Intel’s bad quarter

[dada_dave]

4070 ti in 4K is 10% slower than 7900xt and 8% slower in 1440p.

MSI GeForce RTX 4070 Ti Suprim X Review

With the Suprim X, MSI offers an impressive RTX 4070 Ti custom design. Thanks to its huge quad-slot cooler, and an excellent fan profile, the card runs virtually inaudible, even under full load. Overclockers can manually increase the power limit up to 365 W, more than on any other RTX 4070 Ti.

www.techpowerup.com

This link shows the relative performance.

Interesting slightly higher boost than clock, heavy duty cooling but otherwise “standard” wrt memory. Not sure then, I would’ve expected a higher difference at 4K though the closeness at 1440p given the above doesn’t surprise me. The AMD GPU only has GDDR6 rather than 6X but overall bandwidth is good. My guess the games that AMD has the biggest leads are indeed the ones that impact memory the hardest but not every game tested does so and the average reflects that.

 

[Colstan]

Shouldn't the comparison be against small form factor PCs sold by Dell, HP and other PC computer manufacturers. Not individual parts that make up a PC.

No.

Please reread the context:

I've had a conversation about this with a friend who is very much non-tech. He refers to a screenshot as "print screen" and will get lost if I go off even slightly into the weeds. However, I appreciate his feedback, because his ignorance is the most unbiased opinion I can get. There's a lot of tribalism in tech, so I like to hear what he has to say.

I was comparing the price of the 4080 and the cost of an entire M2 Pro Mac mini to each other so that my non-tech friend could understand my argument, which is simple. Building a PC has gone up in relative cost when compared to purchasing the average Mac. My friend doesn't care about gewgaws and doohickies, but he does care about money. This is an analogy that he does understand and can wrap his mind around.

Because he is an unbiased source, and was able to put it into perspective for me, I concluded that PC parts are over-priced, and therefore not worthy of my consideration. It's as simple as that.

 

[exoticspice1]

I was comparing the price of the 4080 and the cost of an entire M2 Pro Mac mini

Why? Why not something more comparable in perfromance to a M2 Pro GPU(6.8 TFLOPS) . Like a RTX 3050( 9.1 TFLOPS).
The RTX 4080 has 48TFLOPS and RT cores its in a different league compared to the M2 Pro GPU which doesn't even Ray Tracing cores.

These serve different users and needs. A person who intends to buy a RTX 4080 will NEVER look at a Mac mini and vice-versa

Building a PC has gone up in relative cost when compared to purchasing the average Mac

This what I am confused about. Comparing a whole product that costs the same as a Mac mini is not a good way to make judgement. Why not compare an average PC build that is around the same performace ratio as the Mac mini.

Because he is an unbiased source, and was able to put it into perspective for me, I concluded that PC parts are over-priced, and therefore not worthy of my consideration. It's as simple as that.

The only overpriced market right now in the PC space is the high end GPU market. The other markets such as CPU, RAM and Storage markets are quite competitive.

Oh well, you made your mind and but I will a provide an example that overall PC build is still cheaper than a base Mac Studio(Frankly, I overspecced the GPU).

Part List - AMD Ryzen 7 7700, GeForce RTX 4070 Ti, Fractal Design Meshify C ATX Mid Tower - PCPartPicker

Part List - AMD Ryzen 7 7700, GeForce RTX 4070 Ti, Fractal Design Meshify C ATX Mid Tower

pcpartpicker.com

The price includes MS Windows.

 

[Colstan]

Why?

Because I was explaining it to a friend who knows nothing about technology. I was using simple price points. I could have been talking about canned beans, lizards, or soda pop. It was about the economics of building a PC, not the specs.

This what I am confused about. Comparing a whole product that costs the same as a Mac mini is not a good way to make judgement.

For my friend, it was perfect for his understanding. He didn't realize how much a PC graphics card, any graphics card, costs. This put it into perspective, even though he still has no idea what a graphics card actually is.

The only overpriced market right now in the PC space is the high end GPU market. The other markets such as CPU, RAM and Storage markets are quite competitive.

Depends upon one's viewpoint. I think all PC gear is too expensive.

Oh well, you made your mind and but I will a provide an example that overall PC build is still cheaper than a base Mac Studio(Frankly, I overspecced the GPU).

I appreciate your enthusiasm, but I think we are arguing different things. My interest in the PC side of the pond is purely based upon curiosity. It's the same thing as watching the circus so that I can gawk at the carnies. It's also the same reason that I give every new version of Windows a shot, even though it's like tossing rocks at the village idiot.

I'm not deriding your personal interest in PC products or owning such devices. We simply have different priorities and opinions on the subject, which is fine, because the world would be awfully boring if we all thought the same way.

 

[mr_roboto]

That’s not to say Apple will or should be immune from criticism: if I remember right, the M2 SOC got criticized for effectively cutting memory bandwidth in half on some of the tiers by using one double sized module instead of two smaller ones.

It was SSD flash, not DRAM. Base model M2 Macs with the smallest 256GB SSD have half the sequential R/W throughput compared to everything larger.

IMO, the criticism for this downgrade was often a bit overwrought. To read some of it, you'd think it was the end of the world, but once you're over about 1GB/s sequential (which base model M2 Macs still provide), light-duty users (most of the people who are buying base-spec Macs) will never notice more.

 

[dada_dave]

It was SSD flash, not DRAM. Base model M2 Macs with the smallest 256GB SSD have half the sequential R/W throughput compared to everything larger.

IMO, the criticism for this downgrade was often a bit overwrought. To read some of it, you'd think it was the end of the world, but once you're over about 1GB/s sequential (which base model M2 Macs still provide), light-duty users (most of the people who are buying base-spec Macs) will never notice more.

Yeah overwrought does seem to be the baseline for a lot of tech complaints or even accolades , the stridency either way only occasionally justified. And if you were writing about MR that’s even a tame description: it’s typically apocalyptic or is it apoplectic? … end of the world as you said

 

[Citysnaps]

Yeah overwrought does seem to be the baseline for a lot of tech complaints or even accolades , the stridency either way only occasionally justified. And if you were writing about MR that’s even a tame description: it’s typically apocalyptic or is it apoplectic? … end of the world as you said

99% of laptop users will likely never notice the difference.

But... maybe I'm being unfair. Who knows what they do with their laptops. Perhaps they're finding the next prime number, mining Bitcoin, or decrypting secret Russian communications channels. Sounds legit to me...

 

[dada_dave]

And that’s part of @Colstan ’s point: Apple by designing the whole widget can put in downgrades and offer “upgrades” on other features to protect margins and justify price bumps that other companies would struggle with. Now occasionally those upgrades are not great, the price bumps are still high (iPhones have been increasing in price quite a lot), and the downgrades are noticeable. But other companies just selling chip components are even more at the mercy of increasing node costs with no ability to either maintain margins or stop the concomitant price spikes. I think that’s a fair statement.

As for Intel itself or even more narrowly x86 I’m not in a position to declare them dead or going the way of IBM - such things may happen, but those predictions are too big and about an event that’s too far away to be useful. Its primary competitors AMD and ARM have structural problems of their own. The only thing I’d say is that I’d rather be on the AMD/ARM executive team (not that I’d be allowed anywhere near) trying to solve their long term problems than Intel’s. Intel is going to go through the wringer over the next decade and will have to have an amazing product line during that time to stave off continued fall.

 

[Colstan]

And that’s part of @Colstan ’s point: Apple by designing the whole widget can put in downgrades and offer “upgrades” on other features to protect margins and justify price bumps that other companies would struggle with.

This is a company that survived the death of its genius co-founder, went through the most successful succession of power in history, became even more innovative since those uncertain times, built a war chest larger than the GDP of many nations, developed the strategic goal to control every component in their product line, ruthlessly manages the supply chain, and has the ability to pivot like a startup despite being the largest dreadnaught on the high seas.

As for Intel itself or even more narrowly x86 I’m not in a position to declare them dead or going the way of IBM

Intel is the derelict sunk down to Poseidon's watery abyss. They should cleave themselves asunder, much as AMD once did. Intel should spin off its foundry business and concentrate on design. Talos long ago lost his bronze nail, has sprung a leak, and Gelsinger isn't the man to patch him up.

 

[Cmaier]

This is a company that survived the death of its genius co-founder, went through the most successful succession of power in history, became even more innovative since those uncertain times, built a war chest larger than the GDP of many nations, developed the strategic goal to control every component in their product line, ruthlessly manages the supply chain, and has the ability to pivot like a startup despite being the largest dreadnaught on the high seas.

Intel is the derelict sunk down to Poseidon's watery abyss. They should cleave themselves asunder, much as AMD once did. Intel should spin off its foundry business and concentrate on design. Talos long ago lost his bronze nail, has sprung a leak, and Gelsinger isn't the man to patch him up.

I think the opposite. They should focus on fab and let companies that are good at design do design. There’s a low barrier to entry for design and a high one for fab. And design will increasingly move to their customers.

 

[Colstan]

I think the opposite. They should focus on fab and let companies that are good at design do design. There’s a low barrier to entry for design and a high one for fab. And design will increasingly move to their customers.

Okay, spin off the design group, and concentrate on fabs. It's the same thing, in my eyes. I framed it in terms of mythology, not strategy.

"When people see a strong horse and a weak horse, by nature they will like the strong horse."
- Osama bin Laden

(Detestable creature that he was, but not wrong, in that regard.)

 

[Colstan]

Unlike Intel's disastrous quarter, AMD slightly beat analyst forecasts.

AMD Q4 earnings beat expectations, despite slowing PC sales

Chip maker AMD (AMD) reported its Q4 2022 earnings after the bell on Tuesday, beating analysts' expectations. The report comes just days after rival Intel (INTC) announced disappointing earnings and guidance.

finance.yahoo.com

Seeking Alpha has a short article about how Intel may never see "peak PC revenue" again, as a result of competition from Apple and AMD.

Intel may never see 'peak' PC revenue again, thanks to Apple, AMD: Bernstein

The correction in the PC market is firmly underway, as evidenced by results from AMD (AMD) and Intel (INTC), the latter of which may never see its peak from the PC market again.

seekingalpha.com

Intel is getting squeezed in server and client by AMD. (Never mind the Arc GPUs being a massive money pit.)

Regarding the fruit company, when Intel lost Apple as a customer, over 10% of the computer market went with them. Unlike the Dells and HPs of the world, Apple doesn't sell cheap crap, hence Intel lost its most premium customer when the Mac transitioned to Apple Silicon. (Excluding the handful of Xeons that shipped inside of all six of the Mac Pros that Apple sold last quarter.)

A recent photograph was smuggled out of Intel's headquarters, with the engineering team attempting to decide who is to blame for the chip company's repeated failures:

 

[Cmaier]

Unlike Intel's disastrous quarter, AMD slightly beat analyst forecasts.

AMD Q4 earnings beat expectations, despite slowing PC sales

Chip maker AMD (AMD) reported its Q4 2022 earnings after the bell on Tuesday, beating analysts' expectations. The report comes just days after rival Intel (INTC) announced disappointing earnings and guidance.

finance.yahoo.com

Seeking Alpha has a short article about how Intel may never see "peak PC revenue" again, as a result of competition from Apple and AMD.

Intel may never see 'peak' PC revenue again, thanks to Apple, AMD: Bernstein

The correction in the PC market is firmly underway, as evidenced by results from AMD (AMD) and Intel (INTC), the latter of which may never see its peak from the PC market again.

seekingalpha.com

Intel is getting squeezed in server and client by AMD. (Never mind the Arc GPUs being a massive money pit.)

Regarding the fruit company, when Intel lost Apple as a customer, over 10% of the computer market went with them. Unlike the Dells and HPs of the world, Apple doesn't sell cheap crap, hence Intel lost its most premium customer when the Mac transitioned to Apple Silicon. (Excluding the handful of Xeons that shipped inside of all six of the Mac Pros that Apple sold last quarter.)

A recent photograph was smuggled out of Intel's headquarters, with the engineering team attempting to decide who is to blame for the chip company's repeated failures:

View attachment 21862

I texted that photo to my daughter yesterday with the caption “big crowd on beaker street.”

 

[dada_dave]

Unlike Intel's disastrous quarter, AMD slightly beat analyst forecasts.

AMD Q4 earnings beat expectations, despite slowing PC sales

Chip maker AMD (AMD) reported its Q4 2022 earnings after the bell on Tuesday, beating analysts' expectations. The report comes just days after rival Intel (INTC) announced disappointing earnings and guidance.

finance.yahoo.com

Seeking Alpha has a short article about how Intel may never see "peak PC revenue" again, as a result of competition from Apple and AMD.

Intel may never see 'peak' PC revenue again, thanks to Apple, AMD: Bernstein

The correction in the PC market is firmly underway, as evidenced by results from AMD (AMD) and Intel (INTC), the latter of which may never see its peak from the PC market again.

seekingalpha.com

Intel is getting squeezed in server and client by AMD. (Never mind the Arc GPUs being a massive money pit.)

Regarding the fruit company, when Intel lost Apple as a customer, over 10% of the computer market went with them. Unlike the Dells and HPs of the world, Apple doesn't sell cheap crap, hence Intel lost its most premium customer when the Mac transitioned to Apple Silicon. (Excluding the handful of Xeons that shipped inside of all six of the Mac Pros that Apple sold last quarter.)

A recent photograph was smuggled out of Intel's headquarters, with the engineering team attempting to decide who is to blame for the chip company's repeated failures:

View attachment 21862

I texted that photo to my daughter yesterday with the caption “big crowd on beaker street.”

Is the one in the dark jacket emo or corporate?

 

[Colstan]

More fallout from Intel's latest earning report.

Intel Leadership Shuffle: Stuart Pann in for IFS, Raja Koduri out for GPUs & off to AI Startup

www.anandtech.com

Dr. Randhir Thakur is being replaced by Stuart Pann as head of Intel Foundry Services. This was a previously announced departure, so it's not a surprise. What is interesting is that Pann is a businessman, not someone with an engineering background. That hasn't historically worked out well for Intel.

More notable is that Raja Koduri is out as GPU chief. After getting fired for the failures of Vega over at AMD, Intel is now showing him the door for the failures of the Arc devision. Currently, Intel's graphics card efforts are on life support, with major price cuts on the A750 model. (Recently down to $225 at Newegg.)

Tom at Moore's Law is Dead has been covering this for some time, saying that "Arc is effectively dead". He got the leaked internal memo that Pat Gelsinger sent out to employees, stating that he had "mixed feelings" about Raja leaving. Typically, these things are announced in advance, but this was barely a footnote. Most likely, this was not an amicable departure.

Concerning the future of Intel's Arc plans, Tom says there have been revisions to the current gen Alchemist GPUs, but Intel may not put those into production, since current Arc products are sitting on store shelves. Next gen Battlemage won't be out until late 2024, and after that, Celestial in 2026, both with a single die planned. The one bright spot for Arc is that Gelsinger hasn't canceled the final product in the series, Druid, with a full line planned for 2027. It may be that Intel is planning to reboot their GPU products in four years, and just keeping it on life support until then.

It is possible that Gelsinger is refusing to completely give up on Arc, for personal reasons, because he was in charge of the failed i740 GPU project from 25 years ago, and doesn't want a repeat.

If you want the full skinny, Tom details what his sources said in the latest video.

 

[dada_dave]

More fallout from Intel's latest earning report.

Intel Leadership Shuffle: Stuart Pann in for IFS, Raja Koduri out for GPUs & off to AI Startup

www.anandtech.com

Dr. Randhir Thakur is being replaced by Stuart Pann as head of Intel Foundry Services. This was a previously announced departure, so it's not a surprise. What is interesting is that Pann is a businessman, not someone with an engineering background. That hasn't historically worked out well for Intel.

More notable is that Raja Koduri is out as GPU chief. After getting fired for the failures of Vega over at AMD, Intel is now showing him the door for the failures of the Arc devision. Currently, Intel's graphics card efforts are on life support, with major price cuts on the A750 model. (Recently down to $225 at Newegg.)

Tom at Moore's Law is Dead has been covering this for some time, saying that "Arc is effectively dead". He got the leaked internal memo that Pat Gelsinger sent out to employees, stating that he had "mixed feelings" about Raja leaving. Typically, these things are announced in advance, but this was barely a footnote. Most likely, this was not an amicable departure.

Concerning the future of Intel's Arc plans, Tom says there have been revisions to the current gen Alchemist GPUs, but Intel may not put those into production, since current Arc products are sitting on store shelves. Next gen Battlemage won't be out until late 2024, and after that, Celestial in 2026, both with a single die planned. The one bright spot for Arc is that Gelsinger hasn't canceled the final product in the series, Druid, with a full line planned for 2027. It may be that Intel is planning to reboot their GPU products in four years, and just keeping it on life support until then.

It is possible that Gelsinger is refusing to completely give up on Arc, for personal reasons, because he was in charge of the failed i740 GPU project from 25 years ago, and doesn't want a repeat.

If you want the full skinny, Tom details what his sources said in the latest video.

I found this upbeat passage from Anandtech at odds with everything else I’ve read about the Intel’s discrete GPU efforts:

Still, it will be interesting to see what kind of impact this has on Intel’s GPU and accelerator efforts. Raja Koduri has been a driving force for Intel’s GPU efforts for the last 6 years, leaving a sizable impression on their efforts in both the consumer and datacenter spaces. Intel is about to have a chart-topping, exascale-class GPU-based supercomputer to their credit with the nearly finished Aurora system, and Intel’s discrete GPU shipments for consumers are already within closing distance of AMD’s. All of which come from projects overseen by Koduri.

I mean they have Aurora yes, but that their consumer dGPU shipments are “within closing distance of AMD” is unexpected. Though the wording of shipments is potentially obfuscating as that can just mean shipping to sellers rather than moving units to customers. Still, interesting.

 

[Colstan]

I found this upbeat passage from Anandtech at odds with everything else I’ve read about the Intel’s discrete GPU efforts:

It may be a misleading statement, because it's an assumption based upon Intel's earnings report.

Intel Looks to Be Catching Up with AMD's Discrete GPU Market Share

Nvidia predictably retains the discrete GPU crown in Q4 2022.

www.tomshardware.com

"The Intel numbers are an estimate based on their financial report," said Jon Peddie, the head of Jon Peddie Research. "I would not get too excited about the closeness in shipping levels as Intel’s numbers are influenced by ASPs and that is soft data."

Put another way, this data represents sell-in for dedicated graphics cards into the channel. Sell-out — the number of dedicated graphics cards sold to actual consumers and businesses — would tell us what people are actually using, but we don't have that data. Right now, we know sell-in is higher than sell-out and that inventories at OEMs, distributors, and retailers have increased, but we don't know exactly how much inventory is available in the channel. It's possible, for example, that Intel has pushed a ton of Arc inventory into the channel but that it hasn't been sold yet.

 

[dada_dave]

It may be a misleading statement, because it's an assumption based upon Intel's earnings report.

Intel Looks to Be Catching Up with AMD's Discrete GPU Market Share

Nvidia predictably retains the discrete GPU crown in Q4 2022.

www.tomshardware.com

Yeah that’s what I figured. As I wrote above, it could mean shipping to sellers not necessarily customers and that’s what Tom’s is basically confirming with the caveat that we don’t have good data yet on the latter.