Intel CEO Pat Gelsinger Declares 'AMD's Lead Is Over' After Alder Lake, Sapphire Rapids

[leadeater]

33 minutes ago, porina said:

Don't forget on Intel, (peak) power consumption and therefore power efficiency is a choice for the system builder. Just most enthusiasts leave it on unlimited, trading more performance for worse efficiency. AMD doesn't give you that choice without voiding warranty.

I don't think it matters though, you can uncap Ryzen CPUs but they won't use any more power because they are temperature clock limited so you have to go down real cold to get any increases in power usage. Zen/TSMC 7nm is much like the fuel flow restrictor used in motor racing.

[Lightwreather]

7 minutes ago, Master Disaster said:

Agree to disagree, macOS jumping to ARM has already started the ball rolling. I don't see RISC making much of an impact in the server space but for us normies at home RISC will become our standard platform by the end of this century. Why wouldn't it? Apple have shown it can be competitive and RISC devices are easier and cheaper to manufacture than X86 plus X86 is reaching its limits anyway.

 

Even after everything that's happened with Linux over the past 5 years there are still some notable devs who refuse to port anything over to it, M1 is a first gen platform and it already has most of the missing stuff built and ready.

 

This is why Nvidia want to buy ARM and both Intel & AMD have stated their interest in creating ARM (or RISC V) platform, they all see the writing on the wall.

If indeed x64 is to die, then I forsee a slow decline not one that is sudden adn exponential.

[mr moose]

29 minutes ago, porina said:

I don't think Intel ever got complacent. Their 10nm was broken for a long time, and that dictated what they could or couldn't do in that timeframe. If you have lemons, make lemonade.

 

 

Agree, complacency rarely happens at this level.  Making sure you are at the bleeding edge of technology development is a kin to olympic athletes making sure they are at peak fitness.  Neither one risks taking a day off because the result is major losses.

 

Would you say Intel are planning to put more resources into support for system builders,  product supply and general make their products more appealing over all rather than relying on raw performance in order to give their products and advantage?  Because that sounds like a very logical move given how wild the industry is right now.  

 

 

[porina]

4 minutes ago, leadeater said:

I don't think it matters though, you can uncap Ryzen CPUs but they won't use any more power because they are temperature clock limited so you have to go down real cold to get any increases in power usage. Zen/TSMC 7nm is much like the fuel flow restrictor used in motor racing.

Point was, often when looking at power consumption numbers they are presented without sufficient context. Like AVX-512 peak power on Intel can be astronomical, but it is also doing much more work in that state, and efficiency wise isn't much different than without. AMD limit their "stock" operation to a somewhat efficient (or less inefficient) area than Intel.

 

I'm not saying that Intel desktop products up to now are more power efficient than Zen 2/3, as they're not. But the difference is nowhere near as big as only looking at peak power would suggest.

 

2 minutes ago, mr moose said:

Would you say Intel are planning to put more resources into support for system builders,  product supply and general make their products more appealing over all rather than relying on raw performance in order to give their products and advantage?  Because that sounds like a very logical move given how wild the industry is right now.  

The original interview did cover some of these points, so I'd suggest going over that rather than me mis-remembering or presenting an interpretation that might be wrong.

[leadeater]

6 minutes ago, J-from-Nucleon said:

If indeed x64 is to die, then I forsee a slow decline not one that is sudden adn exponential.

Makes no difference what anyone does other than a single company, Microsoft. Microsoft and Windows is the linchpin here, without any changes on this front x86 is here to stay and largely the market share will be around what is is now. Microsoft really doesn't want to make any changes either otherwise the backlash is going to be immense, it's not as simple as making an efficient translation layer like Apple did, Windows ecosystem is just more complex and littered with mines and unexploded bombs.

 

And the next comment about Linux or Mac OS taking over I'll add a penny to the entire ocean of pennies from the utterances of this in the past with nothing ever changing.

 

 

[MultiGamerClub]

17 hours ago, givingtnt said:

Not like they would decimate the market.. right?

[Blademaster91]

1 hour ago, Mark Kaine said:

This is exactly what this means:

Their big little won't be able to compete in certain (or all) work loads and that's why they don't want to have these comparisons. Otherwise this statement (did intel actually say this?) doesn't even make sense. 

 

 

Btw i understand why, this is like a new start for intel, a completey new concept which long term might evolve to something that "can" beat the competition,  but,  its actually really bad PR. Depends on what AMD is doing , how bad. Maybe AMD has suddenly similar ideas... who knows. 

I think that makes sense, Intel has been kind of pushing away from benchmarks they used to put a lot of marketing into, even saying benchmarks don't matter.

But I am hopeful Alder Lake can at least be on par with Zen 3, that would at least drive prices down. I think Alder Lake is interesting because Intel finally has something new instead of adding more cores onto a Skylake based CPU, though I might just upgrade to Ryzen 5000 instead as DDR4 is cheaper. I have seen rumors that Zen 4 is going to have big & little cores, and every Zen 4 cpu might have an iGPU which would be nice as I don't expect GPU prices to go down to what they were with the RTX 2000 series.

1 hour ago, porina said:

Don't forget on Intel, (peak) power consumption and therefore power efficiency is a choice for the system builder. Just most enthusiasts leave it on unlimited, trading more performance for worse efficiency. AMD doesn't give you that choice without voiding warranty.

Most leave it on because board manufacturers leave Intel boosting with unlimited power on by default, which really doesn't help and Intel does have the choice to tell board manufacturers to stop for efficiency reasons. I don't really see the point in increasing power limits on AMD, their turbo boosting is good enough, you actually need to lower EDC and PPT if you want the best performance out of a Ryzen 5000 cpu.

24 minutes ago, mr moose said:

 

Agree, complacency rarely happens at this level.  Making sure you are at the bleeding edge of technology development is a kin to olympic athletes making sure they are at peak fitness.  Neither one risks taking a day off because the result is major losses.

 

Would you say Intel are planning to put more resources into support for system builders,  product supply and general make their products more appealing over all rather than relying on raw performance in order to give their products and advantage?  Because that sounds like a very logical move given how wild the industry is right now.  

 

 

If you mean complacency in terms of architecture or process nodes, Intel was ahead, but Intel was complacent with advancements on mainstream until 8th gen, until Ryzen came out.

[leadeater]

4 minutes ago, porina said:

Point was, often when looking at power consumption numbers they are presented without sufficient context. Like AVX-512 peak power on Intel can be astronomical, but it is also doing much more work in that state, and efficiency wise isn't much different than without. AMD limit their "stock" operation to a somewhat efficient (or less inefficient) area than Intel.

 

I'm not saying that Intel desktop products up to now are more power efficient than Zen 2/3, as they're not. But the difference is nowhere near as big as only looking at peak power would suggest

True, my comment was more specifically about Ryzen power limit than the parts about efficiency. AMD's limit is less of a limit and more of a safeguard and even when you remove that safeguard (PPT) the inbuilt in archecture boost and power controls end up imposing the same limit as the junction temperatures across the sensors in the die aren't low enough for the algorithm to allow higher clocks and thus higher power. The PPT limit prevents spikes in power and Precision Boost 2 prevents sustained power increase due to those temperature sensors.

 

Precision Boost 2 is why there isn't a fixed "All Core" or "per Thread/Core" boost table like Intel has, it's just whatever it is based on workload. It's "better" but also less fun for tinkerers.

[leadeater]

5 minutes ago, Blademaster91 said:

If you mean complacency in terms of architecture or process nodes, Intel was ahead, but Intel was complacent with advancements on mainstream until 8th gen, until Ryzen came out.

Complacent only in terms of ramming more cores in to dies and increasing the size and cost of them. Intel never stopped archecture improvements or process improvements. Intel ran in to problems because architecture improvements are planned along side process improvement roadmaps and the two together become part of the architecture design so if the process fails to materialize then Intel has an useable architecture because it can't be manufactured on anything other than what it was designed for. Rocket Lake was a significant porting effort to get it on to 14nm.

 

Intel at any point could of added more cores to their products, the Ring Bus and archecture could and were easily scaled out (aka Xeon). It's just there was little demand for more cores, the increased cost, or application utilization. There's a reason AMD waived the Blender Flag all day every day for Ryzen 1000 as well as a side flag of gaming and game streaming on a single system, these were the only things close to relevant to the more general consumer that could show a benefit to more than 4 cores at the time.

 

Remember it's not as if Intel gave as more cores on the consumer platform for no increased cost.

[mr moose]

55 minutes ago, Blademaster91 said:

If you mean complacency in terms of architecture or process nodes, Intel was ahead, but Intel was complacent with advancements on mainstream until 8th gen, until Ryzen came out.

I mean complacency as in intentionally not trying harder because they didn't need to.    The type of character that becomes CEO/director or even upper management of any company is a workaholic 16+ hour day goal oriented psychopath,  complacency is the devil to them and they do not tolerate it in their organization.   People seem to think that poor performance is only caused by complacency. 

 

 

[RejZoR]

30 minutes ago, comander said:

One caveat. 

This is generally true for 1CCD parts. It's generally not true for 2CCD parts. 

Anecdotally speaking -0.1VID helped my CPU boost higher because it was no longer wattage limited. 

Ryzen chips are weird in terms of monitoring how they behave. Voltages and clocks are so far all over the place I don't even know which software to believe as Task Manager, CPU-Z, HWInfo and Ryzen Master all show different values. And same goes for voltages. I've undervolted it and gave it headroom all the way to 5GHz. I've had cases where it was running at 5GHz flat in games and benchmarks and yet it was scoring less than with stock 4.85 GHz. Even in single threaded benchmarks like R23 and CPU-Z. How does that make any kind of sense!?

[IkeaGnome]

2 hours ago, Mark Kaine said:

did intel actually say this?

Yeah, 

https://www.extremetech.com/computing/311275-intel-doesnt-want-to-talk-about-benchmarks-anymore

Quote

We should see this moment [the COVID-19 pandemic] as an opportunity to shift our focus as an industry from benchmarks to the benefits and impacts of the technology we create. The pandemic has underscored the need for technology to be purpose-built so it can meet these evolving business and consumer needs.

 

[da na]

8 hours ago, WolframaticAlpha said:

you guys are making phenom-enal puns

And yours are also Optemal

Idk that was bad

[CTR640]

On 10/5/2021 at 5:56 PM, Mel0nMan said:

Epyc joke

Indeed. In the end after the battle, one will get the victory and the other will end up drinking Whiskey somewhere in a bar near a Lake.

[Falkentyne]

12 hours ago, comander said:

1. ADL is 5 years of MANUFACTURING advancements and the first time Intel has had a core released on desktop that didn't have latency issues due to complications involving manufacturing. 

2. Conroe had a 90% IPC uplift over the Pentium 4 (the Pentium M and AMD's Athlon 64 were around 70% faster than the P4, the Athlon XP and Pentium III were around 30% faster than the P4 on a per clock basis). On launch, the lowest end, $180ish 1.83GHz Core 2 Duo was practically tied with the $1000 3.73Ghz Pentium Extreme Edition in benchmarks. https://www.anandtech.com/show/2045 - while the 2.93 GHz Core 2 Extreme Edition was pretty fast at stock what REALLY separated it was that you could conceivably take a $300ish CPU (2.4GHz) and push it to ~3.8GHz on a budget board with relatively modest cooling. The PXE would only do something like 4.7GHz or so. OC vs OC the Core 2 was 50+% faster. Also a few months later a quad core version came out.
3 RKL isn't quite 19%, I want to say it was closer to 15% on average. The figure intel put out was UP TO 18% faster... but
4. The big draw with ADL is that it brings around +20% over RKL in ST and around +100% over RKL on MT. 

We haven't had a launch that doubled MT performance and gave us a 20% ST performance gain AND drastically improved power consumption since... around 2006-2007ish. Which was conroe. 

 

The launch of RKL and CML and CFL and SKL and BDW and HSW and IVB and even SB were all kind of modest as you could make arguments that an OCed version of their predecessor did everything about as well. 

Conroe was... big. 
 

There's a VERY VERY real possibility that ADL will do THAT. Make it so that there's virtually no good reason to consider RKL or CML. It'll depend on pricing and availability but... 


Why would you want to buy a CPU with half the MT performance and 20% lower ST performance that sucks around 2x the power?

Oh you're right.  I meant double the IPC or close to it.  not 50%.

I remember now.  I had a Pentium Northwood (EE, because I was stupid and wasted extremely expensive amounts of money back then) at 3.7 ghz and compared it to my Core 2 duo (before I wasted even MORE amounts of money on a Core 2 QX9650 extreme edition--OH DID I TELL YOU MY CORE 2 DUO WAS AN X6800 EXTREME EDITION TOO?  If i had that money back, I could have just bought a car...), and in a 100% CPU limited game back then, my FPS was almost double.

 

I think I remember I had to set my core 2 duo to 2.0 ghz, which sort of matched the Pentium 4 @ 3.7 ghz in FPS.  I forgot what I was running.  Maybe it was some sort of SNES emulator, I don't remember anymore...But yes you're right.

[soldier_ph]

Quote

"AMD's Lead Is Over"

Haha, no.

[XGoodGuyFitz]

On 10/5/2021 at 1:58 PM, valdyrgramr said:

  The 3090 is not a gaming card it's a workstation card they label as basically a Titan replacement for gamers.   Unless you're Linus, I don't see how the 3090 is one.

i dont know what you mean, the 3090 is for school man. 

[Mark Kaine]

Tbh, this sounds incredibly weak, declaring win before you delivered the goods is something intel would do. It just shows they have not overcome their hybris yet, more mutilation needed, huh...