to your best products while showing impressive performance increases. Apple did that several times and I think that's the biggest big balls way. AMD did it too during X3D launches and 7000 series iirc.

 

It's kinda unwritten rule to not badmouth the competition. And Intel isn't getting that memo for whatever reason. They did it for the "glued dies" thing when AMD launched chiplet Ryzens and they are doing it again. Intel, WHY? JUST WHY? Shows you're weak and have zero confidence in your products if you have to go and bash on competition this way.

[starsmine]

20 minutes ago, RejZoR said:

That's pretty cringe. What oozes confidence the most is when you don't mention competition AT ALL. You mention and compare your own stuff only to yourself and you do this to your best products while showing impressive performance increases. Apple did that several times and I think that's the biggest big balls way. AMD did it too during X3D launches and 7000 series iirc.

 

It's kinda unwritten rule to not badmouth the competition. And Intel isn't getting that memo for whatever reason. They did it for the "glued dies" thing when AMD launched chiplet Ryzens and they are doing it again. Intel, WHY? JUST WHY? Shows you're weak and have zero confidence in your products if you have to go and bash on competition this way.

I strongly appreciate it when companies do talk about their competition instead of this amorphous other that they are afraid of giving attention to. When you have winners its easy to do it respectfully and gain mind share and it shows confidence. 

Its just when you say things that are not true OR are intentionally obfuscating what you are doing that its in bad taste. 

Like if Intel showed in slides that per watt or per dollar they perform better sure. But here they went and showed if you spend 700 on an intel you beat a 350 dollar AMD.... They could have shown benchmarks of actual equivalent hardware price-wise, not just compare the worst R5 with the best i5. Better R5s exist, worst I5s exist. 

They could have made a point that it was an R5 which is named that way because it's supposed to compete with i5s, but the fact is other R5s do compete and win.  There was a way Intel could have communicated this and not come off as slime. The fact is its a small blip in the AMD line up.
 
They could have pointed out that with the very wide range of R5s on the 7000 line its hard to know what you are getting when it all has an R5 sticker without all this bullshit around it. But then again, I dont think they could have done that without showing pheonix r5s kicking its ass with its orange sticker. But even then, arguing that R5 zen2 chips should have been called R3 still is not a strong argument. (like keep AMD 7320, but change the AMD 7520 to a AMD 7325 or something.

[RejZoR]

16 minutes ago, starsmine said:

I strongly appreciate it when companies do talk about their competition instead of this amorphous other that they are afraid of giving attention to. When you have winners its easy to do it respectfully and gain mind share and it shows confidence. 

Its just when you say things that are not true OR are intentionally obfuscating what you are doing that its in bad taste. 

Like if Intel showed in slides that per watt or per dollar they perform better sure. But here they went and showed if you spend 700 on an intel you beat a 350 dollar AMD.... They could have shown benchmarks of actual equivalent hardware price-wise, not just compare the worst R5 with the best i5. Better R5s exist, worst I5s exist. 

They could have made a point that it was an R5 which is named that way because it's supposed to compete with i5s, but the fact is other R5s do compete and win.  There was a way Intel could have communicated this and not come off as slime. The fact is its a small blip in the AMD line up.
 
They could have pointed out that with the very wide range of R5s on the 7000 line its hard to know what you are getting when it all has an R5 sticker without all this bullshit around it. But then again, I dont think they could have done that without showing pheonix r5s kicking its ass with its orange sticker. But even then, arguing that R5 zen2 chips should have been called R3 still is not a strong argument. (like keep AMD 7320, but change the AMD 7520 to a AMD 7325 or something.

I didn't mind AMD's straight to the point Cinebench showcase of having a lot less seconds of run time than top end Intel's product. It's like this is our performance and this is theirs. And that was it. That's perfectly ok. But the crap Intel is pulling here is just lame AF and instantly makes me not want to buy Intel. And I owned Intel CPU's most of my life with several AMD in between and currently on AMD after a very long time of prior Intel's. This is NOT how you win over customers.

 

And I wouldn't mind if Intel did the same by just showing honest numbers and showing they are just better. Now that people are digging into this cringefest, it seems not even Intel is being honest about it which is extra bad.

[RejZoR]

2 hours ago, LAwLz said:

On one hand, I have been saying AMD should stop their bullshit naming schemes forever. It really feels like they named things in a way to confuse people and it's far from the first time AMD has done this in recent times.

 

On the other hand, it is a very bad look in my eyes when companies try and throw mud at each other. I much prefer when a company talks about how good they are, instead of talking about how bad their competitor is.

Intel has a lot of good things going for them right now. In a lot of cases, I'd even say they are better buys than AMD (although it heavily depends on regional pricing). They should focus on that instead.

Their desktop models are actually fine, pretty straight forward, but their mobile ones are absolutely horrid. From last gen architectures being named as current gen by numbers down to just straight up confusing numbers entirely. And they didn't used to be this way which is annoying. I still have my old Ryzen 5 2500U powered laptop and those were pretty straight forward numbered.

[Fnige]

The naming scheme for laptops that AMD has is shit but I honestly cant think of a better naming scheme without going down the path of "x cpus shouldnt exist". 

 

The only actual thing I can think of that would make things better would be some way to better tell the user "oh this is a zen 2 part". Like maybe having some indicator on the sticker or something. But then I dont think the average user knows what a "Zen 2" means so...

[Beerzerker]

Sigh....
You know - I miss the good old days back when AMD was gluing cores together.......

[Forbidden Wafer]

I think it sucks to re-release old products under new names, but you know who did the same? Freaking Intel...

I don't have words to describe how bad this is...

The only thing I can say is that marketing people responsible for this should resign, rather than dragging Intel further down.

[Recharged]

I think AMD's naming schemes are dumb just like Intel's, but this is a rather low blow from Intel.

 

There is clearly no reason to do this if AMD was no threat to Intel at all, completely nullifying the idea that the fab process is what is making AMD faster.

 

It's OK I am sure the same people will think the 14900K is so much greater than 13900K too.

 

38 minutes ago, Beerzerker said:

Sigh....
You know - I miss the good old days back when AMD was gluing cores together.......

Pentium D was my first glued together CPU.

[Rarity]

probably thier bottomline being affected that why this kinda stuff comes out .. 

[HenrySalayne]

If I were AMD, I would call all upcoming Zen5 CPUs 16th gen to battle any possible confusion with older architectures.

[jagdtigger]