News AMD Posts Strong Results on Robust EPYC Sales As Consumer CPU Sales Disappoint

[msroadkill612]

"Datacenter actually wan't that good. The revenue was good, but the gross margins missed expectations. "

on 1.7B$ sales they make $444m operating income. I forget intel's numbers but the contrast is stark. AMD make say 40% as much from a fraction of the sales volume. You have singled out their most outstanding result for criticism.

 

[digitalgriffin]

Datacenter actually wan't that good. The revenue was good, but the gross margins missed expectations. Where AMD is really winning is with the Xilinx acquisition. The "embedded" segment (largely automotive and aerospace) killed it on all metrics. The embedded segment was more than 50% of their total profit while only being about 25% of their total revenue.

Intel has abandoned embedded fpga. Trying to get a hold of fpga chips in any large qty is near impossible.

 

[msroadkill612]

Given how diversified AMD has been trying to be (just look back at their CES presentation), they've been threading the rough waters a bit better, for sure.

As someone already mentioned, while having Xilinx is making them pay a lot QoQ, it is stil paying itself quite nicely and it'll definitely become a big part rather soon.

Server, well, we all know how that is going. AMD is truly competitive, if not flat out kicking Intel, so they have a strong foot there and it seems they'll be staying for a long while. Wendel in his podcast with Tom (MLiD) actually talked about how this is going from being in meetings where this is actually decided and shed some very interesting insights on the decision making for Corps. He had a very positive potential outlook for AMD, provided they keep doing what they have and don't go backwards on their pacing. Knowing AMD, that is not completely out of the scope of possibilities, LOL.

Anyway, on the consumer... Well, nothing much to say there. As many have already pointed out, these past few years, including "lockdown season" and the fierce competition between AMD and Intel on this front, nothing much either can do than just take the plunge and consider this a marketing expenditure (joking, obviously).

Let's hope they keep up the good (ish) work and bring competition.

Regards.

yep - that Wendel youtune was about the meatiest & best i have seen -2 very bright comp science phdS? who are active & connected talking shop - i loved it.

 

[digitalgriffin]

AMD is probably the only tech company that didn't do a "mass layoff" and that is due to the positive financial reports.

Good God man you just jinxed the employees.

 

[TerryLaze]

"Datacenter actually wan't that good. The revenue was good, but the gross margins missed expectations. "

on 1.7B$ sales they make $444m operating income. I forget intel's numbers but the contrast is stark. AMD make say 40% as much from a fraction of the sales volume. You have singled out their most outstanding result for criticism.

Are you getting quarters confused with full year?!
Intel made about ten times as much revenue at 19.1 bil and made much more than ten times as much in operating income at 6.2 bil

 

[JamesJones44]

"Datacenter actually wan't that good. The revenue was good, but the gross margins missed expectations. "

on 1.7B$ sales they make $444m operating income. I forget intel's numbers but the contrast is stark. AMD make say 40% as much from a fraction of the sales volume. You have singled out their most outstanding result for criticism.

My post isn't an "Intel" vs "AMD", it's stating fact. Their margins were lower than they guided to last quarter, signaling pricing pressure. Also, maybe you missed the part where embedded was more than 50% of their profit?!?!?

 

[JamesJones44]

Are you getting quarters confused with full year?!
Intel made about ten times as much revenue at 19.1 bil and made much more than ten times as much in operating income at 6.2 bil

I'm not talking about Intel, no where in my post is there a word about Intel. AMD last quarter said they would have a 30% operating margin for DC, they came in at 27%. No one is saying it's "bad", it's just not as good as the headline suggests. Again, the big surprise was embedded, it not only beat AMD's estimates it also beat even the most bullish analysis estimates. Leading off with DataCenter was great really isn't telling the real story. DataCenter was good, embedded was amazing.

 

[Jimbojan]

AMD datacenter business was able to increase $100M from last quarter ( from $1.6B to $1.7B), so did Intel ( from $4.2B to $4.3B), but as Intel’s Sapphire Rapids is just coming on line in 1Q23 and potentially next gen Emerald Rapids in 3Q23, AMD’s datacenter business will likely to go down just as its PC business did in the last year or two, if not were Xilinx business, AMD would have a big loss. I am sorry, I can not see AMD can have such high PE nor has any bright future as Intel is moving rapidly in the design and fab, which will pass TSMC in a year or so.

 

[TJ Hooker]

Are you getting quarters confused with full year?!
Intel made about ten times as much revenue at 19.1 bil and made much more than ten times as much in operating income at 6.2 bil

They were talking about datacenter specifically. In the last quarter AMD had higher DC operating income, $444M vs $371M for Intel. Intel's DC revenue for the quarter was about 2.5x that of AMD.

 

[TJ Hooker]

I'm not talking about Intel, no where in my post is there a word about Intel. AMD last quarter said they would have a 30% operating margin for DC, they came in at 27%. No one is saying it's "bad", it's just not as good as the headline suggests. Again, the big surprise was embedded, it not only beat AMD's estimates it also beat even the most bullish analysis estimates. Leading off with DataCenter was great really isn't telling the real story. DataCenter was good, embedded was amazing.

Where did they give guidance for operating margin specifically for DC for Q4? All I'm seeing in their Q3 release is guidance for gross margin for the company as a whole.

 

[waltc3]

There's nothing to see here at all, really. It's a good quarter for AMD. Consumers bought out last year; for most of '22 they sat pat. Lisa Su has stated she thinks the second half of this year will see a big pickup in consumer business (newer CPU x3Ds/midrange GPU products on the way.) For the last 20 years at least I've been hearing that "the PC is dead (or dying)" it's not. It never has been. These things come in cycles, always. Right now we have idiots in Washington, and inflation is king, and so on. Product cycles are product cycles. This notion that it's always supposed to go straight up like gangbusters is rather silly, as it's never been that way. PC markets for games and hardware have never been bigger. And the server-side and data center stuff is amazing--from AMD, anyway.

 

[RichardtST]

No worries. I've got my eye on that 7950X3D... I'm sure I'm not the only one who has been holding out and waiting for it. Now... where is all this excess DDR5 memory everyone is whining about? I can't find it!

 

[bit_user]

With 40 year record high inflation in the U.S., $5/gal. gas and $8 for a dozen eggs consumers simply have less discretionary income.

Except the egg price spike is due to avian influenza, and therefore shouldn't be taken as an indicator of inflation. Once that passes and hen stocks have a chance to recover, you'll see egg prices return to long-term trends.

With tens of thousands of tech job layoffs being announced almost daily by Microsoft, Apple, car makers, Intel and more it's inevitable that there will be significantly lower consumption because most unemployed workers don't make unnecessary purchases when they don't know when they will have a job again

In terms of gross numbers, the layoffs announced so far haven't really made a dent in the overall unemployment rate. However, they are focused on a mid/high-income, tech-oriented segment of the workforce, which tends to do a disproportionate amount of the spending on tech products. Furthermore, the rest of that workforce that's still employed will be more wary of making big, unnecessary purchases on PCs and accessories, until the overall health of this sector improves. Too often, one round of layoffs is followed by another.

More importantly, corporate spending on IT products will also be down.

 

[bit_user]

how to pay 1700 dollars for something will be obsolete next year?

Two years. Nvidia has been refreshing their consumer GPUs every 2 years, for several generations now.

 

[bit_user]

Yes, AMD's data center business will start to lose share, as Intel's Sapphire Rapids just came out and a new gen with Intel 4 in 6 month will gain significant market share as Intel's data center and other CPUs will have a better power efficiency than AMD's 5nm chip with better performance.

If you really believe that, then you need to study the Geomean and aggregate power consumption stats on the last couple pages of these benchmarks:

Intel Xeon Platinum 8490H "Sapphire Rapids" Performance Benchmarks Review - Phoronix

www.phoronix.com

AMD Genoa wins most of the Geomean rankings, sometimes by a lot, and yet Intel's Sapphire Rapids uses significantly more power. This is not a good situation for Intel.

as Intel’s Sapphire Rapids is just coming on line in 1Q23 and potentially next gen Emerald Rapids in 3Q23, AMD’s datacenter business will likely to go down just as its PC business did

No, Sapphire Rapids was competitive against Milan (Zen 3), not Genoa (Zen 4).

 

[bit_user]

For the last 20 years at least I've been hearing that "the PC is dead (or dying)"

Um, 20 years? No... I can't believe that. 20 years ago, we didn't (really) have tablets or smart phones, much less chromebooks. I guess consoles were starting to nibble at some PC gaming, but unless you mean specifically desktop PCs (i.e. not laptops), I don't believe it.

Right now we have idiots in Washington,

There are always idiots in Washington. Not only right now! The political process is a popularity contest, and that tends to select the wrong kinds of people for the job. We need a better way to choose lawmakers.

Product cycles are product cycles.

I think you mean "business cycles".

 

[bit_user]

AMD is probably the only tech company that didn't do a "mass layoff" and that is due to the positive financial reports.

They're also the only one that doesn't pay out dividends. Coincidence?
🤔

 

[TerryLaze]

If you really believe that, then you need to study the Geomean and aggregate power consumption stats on the last couple pages of these benchmarks:

Intel Xeon Platinum 8490H "Sapphire Rapids" Performance Benchmarks Review - Phoronix

www.phoronix.com

AMD Genoa wins most of the Geomean rankings, sometimes by a lot, and yet Intel's Sapphire Rapids uses significantly more power. This is not a good situation for Intel.


No, Sapphire Rapids was competitive against Milan (Zen 3), not Genoa (Zen 4).

Sapphire rapids is still better in some things, namely AI which is an extremely small market that nobody cares about /s .
And cpu max that will come out in 6 months, or whenever it will come out, will make intel an even better choice.
From your link:

The Xeon Platinum 8940H outright dominated in the AI benchmarks with the likes of oneDNN, DeepSparse, and OpenVINO able to fully leverage the Sapphire Rapids capabilities with AMX. There was also very strong showings out of the 4th Gen Xeon processors with the likes of Open Image Denoise for image denoising and OSPRay, GraphicsMagick, Python, PHP, OpenJDK Java, etc. The 4th Gen Xeon Scalable performance struggled when it came to HPC/server workloads that scale well to high thread counts where the 96-core / 192 thread EPYC 9654 could benefit, even in 2P configurations up to 384 threads. Some of the memory-intensive workloads also did better thanks to AMD EPYC "Genoa" supporting 12 channels of DDR5 memory, but for those same workloads it's also where the Intel Xeon CPU Max Series with HBM2e should perform very well too. Unfortunately, no hands-on access to the Sapphire Rapids HBM2e SKUs yet for benchmarking there.

 

[bit_user]

Sapphire rapids is still better in some things, namely AI which is an extremely small market that nobody cares about /s .
And cpu max that will come out in 6 months, or whenever it will come out, will make intel an even better choice.

Yes. If you're running AI workloads on a CPU, then Xeon Max will clearly excel.

Right now, investors clearly seem to think Nvidia is in a better position to capitalize on the ChatGPT boom.

ChatGPT Has Turned NVIDIA's AI GPUs Into Literal Gold, Company Stock Soars Over 40% This Month

The rising popularity of ChatGPT, an AI-powered language tool, has turned NVIDIA GPUs into literal gold for the booming AI business.

wccftech.com

 

[TerryLaze]

Yes. If you're running AI workloads on a CPU, then Xeon Max will clearly excel.

Right now, investors clearly seem to think Nvidia is in a better position to capitalize on the ChatGPT boom.

ChatGPT Has Turned NVIDIA's AI GPUs Into Literal Gold, Company Stock Soars Over 40% This Month

The rising popularity of ChatGPT, an AI-powered language tool, has turned NVIDIA GPUs into literal gold for the booming AI business.

wccftech.com

Sure, but nvidia is also a relatively small company that can't supply the whole market, there is plenty of room for arm gpus and intel and many others in the AI market.
This is also another reason that intel started their own GPU division, they want in on all of that money, and with all of the FAB space coming up intel will be in a position to supply most of that market.

 

[digitalgriffin]

Are you getting quarters confused with full year?!
Intel made about ten times as much revenue at 19.1 bil and made much more than ten times as much in operating income at 6.2 bil

Whether you want to admit it Intel is getting it's lunch eaten on all sides. Alibaba, google, and Amazon created their own arm chips. NVIDIA has a power hold on AI accelerators. AMD has a better more space efficient and power efficient platform that leads to better TCO. 256 threads in a single rack mount? Manageable heat and power? Lowest TCO? 2 cpu core Epyc.

 

[digitalgriffin]

Yes. If you're running AI workloads on a CPU, then Xeon Max will clearly excel.

Right now, investors clearly seem to think Nvidia is in a better position to capitalize on the ChatGPT boom.

ChatGPT Has Turned NVIDIA's AI GPUs Into Literal Gold, Company Stock Soars Over 40% This Month

The rising popularity of ChatGPT, an AI-powered language tool, has turned NVIDIA GPUs into literal gold for the booming AI business.

wccftech.com

Technically speaking spec wise, AMDs mi300 accelerators are significantly more powerful than NVIDIAs But the tool chain isn't there as every major toolset is heavily invested in CUDA like libraries.

This will s why NVIDIA has a death grip over the market.

 

[bit_user]

Whether you want to admit it Intel is getting it's lunch eaten on all sides. Alibaba, google, and Amazon created their own arm chips. NVIDIA has a power hold on AI accelerators.

Don't forget that Nvidia now has its Grace ARMv9 CPUs. So, their DGX systems can soon ship without any x86 CPUs in them.

https://www.tomshardware.com/news/nvidia-details-grace-hopper-cpu-superchip-design-144-cores-on-4n-tsmc-process

AMD has a better more space efficient and power efficient platform that leads to better TCO. 256 threads in a single rack mount? Manageable heat and power? Lowest TCO? 2 cpu core Epyc.

Genoa has up to 96 cores, which gives you 384 threads in 2U. Perahaps you're referring to Begamot providing (presumably) 512 threads in 2U, or 256 threads per 1U?

As an aside, I wonder if CXL.mem will finally be the end of the dual-CPU configuration, as the server mainstay. It seems the main source of its popularity is the memory scaling provided by the second socket. However, if you can attach banks and banks of memory via a CXL switch fabric, then a single CPU could host more memory than even conventional 2P setups.

 

[bit_user]

Technically speaking spec wise, AMDs mi300 accelerators are significantly more powerful than NVIDIAs

I don't think you can say that (yet). I can't find detailed specs on the MI300's compute capabilities, can you? And the MI300 isn't set to ship until mid-2023, whereas the H100 officially launched way back in March 2022, making it a slightly mismatched comparison for other reasons.

Also, AMD's messaging around the MI300 has been that it's a specialty product, rather than something they'll necessarily offer on the open market. The H100, on the other hand, you can even buy on a PCIe card and slap it in a bog standard server, if you don't want a more bespoke SXM-based solution.

But the tool chain isn't there as every major toolset is heavily invested in CUDA like libraries.

This will s why NVIDIA has a death grip over the market.

AMD has a CUDA clone (i.e. HIP) + porting tools. That's enough for some of their customers. They're porting an increasing number of machine learning, HPC, and rendering libraries and frameworks to HIP. Intel is also doing the same, with its oneAPI.

 

[TerryLaze]

Whether you want to admit it Intel is getting it's lunch eaten on all sides. Alibaba, google, and Amazon created their own arm chips. NVIDIA has a power hold on AI accelerators. AMD has a better more space efficient and power efficient platform that leads to better TCO. 256 threads in a single rack mount? Manageable heat and power? Lowest TCO? 2 cpu core Epyc.

There is still plenty of lunch left.
What you are talking about isn't a recent thing but has been going on for decades.
There is almost no customer that can do everything on one platform alone, arm is for a completely different workload than what people would buy x86 for, and not everybody in the server world needs the best efficiency, everything else be damned, amd has almost zero support so if you buy from them you are on your own to figure any issues out, intel has support out the wazoo for both software and hardware.
ARM can't do everything, nvidia AI can't do everything, amd can't do everything and intel also can't do everything.

I think custom arm has hurt traditional server suppliers that used RISC from the get go (like IBM) more than it did hurt intel.