That says more about Intel's shambolic efforts than it does about AMD's road to damascus moment.
Honestly both of them have demonstrated that they have little interest in AI. Intel gave up on Optane right as (a much-improved successor) would've helped with MoE offsetting or even as a highly durable drive for KV caching.
AMD have... turned out some decent gaming graphics cards and fine server CPUs but little else.
People are out there solding SMDs on old Nvidia mining rig GPUs to use them for inference; their hardware is not depreciating at the normal rate you'd expect from GPUs. DGX spark is finally useful now that AI can write multi-machine inference tooling optimised to it. Where's AMD in all of this? Nowhere.
AMD is doing 5 day RTO for almost everyone with 3 months of notice in an obvious attempt to get people to quit before vesting their 5x stock grants.
The rule includes >90% of "remote" workers who have discovered they were classified as hybrid (a classification now eliminated) and their manager didn't make it official. Now, remote approvals are SVP level.
The issue is, AMD still pays significantly less than any company with an RTO mandate, even with the stock appreciation. And the stock has gone down 25% in the past few weeks so it's even less competitive. The typical engineer was not given an equity-weighted package, even L9 only 30% of your TC is in stock.
It's a recruiting bonanza right now and everyone's getting calls offering 50% pay bumps from 5-day office semi corps like Broadcom or Qualcomm. The better people are getting offers at Google or Apple for 3-day RTO for even greater TC bumps.
The board should've sacked the entire leadership team of Intel and AMD years and years ago.
The fact that you can sit through both a mining boom and now a generational AI boom and have little to show for it and still keep your job is a Yahoo!-level fumble.
They can't find new leaders! Sharon Zhou, VP who handles all our engagements with the frontier labs quit today to be an MTS at Anthropic. https://lnkd.in/p/g53S3-fG
Sharon was the AI Chief of Staff brought in through an acquihire, managed all the high-visibility external/internal AI engagements, reported directly into Lisa Su, and was being groomed to be Lisa's successor. Even with all that, she couldn't make it 2 years at AMD.
I don't see any way out of this hole. We can't fire our execs and promote internally because our internal directors/VPs are mostly non-technical. We've been trying to pull leadership in with acquihires but we've been losing them too. Best win was replacing Alan Lee w/ Anush Elangovan which finally got ROCm working but there's a massive re-org coming he might not survive.
I remember buying an AMD K6-2 back in the day as it was cheaper and better. I've dropped in and out of the AMD sphere depending on who was winning at the given moment.
Ryzen series chips might not have been released if AMD's entire leadership team was sacked many years ago.
The heavy investments into HMB between SK Hynix & AMD might also not have happened. They went down this road before HMB demand popped off and it was a risky bet.
"The fact that you can sit through both a mining boom and now a generational AI boom..."
It's not necessary for every company to 'revenue maxx' by jumping on every trend.
AMD made $3.5 billion in profit last year, money left over after all salaries were paid and all expenses were settled. It might be galling to this generation of avarists, but maybe that's enough?
'Doing a decent job' seems to have become seen as archaic these days, which explains much of how the stock market works. The line doesn't _have_ to keep going up faster and faster.
They brought back some questionable former managers like the guy responsible for 14nm, Sanjay Natarajan, which was the first process to get stuck with a protracted period of low yields.
AMD has their Strix/Gorgon Halo as a competitor to DGX Spark, which is at least in the ballpark in terms of bandwidth, and being x86 has some compatibility benefits.
If you wanted a GPU with 32GB of memory for AI tasks, the AMD AI Pro R9700 was (up until about a month ago) reasonably priced at ~$1300. It's not a RTX 5090 in terms of bandwidth and compute, but for the price and being actually purchasable it's a reasonable pick.
AMD has an entire division dominating in HPC and #2 close behind nVidia in datacenters with extremely good margins. Over half their revenue is from those segments. Not a lot comes from consumer devices right now. That's why they did so well against Intel right now. Intel is the one that is failing in those markets.
I never understood why some fraction of technologists seem convinced that Optane was this amazing thing that would've been awesome if only Intel hadn't killed it.
It never made sense. It's slower than RAM, and what do I get for it? Persistence? I don't want persistent RAM, I like the ability to reboot to solve problems, and storing stuff on a non-volatile mediums has all sorts of legal implications which it's easier to avoid.
It's more expensive and has less storage than SSDs, even though the main downside of SSDs is their high price per gigabyte and relatively low capacities.
It's going in the wrong direction for both use cases. I guess it could have a very niche use case as a super fast persistent caching layer in front of spinning rust but that's kinda it.
Having a very low latency, durable write destination is a superpower for lots of database architectures (or things that share architectural details with databases, like some filesystems).
Optane had much lower latency than other types of durable storage, and also had very good endurance compared to NAND.
At a prior employer, we made extensive use of them as ZFS SLOG drives for large database servers. They made fsync very inexpensive.
Yep. I use a P4800X for ZFS special vdev. It's a monster. The thing has write endurance for many petabytes of write. The queue depth alone make for fast small reads; and the architecture means sustained reads and writes are... sustained. Try that with a normal SSD/NVME.
If Intel had volume production of Optane and had been able to quickly reconfigure it to produce a device that could stream out model weights fast enough for inference or could have gotten a read-write device that could hold KV caches effectively, they could have made an absolute killing in the current market.
HBF is slowly going after the former market. I'm not sure flash has a great answer for the latter due to endurance issues.
I can't think of any company that is as good as Intel at having a technology that's ready or almost ready at or before the time that the market is ready for it and then completely failing to sell the product into the newly developed market.
Huh. I wrote a whole thing here about serving a niche in serving local LLMs, but turns out, any niche it could have served is probably better served (per dollar) by either more RAM or modern SSDs.
There's an argument to be made for the persistent memory being a cheaper substitute for more RAM, but imo the gap isn't big enough to justify it when considered as a % of the rest of the system you're building around it and the performance tradeoff.
Important reminder to check your biases. I was sure Optane had missed its mark by dropping out just before LLMs became a primary use case for high $ systems.
There was a version of optane (optane DC?) that plugged into DIMM slots and cost less than RAM, with possible higher capacities. That would be a useful tradeoff had it kept up the advantage.
Intel had multiple false starts on getting their platform ready for these devices to be used in a sensible manner. I have one of these false starts in a closet somewhere, and I did some kernel dev work for it. It was an incredible mess.
(I didn't actually have an Intel DIMM module -- I had a different vendor's. But fundamentally there needed to be a clean way to get data from software into the durable memory domain, and there wasn't. There also needed to be a clean way to talk to the module to get health information and to configure it to the extent it was configurable, and that didn't exist either. I wrote the kernel driver for the latter and submitted it upstream, and it probably could have gotten merged if Intel had assisted with the gnarly bits involving coordination with platform thermal firmware that shared the same SMBUS lines, but they didn't.)
That's the persistent memory modules; they were roughly half the price per GB compared to DDR4 from my quick research, but substantially higher memory latency, which would significantly affect usage as a KV cache, which would probably be their primary use case since SSDs beat it in raw data throughput. Who knows how it would have developed over the years since it was discontinued, but my gut reaction to the numbers said you'd rather just invest in more RAM based on the numbers I saw.
More like they stopped investing in improving that tech so flash eventually caught up on throughput. It didn't catch up on latency and has never even entered the ballpark when it comes to endurance.
The other answers are hinting at it, but Optane was fundamentally much much lower latency than NAND flash used in SSDs.
SSDs use lots of tricks to hide this latency (using RAM). However, when the data is super critical, you can't just throw some data in RAM and hope that it gets written out.
Enterprise flash drives have battery backups and keep the system alive for a moment to flush out the writes, but Optane doesn't need special tricks to get this sort of durability, it just has it. The Optane can keep accepting writes, eventually the RAM of the SSD is filled, and then the write performance falls off a cliff. However, if your workload gives the storage device a breather to write out to flash, you wouldn't notice the difference.
If you have a large stream of critical data flowing in that has to be stored and can't be lost- Optane suddenly looks way better than SSDs
Not to forget AMDs dominance in performance per dollar in CPUs of any variety up until very recently on basically all fronts. Consumer, handheld/low power, server, you name it.
Intel finally has really great low power/budget answers, but still loses greatly to 3d cache chips on most workloads.
Intel's newest gen is a pretty clear favorite for laptops (especially with their better wifi/bluetooth/USB chipsets). Outside of laptop though, Intel has nothing compelling.
Totally. That's what I was vaguely (from mobile, sorry) trying to give them credit for. Their latest low power stuff is genuinely good, and their iGPUs are catching up fast.
Persistent memory will be back, it's just too compelling to give up on, and thanks to the research industry it spawned we have a pretty good idea of how to build systems that effectively exploit it. I can't wait until every DB can use pmem to achieve the holy grail of low-latency transactional durability (which is impossible today even with NVMe).
The community is basically doing this now with the kickflips in driver/rocm/etc "stuff". It's gotten so good and well-embraced that even AMD is backing some of these community efforts.
It's due to ROCm, think equivalent of CUDA lower level drivers, which is well problematic.
There's huge upswing in using Vulkan Compute which works across NVIDIA, AMD, and Intel (and even others), and which is apparently much easier and stable to run on AMD
Huh? Instinct is basically the only real alternative to Nvidia in the market. And they have the only real rack scale competition with Helios. This revenue is not because of gaming gpus and server cpus. This literally is AMDs road to damascus moment with data center GPUs.
What's more interesting to me is how blisteringly evident it is that GPUs depreciate a lot slower than the analysts assumed. But Nvidia GPUs depreciate slowly because a lot of thought and resources were dedicated into building that ecosystem.
AMD has always been an also ran in this space and it shows to this day, but the true believers cannot be dissuaded of their faith so whatevs.
I honestly think Nvidia needs a worthy nemesis right now more than ever and I wish it existed but I don't think it does... Yet...
They're around, selling AMD Strix Halo (up to 128GB unified memory) on the low, equivalent to DGX Spark, level; and MI* chips for enterprise AI inference.
The problem they have and they've always had is that everyone uses CUDA. ROCm works, but everyone's toolchain is CUDA exclusive or CUDA first, so it's much harder to use AMD hardware.
Using higher level tooling (vLLM, llama.cpp) I have DGX Spark comparable performance for 1/3 the price (but without the same software out of the box experience nor the clustering network options).
>Honestly both of them have demonstrated that they have little interest in AI.
Awesome? Maybe this means supply wont be ruined and I'll be able to buy CPUs.
Optane really died before LLMs went big. I was at an enterprise using a TON of them, often for SAP HANA workloads.
> AMD have... turned out some decent gaming graphics cards and fine server CPUs but little else.
I'm on a AMD desktop, using an AMD gpu, next to a gaming console using an AMD CPU and GPU, making large DC purchasing decisions where intel hasn't even shown up as a factor.
Ahh TIL! I used ChatGPT to find out what your "..road to damascus moment." expression meant, it was quite complimentary to you;
> And honestly, "Intel's shambolic efforts" alongside "AMD's Road to Damascus moment" is rather delicious phrasing. The author has managed to deliver quite a sophisticated insult in a single sentence.
I have avoided Intel products since the time some law enforcement forensic team used DCILeech to read the RAM of my server hosted at Hetzner and extracted a LUKS key from it. That was a lifelong lesson. There is no guarantee AMD won't come up with similar shenanigans like Intel's DCI soon and of course we don't know what PSP does behind the scenes, but at least for now, they are generally better for anti-forensics-tuned Linux kernels.
Telegram chats not being end-to-end crypted are encrypted at rest on their servers. In this case telegram definitely controls the encryption keys but distributes them across jurisdictions to mitigate some risks.
Look into the WireCard scandal. WireCard was the Stripe of EU worth billions. They were even audited by one of the big 4 firms and they confirmed it was a real company.
Turns out it was a front for Russian intelligence. Even after the German government found out, the guy still somehow escaped to Russia. There’s a solid chance that Russia operatives have infiltrated the German government. They was even a big hack Russia’s state hackers did and a lot of information was leaked.
I bring this up because Hetzner was asking for my passport to buy a $10/m server. Simple opsec says to just pay a bit more and use DigitalOcean or some other VPS.
I have no idea what this has to do with the topic at hand. What does KYC from Hetzner have to do with Wirecard? Do you mean to say that you suspect that Russia is behind Hetzner?
That case was so baffling, it left me forever doubting German companies, especially ones that ask for my passport. A lot of businesses used WireCard, even the government itself.
Yeah I tried to move to Hetzner, but bounced off when they wanted my passport info. I've since moved to OVH, where I get significantly more resources for the same money compared to DO, they're a European company hosting in Europe (getting away from US tech was the primary reason I was looking for DO alternatives), and they never asked for my passport!
There are many good non-Hetzner European hosting companies, luckily.
You're twisting the Wirecard to such an extent it's de facto wrong.
Wirecard were a German fintech, but didn't have big market penetration across Europe. They were a real company with real operations, but a decent part of its balance sheet was fake (supposedly it was in an Asian subsidiary, there was no money there). It wasn't a front for Russian intelligence, it's COO was a Russian operative (and nowadays runs a shadow intelligence agency from Moscow, using incompetent morons to try to kidnap journalists, or scheming to hire ISIS suicide bombers to blow them up); those connections only surfaced after the whole thing imploded. But the whole company was an actual real place with real operations and real non-russian intelligence employees. It's still not certain if the CEO knew about the COO for instance.
Read the FT series on the whole thing, their investigation started the collapse.
AMD specifically calls out high sales of Instinct products as leading the sales growth. I wonder how their Epyc and Ryzen lines compare to Intels Xeon and Core lines. AMD's Instinct sales are practically nothing compared to competing Nvidia products, but no one† buys Intel Arc accelerators, so that may be the main reason AMD is beating intel.
do people / businesses even buy Xeons anymore? Any used server from the last 5 years I see is always EPYC and all VM services I see are EPYCs unless something ancient that is still on a Xeon E5 from like 2015
There are many years since AMD's EPYC CPUs have been greatly superior to Intel's Xeons and Intel has not succeeded to match them (Intel has a hope with the future Diamond Rapids Xeons, which are their first CPUs since 2019 that have a good chance to match AMD's).
In the last few years Intel's division that makes Xeons had huge losses, because in order to retain their big customers they had to accept to sell at big discounts, otherwise the customers would have switched to AMD.
All has changed this year, due to the AI craze.
AMD has sold all the server CPUs that it could make, within its allocations at TSMC. Everybody who did not succeed to buy EPYCs has bought Intel Granite Rapids Xeons, because they are not made at TSMC, but in Intel's fabs, so they did not compete with other TSMC customers for allocations.
So Intel has also sold every server CPU that they could make, and they are now trying to retool some fab that made new Panther Lake CPUs back to the obsolete Intel 3 CMOS manufacturing process, in order to make more Xeons, to sell them as long as this huge demand persists.
Thus in the last quarter both AMD and Intel have sold less consumer CPUs than in the past, because all who could did delay their upgrade plans due to the unacceptable memory prices, but on the other hand both AMD and Intel had record revenues and profits from selling server CPUs.
Last I tested about a year ago in Google cloud the best Intel vs best AMD same machine config otherwise, video transcoding speed was practically identical.
AMD was at over 46% of the "server" market 1Q2026.
Intel isn't even condered a player in HPC anymore. 147 of TOP 500 is AMD (it's really custom chips, which AMD also does with it's Xilinx and Instinct divisions, and nVidia that are chewing up the Intels)
Good. I was bitten by Intel's voltage issues in their 13th gen CPUs. Then again in the 14th gen replacement. They appear bloated, complacent and I hope AMD eats their lunch.
Having strong competition between the CPU vendors is always a good thing. AMD has made mistakes, but it's hard to argue they don't deserve to outsell Intel for a change.
Hopefully they don't completely swap roles completely, with Intel being incapable of competing and AMD stagnating without competition. I'd rather they both be competitive.
Absolutely. It's also why I like seeing the Arm and Risc processor pressure; forces them to keep upping their game and not get fat and lazy (like Intel did).
I have reservations about AMD due to repeatedly poor Linux support. My 8700GE APU has display freezes and can't recover 1-2 times a week. For a chip that was released 2.5y ago, that's nuts.
Is that a known issue? I had something vaguely similar happen with a GPU that had intermittent hardware failure.
Granted their rocm support was abysmal several years back and they lost my trust as a result. But nvidia is hardly any better and intel doesn't seem to be a viable alternative ...
I hope they use much of this windfall to invest in improved products in the future. It's unlikely to impossible that Radeon or Instinct will ever be more popular or faster than its competition, but it shouldn't languish, and the Zen series should keep running circles around Intel.
There is arguably no particularly strong competition. There would be strong competition if Intel and AMD had to compete with x86 CPUs from Apple and Qualcomm.
Or if Intel and AMD GPUs had to compete with Nvidia GPUs ... wait, they do, and they are absolutely losing.
The memory crisis has delayed both Intel's and AMD's consumer CPU launches, i.e. Nova Lake for Intel and Zen 6 for AMD, to next year, but when that will happen there are good chances to see a much stronger competition between them than during the last 6-7 years, because Nova Lake should finally bring back AVX-512, a.k.a. AVX10, to Intel's consumer CPUs, and it also has many enhancements that AMD has not implemented yet (e.g. APX and FRED), so unless Intel fails at execution, the competing Intel and AMD CPUs should be much more balanced than in the last few years.
Also Intel decided to not improve for a long time. They were just happy collecting rent and protecting margin. Only the paranoid survive. Intel used to be a memory company and they pivoted cause Andy grove could see what was coming.
Intel didn't decide not to improve. Their CPU R&D has a long pipeline and they made the wrong prediction on tech (10nm), which resulted in a decade long pipeline stall. They're just about through that now, we'll see if they've got their groove back in the next few generations.
Before the 10nm issues, they were making consistent IPC improvement every generation, and although core counts were pretty stagnant for consumers, they would add a couple cores every generation on top Xeon skus. Maybe they could have done bigger changes with more time between, but it was nice being able to expect new server CPUs every 12-18 months that were meaningfully better.
> they made the wrong prediction on tech (10nm), which resulted in a decade long pipeline stall
I don't think their node problems were anywhere near that simple. You also left out the part where AMD moved to chiplets while intel stubbornly remained monolithic for quite a while. And a whole list of other missteps.
It wasn't just that. Intel could have survived a failed node, or a few of them. Certainly wouldn't be their first.
Krzanich "optimized" the company for profit, by cutting out things he considered to be excess. Including a lot of their irreplaceable validation & verification talent, that AMD instantly snapped up. The lack of validation resources is directly responsible for the failures of many key Intel products during his tenure and later, because major bugs that could have been spotted earlier slipped through, and then resulted in horrible delays and even cancellations. He then spent the surplus of the company on 30 major acquisitions, of which only two were in any reasonable degree successful.
There are CEOs who have destroyed more value than Krzanich, but not many.
AMD benefited greatly from Intel's self inflicted wound: laying on their laurels thinking the leverage they had with OEMs and inertia will carry them through the hard patch like it did in their P4 times.
Intel stopped advancing in 2014-2015 for a long time. They flopped on the fabrication side, they stuck to the 2-4 core CPUs and charged an arm and a leg for anything with more than that. This gave AMD an opening with their cheap and yet still decently fast higher core count CPUs.
That’s the leverage I was mentioning. That’s what worked for them in Pentium days and thought will just work again. It helped a lot that AMD had a compelling product to sell.
They used the same tricks when AMD landed useful 64bit mobile CPUs (turion X2 series and the like). But by zen era even big box retail is selling and models
F in chat to my Intel investments, their stock price is doing bad right now. And now they're also cutting off B580, what unfortunate times. Nova Lake, where are you?
AMD is reaping the reward of pragmatism, their best gaming CPU skus often have 8 full cores with a large cache for desktops plugged into a wall outlet.
Intel has been chasing higher and higher core counts which look good on spec sheets and dazzle in rarely utilized workloads, but said core counts do nothing to run the apps and games that most people actually use. I think getting dumped by Apple broke them, they're competing with Apple on paper and AMD is eating their lunch in reality.
It's amusing because i915-derived integrated graphics had the largest GPU share for a long time, but they still tried to make an x86-based GPU instead.
They also have baffling policy of just introducing a feature and then snuffing people not paying enough from using it. "Oh for that you need more expensive model" "Oh yeah ? fuck you I'm buying AMD"
First it was virtualization instructions, then AVX stuff
This! All this shit that's happening with AI and core hardware component price hikes, is just blantant evidence that our current best *ism (we're now on level capital) is not now nor it will ever be, the thing that'll take us all anywhere worthy.
Companies bending over and submitting to the AI/LLM/Datacenter hyperscalers at the expense of completely depleting all avenues for regular everyday people to have affordable access to basic computing (imho basically just another utility these days) is something that I hope people remember once the well dries and they try once again to sell us their RGB powered wares.
So... F Nvidia, F Micron and F all these money hungry asshole companies, including the weirdos thirsting for your sticks of ram.
AMD relies on geopolitically dicey Taiwanese fabs. Why is nobody else worried about this? Berkshire divested all TSMC investments a few years ago so at least they're worried.
Those same "geopolitically dicey Taiwanese fabs" are used by almost every major chip company in the world, including Intel which uses TSMC to manufacture its Panther Lake graphics tile and all the tiles in Arrow Lake except the base tile. Nvidia, Apple and Qualcomm are just as dependent on Taiwanese fabs as AMD.
TSMC dominates world wide with 80% of the general chip market and nearly 100% in some specific markets.
Due to Trump policies (and geopolitical realities), they are building or helping to build fabs in the US and elsewhere. It's takes years to build a fab.
The truth is this is more of the same tech inflationary pressure driven by AI. AMD is winning for showing up. Don't get me wrong, AMD does have technically superior products for specific applications, but that never really mattered.
Ryzen and EPYC have been doing crazy kickflips since around 2018 in both value and performance. Since Ryzen 3000/Zen 2 things hit parity on IPC generally, and Ryzen 5000/Zen 3 and beyond has just been total crush mode. Especially after 3d cache was intro'd.
For me, it was core count, price and energy efficiency (especially after a hefty power limit reduction that costs barely any performance). Intel made faster single cores most of the time, but at ridiculous power consumption and lower total core count.
Zen 1's $/perf was so good. After I booted a freshly-unboxed Zen box and ran some benchmarks for work (RIP on-prem server rooms), I bought a bunch of $AMD at the next market open, for about $12 IIRC. It's shocking to me it took almost a decade for AMD to outsell Intel, the value proposition was evident on day 1.
Intel has always been better at marketing and sales than AMD. I'd say, it still is and that continues to be AMD's weak point. In my opinion, it should be doing better than it is given the state Intel is in currently.
Honestly both of them have demonstrated that they have little interest in AI. Intel gave up on Optane right as (a much-improved successor) would've helped with MoE offsetting or even as a highly durable drive for KV caching.
AMD have... turned out some decent gaming graphics cards and fine server CPUs but little else.
People are out there solding SMDs on old Nvidia mining rig GPUs to use them for inference; their hardware is not depreciating at the normal rate you'd expect from GPUs. DGX spark is finally useful now that AI can write multi-machine inference tooling optimised to it. Where's AMD in all of this? Nowhere.
The rule includes >90% of "remote" workers who have discovered they were classified as hybrid (a classification now eliminated) and their manager didn't make it official. Now, remote approvals are SVP level.
The issue is, AMD still pays significantly less than any company with an RTO mandate, even with the stock appreciation. And the stock has gone down 25% in the past few weeks so it's even less competitive. The typical engineer was not given an equity-weighted package, even L9 only 30% of your TC is in stock.
It's a recruiting bonanza right now and everyone's getting calls offering 50% pay bumps from 5-day office semi corps like Broadcom or Qualcomm. The better people are getting offers at Google or Apple for 3-day RTO for even greater TC bumps.
The board should've sacked the entire leadership team of Intel and AMD years and years ago.
The fact that you can sit through both a mining boom and now a generational AI boom and have little to show for it and still keep your job is a Yahoo!-level fumble.
Sharon was the AI Chief of Staff brought in through an acquihire, managed all the high-visibility external/internal AI engagements, reported directly into Lisa Su, and was being groomed to be Lisa's successor. Even with all that, she couldn't make it 2 years at AMD.
I don't see any way out of this hole. We can't fire our execs and promote internally because our internal directors/VPs are mostly non-technical. We've been trying to pull leadership in with acquihires but we've been losing them too. Best win was replacing Alan Lee w/ Anush Elangovan which finally got ROCm working but there's a massive re-org coming he might not survive.
I remember buying an AMD K6-2 back in the day as it was cheaper and better. I've dropped in and out of the AMD sphere depending on who was winning at the given moment.
Now they're both losing.
why not start with trying to fix that?
The heavy investments into HMB between SK Hynix & AMD might also not have happened. They went down this road before HMB demand popped off and it was a risky bet.
It's not necessary for every company to 'revenue maxx' by jumping on every trend.
AMD made $3.5 billion in profit last year, money left over after all salaries were paid and all expenses were settled. It might be galling to this generation of avarists, but maybe that's enough?
'Doing a decent job' seems to have become seen as archaic these days, which explains much of how the stock market works. The line doesn't _have_ to keep going up faster and faster.
I'm sure they pay more than Intel ;-)
If you wanted a GPU with 32GB of memory for AI tasks, the AMD AI Pro R9700 was (up until about a month ago) reasonably priced at ~$1300. It's not a RTX 5090 in terms of bandwidth and compute, but for the price and being actually purchasable it's a reasonable pick.
Those cards are cheap because they're not really that amazing.
It never made sense. It's slower than RAM, and what do I get for it? Persistence? I don't want persistent RAM, I like the ability to reboot to solve problems, and storing stuff on a non-volatile mediums has all sorts of legal implications which it's easier to avoid.
It's more expensive and has less storage than SSDs, even though the main downside of SSDs is their high price per gigabyte and relatively low capacities.
It's going in the wrong direction for both use cases. I guess it could have a very niche use case as a super fast persistent caching layer in front of spinning rust but that's kinda it.
Optane had much lower latency than other types of durable storage, and also had very good endurance compared to NAND.
At a prior employer, we made extensive use of them as ZFS SLOG drives for large database servers. They made fsync very inexpensive.
If Intel had volume production of Optane and had been able to quickly reconfigure it to produce a device that could stream out model weights fast enough for inference or could have gotten a read-write device that could hold KV caches effectively, they could have made an absolute killing in the current market.
HBF is slowly going after the former market. I'm not sure flash has a great answer for the latter due to endurance issues.
I can't think of any company that is as good as Intel at having a technology that's ready or almost ready at or before the time that the market is ready for it and then completely failing to sell the product into the newly developed market.
Density. PCM is closer in bits-per-unit-area to Flash than to DRAM.
There's an argument to be made for the persistent memory being a cheaper substitute for more RAM, but imo the gap isn't big enough to justify it when considered as a % of the rest of the system you're building around it and the performance tradeoff.
Important reminder to check your biases. I was sure Optane had missed its mark by dropping out just before LLMs became a primary use case for high $ systems.
(I didn't actually have an Intel DIMM module -- I had a different vendor's. But fundamentally there needed to be a clean way to get data from software into the durable memory domain, and there wasn't. There also needed to be a clean way to talk to the module to get health information and to configure it to the extent it was configurable, and that didn't exist either. I wrote the kernel driver for the latter and submitted it upstream, and it probably could have gotten merged if Intel had assisted with the gnarly bits involving coordination with platform thermal firmware that shared the same SMBUS lines, but they didn't.)
SSDs use lots of tricks to hide this latency (using RAM). However, when the data is super critical, you can't just throw some data in RAM and hope that it gets written out.
Enterprise flash drives have battery backups and keep the system alive for a moment to flush out the writes, but Optane doesn't need special tricks to get this sort of durability, it just has it. The Optane can keep accepting writes, eventually the RAM of the SSD is filled, and then the write performance falls off a cliff. However, if your workload gives the storage device a breather to write out to flash, you wouldn't notice the difference.
If you have a large stream of critical data flowing in that has to be stored and can't be lost- Optane suddenly looks way better than SSDs
SSDs are good, but an improved Optane would've been better.
Intel finally has really great low power/budget answers, but still loses greatly to 3d cache chips on most workloads.
There's huge upswing in using Vulkan Compute which works across NVIDIA, AMD, and Intel (and even others), and which is apparently much easier and stable to run on AMD
that's quite a large market and their only competitor has moved on to greener pastures.
AMD has always been an also ran in this space and it shows to this day, but the true believers cannot be dissuaded of their faith so whatevs.
I honestly think Nvidia needs a worthy nemesis right now more than ever and I wish it existed but I don't think it does... Yet...
They're around, selling AMD Strix Halo (up to 128GB unified memory) on the low, equivalent to DGX Spark, level; and MI* chips for enterprise AI inference.
The problem they have and they've always had is that everyone uses CUDA. ROCm works, but everyone's toolchain is CUDA exclusive or CUDA first, so it's much harder to use AMD hardware.
Using higher level tooling (vLLM, llama.cpp) I have DGX Spark comparable performance for 1/3 the price (but without the same software out of the box experience nor the clustering network options).
Great Hope and Crosby movie.
Optane really died before LLMs went big. I was at an enterprise using a TON of them, often for SAP HANA workloads.
> AMD have... turned out some decent gaming graphics cards and fine server CPUs but little else.
I'm on a AMD desktop, using an AMD gpu, next to a gaming console using an AMD CPU and GPU, making large DC purchasing decisions where intel hasn't even shown up as a factor.
> And honestly, "Intel's shambolic efforts" alongside "AMD's Road to Damascus moment" is rather delicious phrasing. The author has managed to deliver quite a sophisticated insult in a single sentence.
But surely intel services are smart enough to get people in there illegally.
Turns out it was a front for Russian intelligence. Even after the German government found out, the guy still somehow escaped to Russia. There’s a solid chance that Russia operatives have infiltrated the German government. They was even a big hack Russia’s state hackers did and a lot of information was leaked.
I bring this up because Hetzner was asking for my passport to buy a $10/m server. Simple opsec says to just pay a bit more and use DigitalOcean or some other VPS.
There are many good non-Hetzner European hosting companies, luckily.
Wirecard were a German fintech, but didn't have big market penetration across Europe. They were a real company with real operations, but a decent part of its balance sheet was fake (supposedly it was in an Asian subsidiary, there was no money there). It wasn't a front for Russian intelligence, it's COO was a Russian operative (and nowadays runs a shadow intelligence agency from Moscow, using incompetent morons to try to kidnap journalists, or scheming to hire ISIS suicide bombers to blow them up); those connections only surfaced after the whole thing imploded. But the whole company was an actual real place with real operations and real non-russian intelligence employees. It's still not certain if the CEO knew about the COO for instance.
Read the FT series on the whole thing, their investigation started the collapse.
https://www.netflix.com/title/81404807
†Disclaimer: I own an Intel Arc GPU
In the last few years Intel's division that makes Xeons had huge losses, because in order to retain their big customers they had to accept to sell at big discounts, otherwise the customers would have switched to AMD.
All has changed this year, due to the AI craze.
AMD has sold all the server CPUs that it could make, within its allocations at TSMC. Everybody who did not succeed to buy EPYCs has bought Intel Granite Rapids Xeons, because they are not made at TSMC, but in Intel's fabs, so they did not compete with other TSMC customers for allocations.
So Intel has also sold every server CPU that they could make, and they are now trying to retool some fab that made new Panther Lake CPUs back to the obsolete Intel 3 CMOS manufacturing process, in order to make more Xeons, to sell them as long as this huge demand persists.
Thus in the last quarter both AMD and Intel have sold less consumer CPUs than in the past, because all who could did delay their upgrade plans due to the unacceptable memory prices, but on the other hand both AMD and Intel had record revenues and profits from selling server CPUs.
Intel isn't even condered a player in HPC anymore. 147 of TOP 500 is AMD (it's really custom chips, which AMD also does with it's Xilinx and Instinct divisions, and nVidia that are chewing up the Intels)
Granted their rocm support was abysmal several years back and they lost my trust as a result. But nvidia is hardly any better and intel doesn't seem to be a viable alternative ...
Or if Intel and AMD GPUs had to compete with Nvidia GPUs ... wait, they do, and they are absolutely losing.
Before the 10nm issues, they were making consistent IPC improvement every generation, and although core counts were pretty stagnant for consumers, they would add a couple cores every generation on top Xeon skus. Maybe they could have done bigger changes with more time between, but it was nice being able to expect new server CPUs every 12-18 months that were meaningfully better.
I don't think their node problems were anywhere near that simple. You also left out the part where AMD moved to chiplets while intel stubbornly remained monolithic for quite a while. And a whole list of other missteps.
Krzanich "optimized" the company for profit, by cutting out things he considered to be excess. Including a lot of their irreplaceable validation & verification talent, that AMD instantly snapped up. The lack of validation resources is directly responsible for the failures of many key Intel products during his tenure and later, because major bugs that could have been spotted earlier slipped through, and then resulted in horrible delays and even cancellations. He then spent the surplus of the company on 30 major acquisitions, of which only two were in any reasonable degree successful.
There are CEOs who have destroyed more value than Krzanich, but not many.
Intel stopped advancing in 2014-2015 for a long time. They flopped on the fabrication side, they stuck to the 2-4 core CPUs and charged an arm and a leg for anything with more than that. This gave AMD an opening with their cheap and yet still decently fast higher core count CPUs.
Intel has been chasing higher and higher core counts which look good on spec sheets and dazzle in rarely utilized workloads, but said core counts do nothing to run the apps and games that most people actually use. I think getting dumped by Apple broke them, they're competing with Apple on paper and AMD is eating their lunch in reality.
They also stopped making 3D XPoint a year before AI would have bought all of it.
Sure, but its not like people were actively seeking that GPU, for long time it was the only option to get in a laptop that isn't 20 pounds and 18".
Micron wanted to sell the fab that manufactured 3D XPoint.
First it was virtualization instructions, then AVX stuff
Companies bending over and submitting to the AI/LLM/Datacenter hyperscalers at the expense of completely depleting all avenues for regular everyday people to have affordable access to basic computing (imho basically just another utility these days) is something that I hope people remember once the well dries and they try once again to sell us their RGB powered wares.
So... F Nvidia, F Micron and F all these money hungry asshole companies, including the weirdos thirsting for your sticks of ram.
Due to Trump policies (and geopolitical realities), they are building or helping to build fabs in the US and elsewhere. It's takes years to build a fab.
I want my money back.
The truth is this is more of the same tech inflationary pressure driven by AI. AMD is winning for showing up. Don't get me wrong, AMD does have technically superior products for specific applications, but that never really mattered.