[SPON: Use "brokensilicon“ at CDKeyOffer to get Win 11 Pro for $23: www.cdkeyoffer.com/cko/Moore11 ] [SPON: Get 10% off Tasty Vite Ramen with code BROKENSILICON: bit.ly/3wKx6v1 ] [SPON: Support MLID w/ AOC T8PLUS Mini PC: www.amazon.com/dp/B0D255YJJB?ref=myi_title_dp ]
It's funny you were talking about the market popping and here we are 12 days latter with the Tech industry seeing a significant chunk of their stock dissolved over the last 3 days lol
I've had to my i9 13900HX laptop replaced due to the CPU. I was getting numerous errors. So, Intel saying laptop CPUs are unaffected gives me pause to say the least.
The 13900HX is essentially a rebadged 13900K; All 24 cores/32 threads of the desktop part; same cache. One would think it may be better binned silicon due to its laptop application (ie. lower voltage for a given frequency, leading to lower heat output; better battery life etc.). If it were a 13900HK, then that would be an interesting outlier (ie. 14 core/20 thread). Intel is suggesting it's a voltage issue - which it well could be - and it'd be fair to expect that the HX series would be affected as well (ie. high core counts + higher voltage fluctuations in response to high frequency changes).
@@heyitsjel It's actually the other way around. Mobile chips tend to be desktop chips that barely didn't make the cut, so they get power limited and underclocked to be tossed into a laptop instead.
@hammerth1421 no, RPL-H is a completely separate die and design from RPL-S (which is used for HX as well). Any microcode issue (and fix) for RPL-S would likely be applicable to HX, but not necessarily H/HK
@@hammerth1421 Intel ;) said the contrary and i know lot of 13900K (like msi board) running less PL than my HX (mine PL1/2 160 PL3/4 250) MSI is running PL1 125 PL2 150W PL3/4 = PL2
Notice how the "fix" is coming out after all the benchmarking for Zen 5 will be done. Bookmark this comment, Intel's "fix" will lower single core performance and they don't want to get fucking destroyed by Zen 5
I understand the logic behind GN testing on current stock settings, but If I was in his position, I would have completely dropped the comparison to Intel with a clear remark as to why.
@@hammerth1421 I already commented on one of HUB's videos (before intel announced this "fix) that they should not include the 1X900 chips until they have fixed their issue. Pretty sure they wont do that but lots of people seemed to have agreed with me so /shrug
intel's response is designed to do multiple things: - shift blame to users (their own consumers, nice move) - obscure the problem by generating videos/articles stating "it's fixed" and therefore create controversy instead of an investigation - appear proactive/communicative in a time of crisis Unfortunately, this is a move that's ultimately designed to confuse future customers and it's already working well according to what I saw on my feed today.
Intel have been milking and stagnating the industry for years with their overpriced quad cores, now they're literally screwing over their customers, they deserve what's coming to them.
@@knabbagluon That's impossible to say as long as the issue isn't known. For all we know, Intel engineers might have had a misconception about something that played into the design of both CPU architectures. Is it unlikely? Sure, but not completely out of the question.
Honestly so thankful to Wendel and GN for continuing their efforts to investigate this issue and report their findings publicly. As somebody with a 13900hx laptop who has experienced game crashes, I am quite concerned. Albeit, these crashes were limited to xdefiant and I believe they were caused by a bad update rather than my 13900hx as they have since subsided with recent game updates. I’m proceeding cautiously as more info becomes available. As of now I have kept everything stock and have disabled the cpu overclock profile which came in the manufacturer software package on my laptop. I will devise a setup with core voltage offsets and popper multipliers, etc. as more info becomes available. Thanks Tom for giving people like Wendel even more of a voice on this platform and for sharing the plethora of information you are able to obtain from across the industry.
@@BrandonMeyer1641 13900HX is literally the best raptor lake chip with the 13980HX. The perfomance per watt is absurd, it makes the desktop raptor lake i9s and i7s look extremely bad.
You kinda got hit by the curse of Murphy's law. Buildzoid of actually hardcore overclocking just did a video because it seems like running a minecraft server will reliably kill a 14900k within 3 months. What that means exactly is another question though. Edit: minecraft server should be high memory usage with low memory speed on the actual hardware. This lines up with what Wendell is saying but isn't clear why. Buildzoid thinks they just push boost voltage too high but I don't think I agree.
Likely this is due to low/light load on the cores; causing them to frequently boost and fluctuate their clock speed (ie. from base clocks up to peak boost). When this happens, the core voltages are quite transient, going from relatively low idle levels (eg. 0.8-0.9v) up to close to 1.5v. Typically under higher load they'll settle back down below the 1.5v mark courtesy of vdroop (eg. 1.35v), but you still have that initial spike in voltage as the frequency is increased/maxed. Lightly used machines see this kind of behaviour constantly (ie. idle --> boost --> idle --> boost). This is likely what Intel is referring to as voltage issues, and I would expect this to even affect relatively lightly used machines (compared to say, a server)... I'm interested to see if the "servers" that have been dying have had restricted peak boost clocks (ie. no eTVB enabled, just the max turbo boost of 5.4ghz), or if they've been allowed to run as per their capability (ie. eTVB enabled; Inte Turbo Boost Max Technology 3.0 etc.). Even server grade boards with lower power limits (eg. 150-200W) could still see a few cores boosting to high speeds (5.6-5.8ghz) if the CPU load is relatively modest, which ultimately leads to higher core voltage. People need to understand that on a fundamental level voltage and power limit are related, but do have some scenarios where you could argue they have less influence on one another. ie. power = volt x amps. Therefore, setting a *lower* *power* *limit* does not mean that the cpu cores won't ever see higher voltages - it just means they're probably less likely to under high load scenarios as the CPU won't have the power budget to sustain high clocks on all cores (ie. the core voltage required at given amps would exceed the power limit/budget for a given clock speed). Under low/moderate load scenarios however, a given set of cores can use the majority of the power budget (ie. higher volts + amps for those cores) and achieve high clock speeds. If my memory serves me correctly, Buildzoid did a video a little while back tweaking his loadline and voltage settings, as the out-of-the-box settings had his 14900k spiking up over 1.5v during light work loads (ie. idle --> light work load).
yeah I suspect that this can be related to other complaints that the AVX offset was not kicked in under ultra light load, thus perhaps the memory intensive but light computing tasks such as minecraft server never trigger AVX offset, thus leading to ultra fast degradation compared to other gaming tasks
If the active core is at full boost but actually stalled on memory, despite not doing work the frequency is at max so max V which is then frying the ring bus with the L3 cache and memory controller on it. It's because of the memory bottleneck keeping the core cool and the CPU cool, while non-processing fetch is at constant max rate. Microcode changes might back off on voltage in these scenarios, which were set too aggressively to hit 6GHz on more dies.
Intel is going the Takata Airbag recall method. Replace faulty with also potentially faulty/degrading part to buy time. Drag this on long enough for it to not matter as people move to other CPU's.
Just purchased a 13700k on prime day for $200 bucks, because people are returning them out of fear. 😂 I'll make sure all my voltage settings are good, and I'll be fine.
be kindness my dear. not every humanitarian from india is janitor. many are xcelent four make the computer or cook the food or xcelent dancing or four xcelent branes.
I am an engineer who has debugged a lot of chip failure. The important information to find the root cause is 1. Does this fault occur at any clock frequency? 2, What is the FIT rate, and does voltage effect the FIT rate? 3. Where and when does this failure occur? (I assume the data is corrupted) There are many existing tests that can isolate the problem. The most difficult bugs to find happens when the failure is not easily repeated. The reason I am interested in the bug is, I worked for Intel and thought how easy it would be for me to sabotage the chip with a bug that can avoid the verification tests, I could create a bug that is near to impossible for a sleuth of engineers to identify. I could have created this bug and made money by shorting the stock. Makes for a movie, but I would never sabotage a chip.
What if Intel's rebrand is actually something they are doing in anticipation of the fallout of this Raptore Lake scandal? That is the only logical explanation for abandoning the i9/i7 branding.
@@saricubra2867 And still fail at an unusually high rate. That's the sketchy thing being pointed out here, Intel has only admitted fault and promised a fix for a tiny sample of all the SKUs having issues which include mobile CPUs as well as "extra stable" server versions of otherwise consumer desktop CPUs.
@@saricubra2867 except the HX series. These are literally desktop chips (ie. 13900HX is essentially a 13900K); presumably with better binning (ie. marginally lower voltages for a given frequency).
"Dave's Garage" have a good explanation of what happened with the Crowd Strike update. He used to work for Microsoft and have a good knowledge of the ins and outs of Windows.
When I saw Bartlett lake, I immediately thought that is the Raptor Lake die fix. But I feel there is another, far more important market for those than desktops or laptops. Intel launched the Xeon E-2400 series late last year. Those are also based upon Raptor Lake, but without the E-cores which are disabled. Now if there is somewhere Intel really needs extra P cores and absolutely no crashes, its there.
@zorexlt8312 APO does not support any produktive applications that would run on Xeon servers. And the Heterogeneous Architecture of 12th gen and later with a different ISA per core architecture is a stupid idea from the beginning in server space. They need to get rid of the e cores anyway..
biggest weakpoint with AI bubble is how narrowfocused stock market is... all magnificent 7 companies are basicly tied to this AI and pushing loads of money into investments and if it pops and revenue isnt enough, they bring down whole stock market. It is like leverage 100 for whole stock market by one single business use case.
@@nowayslowman I dont wish to see it pop, but human fomo seems strong to microfocus on one trend at time creating this weakness. Lot of other industries, aspects of life suffer from lack of funding and innovation now. Thus im criticizing why all money needs to go to just AI. Gamers are not convinced yet companies, studios can deliver something worth promises. Same is with regular companies.
I hope GN still goes ahead with their analysis on 14th gen. There is a statement on Intel saying the oxidation issues were real and supposedly "addressed" in 2023.
At this point AMD should do what the phone companies does when you switch from one carrier to another you get like a discount on a new phone, well AMD should give you a discount if you have a proof of purchase of intel 13th and 14th gen cpu's, give them like 15% off, imagine the flood 😂😂
Brings a whole new meaning to "Broken Silicon" with Intel. I never liked Intel(3% gen over gen lolz) or nvidia(80% price hike 30% performance. Feels like they sell software now.) I never thought I would see this. Intel chips dying like this? Sure older AMD processors(FX) can have issues, but this is a whole new level of crazy Intel. As for Nvidia they are an AI company now.(insert Michael Jacksons They Don't Really Care About Us)
They raised the VID from 1.5 to 1.7 for the 900k parts. Why would suddenly 1.5-7 volts be okay for a CPU? They just could not produce enough good chips, some getting blasted since they run above 1.5v. They probably expected it to take longer, but this simply can´t be a coincidence. Server chips follow Intel guideline so they get blasted with 1.55+ vcore too. So, anyone who did OC their parts and dialed in the voltages to common sense will be far less likely to be affected. 1.45 was always the upper limit, better is below 1.4. I got super lucky my chip runs at 1.34 idle and 1.22 under load. Perfectly stable since launch.
My 14900k has a constant and stable shift between 1.3V-1.45V (max) depending whether I'm putting it under load through doing something intensive like gaming, rendering or some other kind of content creation, then down to 1.3V+ whilst mostly idle on desktop.
So you're an outlier with one of the actual good ones off the production line, or it's coming. When literal firms who definitely know what they're doing with these chips, in motherboard that can't even OC in some cases, are having obtuse stability issues and failure rates, there is a problem. And not user error.
They may not be able to RMA them for two reasons. They dont have a fix that is fully tested, so they wont have any chips at all just yet. Secondly, how much physical effort would it be to replace everything and organise a fix on that scale? It would overwhelm them.
Yep, that's the problem. They have no problem RMAing stuff since that's booked as revenue lost rather than additional costs incurred unlike a mass refund (makes shareholders much less angry despite being more or less equivalent), but they need to have a series of known-good CPUs to replace them with. If Raptor Lake really is fundamentally flawed, Intel themselves understandably don't want to play Raptorian roulette with the replacements either.
@@hammerth1421Raptor Lake isn't the issue, Intel is. Since 8th gen they pushed max CORE VID voltages to the Infinity. 13th gen and 14th gen expose the problem because of the single core turbo boost started to increase since many generations ago.
and the retail where I bought my 13900KS, and basically all retails in my region, don't give even half a fudge about RMA. they even put out a 'store warranty' just so they can have lower pricing. So I'm left with the option to suck it up, sell it to someone gullible, which would be really difficult since the majority would only buy entry level and those who would are knowledgable enough they might already caught up with this drama, OR switch to an x3d system which is practically building a new PC for me.
All of this testing used to be pretty standard in all tech companies. Most tech companies have significantly downsized their QA teams. Double checking things takes time and money. Until a company like Crowdstrike is required to have an insurance policy that covers trillions of dollars of damage for bringing down the internet, that money is not going to get spent.
Anybody remember the TSMC 7nm degradation scare mongering? I have a been a user of Intel, PowerPC and AMD since the P3, G3 and Duron days and this is the worst case of CPU failure I have seen.
Intel updated their release to admit the 13th gen has oxidation issues. And yes they're very sneaky. Waiting till after Zen 5 review window is up for the "fix" as well as simply not being forthcoming about what other issues compounded the problem. And they should recall the 13900K's with oxidation at least. But they're likely hoping they last till warranty is up before they completely die.
Where have they admitted to oxidation issues? The only claim of oxidation is from Steve. Zero evidence yet. So far most of the evidence points to motherboard sending them too much voltage. 🤷
@@DingleBerryschnapps They admitted it was microcode bugs causing the CPU to request for incorrectly high voltages, not just the MB messing with settings and sending high voltages. The problem is on Intel for providing bad microcode. They also admitted, in a hard to find place (PR communication on Reddit, not the original statement), that some of the 13 gen CPUs were indeed effected by an oxidation via issue that they addressed in 2023. They basically confirmed that the oxidation via issue reported by tech outlets was indeed true, and they fixed it sometime in 2023. Why they put this section on Reddit and not their original statement about voltages is anyone's guess.
They had to present a cause/solution so close to the launch of Zen 5, regardless of whether it stated the actual cause or just another possibility. Research into the cause would never have taken so much time if it was simply an error in the microcode.
How many configurations are possible in a desktop system? With the near infinite amount of possible configurations, I'm sure it takes quite a while to figure out the root cause for all of the issues. Especially, if the symptoms for each CPU can be different from chip to chip. There is too much speculation and very little actual evidence in any of these videos.
@@DingleBerryschnapps Did you even watch the video? At 0:48 An Intel employee stated it was an elevated operating voltage stemming from a microcode algorithm giving incorrect voltage requests to the processor. So yes, too much voltage 'allegedly' - caused by a faulty Intel microcode 'allegedly'. The employee also stated an updated microcode to correct the 'alleged' issue is going through validation. Guess time will tell.
@@gbnq2513 do your Research. It was likely a miscommunication. Instead of proving the maximum safe voltage when required boards sent the maximum possible voltage. That’s why on Intels own boards this never really happened. It’s similar to how you can set the powerlimit to a max of 4096W. But that will never hit the CPU. The Mainboard manufacturers weren’t directly responsible for that but in the end the Mainboard did the wrong thing. On mainboards with the right setting this never happened
1. The simplest fix that I can think of right now would be to roll critical updates with windows that nerfs the voltage, just as they did with their security mishaps back in 2019. But that isn't going to be simple either. Introducing a new stepping would not affect existing users, while forcing AIB/retailers for massive RMAs is going to be absolutely ridiculous. 2. As someone who had worked on Silicon chip fabrication, I'd say this is most likely dielectric breakdown. Some of the oxides might have been grown too thin (in bad batch/batches) and slowly breaks down due to high current. 3. I'm certain that there was a huge disconnection between the engineering team and the product team. This is basic knowledge from the 50s, and manufacturing/industrial statistics are always taught in the 2nd year of college. Now, the entire company has to take one for themselves, yet again, in a situation where they really need not to screw up. Oh dear.
@@EntekCoffee wasn’t really an Intel mistake, more like a miscommunication. The Mainboard manufacturers allowed too high voltages because they got Intels spec wrong (and that’s pretty much on Intel). Instead of providing the maximum safe voltage they provided the maximum theoretical possibility voltage which was too much. That’s why it didn’t appear in Intels validations.
Thinking about Medusa, making chiplets work for laptop and desktop may be a chip stacking strategy. You have to go through the L3 MMU before you go to physical ram. Integrate an L3 3D Vcache and an IO die and then stack the chiplets on that. The IO die has the driver circuit that physically drive the external memory bus, All off chip access has to go through that anyway. Plus this gets you CPU cores closer to the heat spreader.
1:32:20 I am working in Fusion 360 and Blender. The experience with AMD GPUs improved drastically over last 2-3 years. In Fusion 360 it improved so much that I can get the job done on R5 5500U laptop with 8GB of RAM. And I work on no small projects! When not on the go, 7900XT just flies through everything. HIP support improves, the newer the Blender version, the faster RDNA 3 gets
People may think GPU doesn't matter for CAD but yes, yes it does. Especially in bigger projects. VRAM often is the limiting factor. With complicated geometry, GPU core also has a lot to do when working in viewport. The same is true in Blender. This gets better with a push towards HIP implementations and optimizations. Now 7900XT is getting where it should be based on in-games RT performance, so at the level of 4070, 4070Ti. The time you need to render an image is one thing. But to get everything set up to the point you actually can render an image or video, and do it fast, efficient and without issues is another thing. For rendering, 4090 is a king, no doubt. OptiX helps it a lot. But when comparing lower end NV GPUs, the performance delta is smaller than you'd expect now. To the point where it becomes somewhat irrelevant for smaller companies and individual users. When working in a viewport, so in 90% of time you actually work as the user, 7900XT is doing surprisingly well too. Plus, when you compare price competitors, 7900XT is doing amazing. In Europe, for the price of really good 7900XT you often can get basic 4070 Super 12GB. High end model or basic 4070 Ti if you are lucky. So again, not only 7900XT isn't really slower than these anymore. When RTX runs out of memory, suddenly 7900XT is faster and also saves a metric ton of time in post-processing. I love how consistent performance I get from this card. And the funny thing is, if or when I need more performance in Blender, I can just add second 7900XT. Not only rendering performance will roughly double, so they will be faster than 4090, it will be in a similar power budget while costing less (also true for 70 class NV cards) and having almost double the VRAM (not true for NV). Before anyone asks, in software like Blender you don't need SLI or crossfire. Not only software splits the work by itself, it also pools the memory. Meaning you can use all the VRAM you physically have on a GPU. And it can work with different GPUs, no need to match them. Though it comes with some inefficiencies.
I'd be surprised if this isn't a tolerance / threshold issue with regards to die binning or acceptable silicon level defects/variances for a given SKU. ie. they're possibly pushing the limits of what is acceptable silicon quality for something like a 13900k or 14900k, in hopes of optimising their yield. For context, I've had my 13900k running since November of 2022, at DDR5 7000 - it could handle 7200MT stable, but required additional voltage, which wasn't worth the benefit IMHO. Recently I shuffled those sticks over to the wife's system (12900k), and went to a 48gb kit at DDR5 8000MT. It's been running at 8000MT without a hitch, at even *lower* system voltages than the other kit required... and for the record, I run various long duration torture tests so it is genuinely stable (Linpackxtreme; OCCT; TestMem5 etc.). Why is this important? It suggests that *not* *all* CPU's are affected per some major design flaw, but more likely something relating to fabrication (such as incorrect design tolerances for silicon level defects/quality).
Same. I've had my 13900k running at 5.8p/4.5e/5.0ring with 4x8gb of 4200c15 ddr4 on a cheap MSI Tomahawk since November 2022 with zero issues. I like it more than my 7800x3d system.
@@jjlw2378I've actually been wondering if *earlier* CPU batches were less/not affected, and as demand has increased they've loosened the tolerances to meet demand and/or had the oxidation issue. This could also explain why it's taken a while for this issue to rear it's head en-masse...
@@jjlw2378also, I too was running an MSI board (MEG z690i Unify, ITX)... I've filled out the GN survey - it would be interesting to also see if there's any correlation with board manufacturers. With the MSI I was typically using the 288w "water cooling" profile, as it's typically been enough for the 13900k. Going to unlimited power profiles maybe gains you 5%-10% more performance in synthetic all-core workloads, for an insane increase in power consumption and temperatures. Recently I switched to an Asus z790i, which by default had *unlimited* power profile settings. I quickly installed the latest BIOS with the less insane default power profiles... so this too could be part of the reason our CPUs aren't wrecked.
@@jorgevaladas788 The thing is. it's not a final fix. Its just to lower the voltages even more. Lowering just prolongs your CPU and lowers your performance more.
@@lukas_ls what didn't age well, that they just released a patch so it doesn't degrade as much anymore? 50% of chips out there (13th and 14th gen) are all degraded. How do we know, Intel just said that 50% of the chips are not affected by degradation. You call that a fix? Are you daft? If this was any other company there would have been a recall but for some stupid reason fanboys think it's fine having 50% of your total sold chips as defective.
I think Intel has always known that they're pushing their CPUs too far, they just couldn't predict how long they'd hold up. You can see it in their ridiculous all core boost behavior, where the only all core bench they manage to really win is CB if you do a single run. If you do a 10min test the 13900k score plummets well below a 7950x and the 14900k is only slightly ahead, closer to tied. Both lose outright in Blender and decompression to the 7950X. Any kind of real all core work is typically going to run past that boost timer, so it's almost like it's designed solely to win CB sprints.
AMD had Opteron A1100-series ARM64 (Cortex-A57) maybe ~7 years ago. It was really an interesting one, especially since majority of new ARM SoC/IP companies didn't have previous real life experience in the datacenter land. AMD on the other hand knew what needs to be done.
1:26:55 if I was asked I'd answer - gaming laptops. Their H and HS CPUs are so good, both showing good performance and fantastic battery life(due to them being the same chips used for u-series CPUs), but they are barely present in laptops, especially high end laptops. Imagine a Legion 7 with a 175w 4090M GPU and Ryzen 9 7940HS. It would be capable of 4K gaming and would have like 10 hour battery life thanks to Ryzen low power consumption at idle and 99 watt hour battery Legion 7 has. But for some reason, such combinations pretty much don't exist. And I don't understand why. Thanks for the stream, guys!
FYI if you're buying an AM5 APU that has PCIe 5.0, 8 lanes of 5.0 is basically the same bandwidth as a full channel of SODIMM DDR5 if you were to enable CXL, and then, say, give the APUs 24 lanes like the CPUs, you could then do a full 16 lanes of PCIe that is almost as fast 2 channels of desktop DDR5 5200 with tighter timings.
There was a dev board that was available with that AMD ARM k12 chip floating around during the Zen 1 years. I wish I would have gotten it but it was expensive at the time and software compatibility wasn't there yet except for a homegrown Linux distro.
the problem is boosting 1 core to 6ghz which uses to much voltage 1.5v or more , solution is to lock all cores to 5.5 or 5.6 ghz also voltage 1.35 - 1.4 max
its been months now, If AMD have done what Intel PR its doing everyone would be cooking them , I want to disagree with you on this one, for me its not that Intel don't know what its going on, they do very well, but what they are doing its gaslighting, they want people to speculate, First they blamed Motherboards, then the the TVB bug, now its voltage bug. That makes them look like they don't know what its going on, that they are "investigating" because as you said they cant RMA everyone chip so make time, it could be that anything mentioned before could be part of the issue of a grand scheme. Remember when AMD was blowing up CPU they own it they make it transparent they didnt throw OEM under the bus and they fixed the problem. I wish that Intel as a BIGGER company with more resources could do better, they should.
@@JynxedKoma No company its perfect, but you didnt see AMD gaslighting everyone, in fact they took things on its hand and did what its right and in the end it fixed the problems, nvidia did the same to some extent with their crappy "user error" connector, meanwhile Intel its denying RMA and not taking responsability at all.
@@eilegz Just to clarify, I'm not weighing in on the former issues AMD might have had. I'm just asking as a first time AMD adopter of AM5 + soon 9950x once it releases in august, if I can be sure the chip won't suffer any instability or heat issues. If at all you have experience with that using previous gen AMD chips + motherboards since this past year. Thanks!
Im glad that you finally invited someone that adresses the Linux side of things , I use linux on my work PC and on my laptop (i only use windows for my gaming rig), so linux is a big deal to me
2:13:13 Actually Apple publicly stated that they were open to getting Windows on ARM running on Apple Silicon, it's just up to Microsoft to agree. Microsoft so far only officially supports WoA on Apple Silicon via Parallels virtual machines though.
if intel issue gets big ,people will loose their trust in them, intel will loose their customers, and i hope AMD won't raise their prices in wake of no competition
That is absolutely not but a smaller percentage of the issues of intels cpus. I would be outraged if i hadnt realized that AMD is the move for my last build. Everyone else is seeing this now. Ryzen will dominate. Even moore!
Only 4 or 5 years? My old laptop gave me 8 years of good usage and is still usable, but feels sluggish with its 4c/8t (but 32GB of DDR3 and 2.5TB of SSD storage was not an issue). I'm hoping my 12900H is going to last as long (hopefully my next laptop can be AMD) Powertop auto tune... and now my USB keyboard and mouse doesn't work 😂
Intel has to stop pushing overvoltage out of the box with an overclock and no powe/current limits with their motherboard partners. That CORE VID of 1.5 volts is embarrasing.
Intel doesn’t want to be genuine and truthful about this whole thing because of the hole they’d put themselves in by admitting it all. And to say it was a foundry problem. That would hurt the foundry business they’re trying to drum up. So both ends of the company are not firing on all cylinders as they’re leading everyone to be. Products always massively delayed and produced at other fabs currently.
i think it is as simple as INTEL is having a hard time competing with AMD and mistakes are happening... pressure causes mistakes. AMD has been applying non stop pressure.
The problem with Intel started before AMD was really back with the first attempt of the 10nm node, goals set too high and no-one stopped the whole thing and asked for a re-evaluation of the goals.
i bought a 14700kf and had the tekken 8 vram error early on, i was using intel extreme utility at the time and then uninstalled it, as well as adjusting nvidia vram amount, i no longer have any issues with tekken 8. this is how the issue started but is now server related and is the log indicating nvram issue as an error for warframe?
![[UA] NAVI vs MOUZ | BO5 | IEM Rio 2024](http://i.ytimg.com/vi/CMBYk0cbGoA/mqdefault.jpg)
[SPON: Use "brokensilicon“ at CDKeyOffer to get Win 11 Pro for $23: www.cdkeyoffer.com/cko/Moore11 ]
[SPON: Get 10% off Tasty Vite Ramen with code BROKENSILICON: bit.ly/3wKx6v1 ]
[SPON: Support MLID w/ AOC T8PLUS Mini PC: www.amazon.com/dp/B0D255YJJB?ref=myi_title_dp ]
It's funny you were talking about the market popping and here we are 12 days latter with the Tech industry seeing a significant chunk of their stock dissolved over the last 3 days lol
STOP TRYING TO HAVE A 'STACHE BRO!!! YOU LOOK TERRIBLE!!! JUST STOP!!! IT JUST DOESN'T WORK FOR YOU :(
Great filming with you as always!
Cheers my friend!
Good to see these guys slapping intel in the B hole when deserved.
I'm a simple man. I see Wendell from Level1Techs; I click the video and smash that like button.
Wendell has a posse!
(WWF reference)
don't smash it if you are using 13th or 14th gen. gently scroll over and lightly tap it to avoid stress on CPU
😂😂😂😂😂@@moldyapple1789
Smash that MF like button 😤💪🏾😤💪🏾😤
I bought a KVM when I barely even need one, it works super well just as advertised
Wendell is on an "Intel Failure" world tour 🙂
Hey, so is Pat Gelsinger! It'd be great of we could get a video with both of them 😅
@@kaeota”we’re seeing a lot of momentum” -Pat Gelsinger every quarter (company is losing money)
These guys are punching intel in the B hole.
@@jonathanjones7751 They said momentum lol never mentioned the direction.
@@jonathanjones7751We expect to see a 2x increase in demand in developing markets.
*Actually witnessing 2x decrease*
A broken silicon on broken silicon!
Good pun!
HA!
I see what you did there. 😅😅😅
I've had to my i9 13900HX laptop replaced due to the CPU. I was getting numerous errors. So, Intel saying laptop CPUs are unaffected gives me pause to say the least.
The 13900HX is essentially a rebadged 13900K; All 24 cores/32 threads of the desktop part; same cache. One would think it may be better binned silicon due to its laptop application (ie. lower voltage for a given frequency, leading to lower heat output; better battery life etc.). If it were a 13900HK, then that would be an interesting outlier (ie. 14 core/20 thread).
Intel is suggesting it's a voltage issue - which it well could be - and it'd be fair to expect that the HX series would be affected as well (ie. high core counts + higher voltage fluctuations in response to high frequency changes).
@@heyitsjel mine is running fine finger crossed
@@heyitsjel It's actually the other way around. Mobile chips tend to be desktop chips that barely didn't make the cut, so they get power limited and underclocked to be tossed into a laptop instead.
@hammerth1421 no, RPL-H is a completely separate die and design from RPL-S (which is used for HX as well).
Any microcode issue (and fix) for RPL-S would likely be applicable to HX, but not necessarily H/HK
@@hammerth1421 Intel ;) said the contrary
and i know lot of 13900K (like msi board) running less PL than my HX (mine PL1/2 160 PL3/4 250)
MSI is running PL1 125 PL2 150W PL3/4 = PL2
Notice how the "fix" is coming out after all the benchmarking for Zen 5 will be done. Bookmark this comment, Intel's "fix" will lower single core performance and they don't want to get fucking destroyed by Zen 5
That's not shady at all. /s
It's 10980XE and Threadripper all over again. Who could have guessed ...
I understand the logic behind GN testing on current stock settings, but If I was in his position, I would have completely dropped the comparison to Intel with a clear remark as to why.
@@hammerth1421 I already commented on one of HUB's videos (before intel announced this "fix) that they should not include the 1X900 chips until they have fixed their issue.
Pretty sure they wont do that but lots of people seemed to have agreed with me so /shrug
If the Vcore is the problem, they have to lower both single and multi core boost/max clocks.
intel's response is designed to do multiple things:
- shift blame to users (their own consumers, nice move)
- obscure the problem by generating videos/articles stating "it's fixed" and therefore create controversy instead of an investigation
- appear proactive/communicative in a time of crisis
Unfortunately, this is a move that's ultimately designed to confuse future customers and it's already working well according to what I saw on my feed today.
Users, developers, & corporations need Intel to state publicly that the 13th/14th gen problem will not be in Arrow Lake!
Yup. All I know is that I’ll never touch intel again
@@tringuyen7519 Arrow Lake is completely different in every way. It would be amazing if it would be affected.
Intel have been milking and stagnating the industry for years with their overpriced quad cores, now they're literally screwing over their customers, they deserve what's coming to them.
@@knabbagluon That's impossible to say as long as the issue isn't known. For all we know, Intel engineers might have had a misconception about something that played into the design of both CPU architectures.
Is it unlikely? Sure, but not completely out of the question.
Honestly so thankful to Wendel and GN for continuing their efforts to investigate this issue and report their findings publicly.
As somebody with a 13900hx laptop who has experienced game crashes, I am quite concerned. Albeit, these crashes were limited to xdefiant and I believe they were caused by a bad update rather than my 13900hx as they have since subsided with recent game updates.
I’m proceeding cautiously as more info becomes available. As of now I have kept everything stock and have disabled the cpu overclock profile which came in the manufacturer software package on my laptop. I will devise a setup with core voltage offsets and popper multipliers, etc. as more info becomes available.
Thanks Tom for giving people like Wendel even more of a voice on this platform and for sharing the plethora of information you are able to obtain from across the industry.
@@BrandonMeyer1641 13900HX is literally the best raptor lake chip with the 13980HX.
The perfomance per watt is absurd, it makes the desktop raptor lake i9s and i7s look extremely bad.
I like Buildzoid's ring bus failure hypothesis. Seems to explain the scattershot failures.
I try to never miss a wendel episode. combines the two best, most technical youtubers imo.
Always a treasure when Wendell is on. The discussion around the AI bubble and how that intersects with business and economics was super fascinating.
You kinda got hit by the curse of Murphy's law. Buildzoid of actually hardcore overclocking just did a video because it seems like running a minecraft server will reliably kill a 14900k within 3 months. What that means exactly is another question though.
Edit: minecraft server should be high memory usage with low memory speed on the actual hardware. This lines up with what Wendell is saying but isn't clear why. Buildzoid thinks they just push boost voltage too high but I don't think I agree.
Likely this is due to low/light load on the cores; causing them to frequently boost and fluctuate their clock speed (ie. from base clocks up to peak boost). When this happens, the core voltages are quite transient, going from relatively low idle levels (eg. 0.8-0.9v) up to close to 1.5v. Typically under higher load they'll settle back down below the 1.5v mark courtesy of vdroop (eg. 1.35v), but you still have that initial spike in voltage as the frequency is increased/maxed. Lightly used machines see this kind of behaviour constantly (ie. idle --> boost --> idle --> boost).
This is likely what Intel is referring to as voltage issues, and I would expect this to even affect relatively lightly used machines (compared to say, a server)... I'm interested to see if the "servers" that have been dying have had restricted peak boost clocks (ie. no eTVB enabled, just the max turbo boost of 5.4ghz), or if they've been allowed to run as per their capability (ie. eTVB enabled; Inte Turbo Boost Max Technology 3.0 etc.). Even server grade boards with lower power limits (eg. 150-200W) could still see a few cores boosting to high speeds (5.6-5.8ghz) if the CPU load is relatively modest, which ultimately leads to higher core voltage.
People need to understand that on a fundamental level voltage and power limit are related, but do have some scenarios where you could argue they have less influence on one another. ie. power = volt x amps. Therefore, setting a *lower* *power* *limit* does not mean that the cpu cores won't ever see higher voltages - it just means they're probably less likely to under high load scenarios as the CPU won't have the power budget to sustain high clocks on all cores (ie. the core voltage required at given amps would exceed the power limit/budget for a given clock speed). Under low/moderate load scenarios however, a given set of cores can use the majority of the power budget (ie. higher volts + amps for those cores) and achieve high clock speeds.
If my memory serves me correctly, Buildzoid did a video a little while back tweaking his loadline and voltage settings, as the out-of-the-box settings had his 14900k spiking up over 1.5v during light work loads (ie. idle --> light work load).
yeah I suspect that this can be related to other complaints that the AVX offset was not kicked in under ultra light load, thus perhaps the memory intensive but light computing tasks such as minecraft server never trigger AVX offset, thus leading to ultra fast degradation compared to other gaming tasks
If the active core is at full boost but actually stalled on memory, despite not doing work the frequency is at max so max V which is then frying the ring bus with the L3 cache and memory controller on it.
It's because of the memory bottleneck keeping the core cool and the CPU cool, while non-processing fetch is at constant max rate.
Microcode changes might back off on voltage in these scenarios, which were set too aggressively to hit 6GHz on more dies.
Intel is going the Takata Airbag recall method. Replace faulty with also potentially faulty/degrading part to buy time.
Drag this on long enough for it to not matter as people move to other CPU's.
That would be a very bold strategy considering how it went for Takata.
Class action lawsuit waiting to happen if they don't come clean soon.
Intel proving why everyone should go AMD for the foreseeable future.
Just purchased a 13700k on prime day for $200 bucks, because people are returning them out of fear. 😂
I'll make sure all my voltage settings are good, and I'll be fine.
@@DingleBerryschnappsOxidation can't be undervolted out of. It's like saying that a building about to collapse can be solved by eating less.
@@DingleBerryschnapps haha . Good luck anyway, Im sure the 13700k will do well for long enough.
lul
@@bulletflight worst case they RMA it... $200 is a pretty damn good deal. Not all CPUs are affected.
Let’s go, Wendell is a fantastic guest! I enjoy both Level1Techs and Level1Linux.
EPYC combo, two of the most informed people on top of the pertinent subjects out there, TY both.
Wendell is a great guest! Always glad to see him.
Just in time to download for my flight, thanks Tom
Have a safe flight and enjoy man 👊👊
Computer janitor is a funny way of describing sysadmin troubleshooting
It's exceedingly common and honestly fairly apt
its funny because its actually true
be kindness my dear. not every humanitarian from india is janitor. many are xcelent four make the computer or cook the food or xcelent dancing or four xcelent branes.
@@RanjakarPatel हमारी त्वचा के रंग से आंका जाना बहुत दुखद है।
And now its combined with DevOops
I’ve only had Nvidia until Sapphire Nitro+ 7900XTX. I play competitive games high frame rate @1440p native. I’ve been super happy with it.
Thanks for this one, having Wendel on at this point is quitte a timely choice. Kudos!
I am an engineer who has debugged a lot of chip failure. The important information to find the root cause is 1. Does this fault occur at any clock frequency? 2, What is the FIT rate, and does voltage effect the FIT rate? 3. Where and when does this failure occur? (I assume the data is corrupted)
There are many existing tests that can isolate the problem. The most difficult bugs to find happens when the failure is not easily repeated.
The reason I am interested in the bug is, I worked for Intel and thought how easy it would be for me to sabotage the chip with a bug that can avoid the verification tests, I could create a bug that is near to impossible for a sleuth of engineers to identify. I could have created this bug and made money by shorting the stock. Makes for a movie, but I would never sabotage a chip.
What if Intel's rebrand is actually something they are doing in anticipation of the fallout of this Raptore Lake scandal? That is the only logical explanation for abandoning the i9/i7 branding.
@@fatidicusaeternus6498 Intel announced that before they knew about any issues
Oh hell yeah, Wendell is an all star guest.
Thanks for pointing out that Intel didn't take accountability for their laptop CPUs, I had forgotten their mobile lineup uses different model names.
Laptop CPUs don't use desktop voltages and clocks.
@@saricubra2867 And still fail at an unusually high rate. That's the sketchy thing being pointed out here, Intel has only admitted fault and promised a fix for a tiny sample of all the SKUs having issues which include mobile CPUs as well as "extra stable" server versions of otherwise consumer desktop CPUs.
@@saricubra2867 except the HX series. These are literally desktop chips (ie. 13900HX is essentially a 13900K); presumably with better binning (ie. marginally lower voltages for a given frequency).
@@heyitsjel worse binning but run at less-than-peak performance, actually
@@erkinalp this makes sense for the 13900hx (peak boost 5.4ghz), but for the 14900hx I'm not so sure (peak boost 5.8ghz)...
"Dave's Garage" have a good explanation of what happened with the Crowd Strike update.
He used to work for Microsoft and have a good knowledge of the ins and outs of Windows.
When I saw Bartlett lake, I immediately thought that is the Raptor Lake die fix.
But I feel there is another, far more important market for those than desktops or laptops. Intel launched the Xeon E-2400 series late last year. Those are also based upon Raptor Lake, but without the E-cores which are disabled. Now if there is somewhere Intel really needs extra P cores and absolutely no crashes, its there.
But how much power would those xeons need to be competitive?
@zorexlt8312 it's not the E cores that pull all the power out of the wall on those i7 and i9 chips...
@zorexlt8312 APO does not support any produktive applications that would run on Xeon servers.
And the Heterogeneous Architecture of 12th gen and later with a different ISA per core architecture is a stupid idea from the beginning in server space.
They need to get rid of the e cores anyway..
@zorexlt8312 APO does not include any produktive Applications that would run on Xeon..
my 14900KS loves running at suicide voltages, thanks intel.
lul
Intel plans on releasing a fix mid-August. I hope they're better at following software release targets than their manufacturing nodes.
biggest weakpoint with AI bubble is how narrowfocused stock market is... all magnificent 7 companies are basicly tied to this AI and pushing loads of money into investments and if it pops and revenue isnt enough, they bring down whole stock market. It is like leverage 100 for whole stock market by one single business use case.
@@nowayslowman I dont wish to see it pop, but human fomo seems strong to microfocus on one trend at time creating this weakness. Lot of other industries, aspects of life suffer from lack of funding and innovation now. Thus im criticizing why all money needs to go to just AI. Gamers are not convinced yet companies, studios can deliver something worth promises. Same is with regular companies.
Definitely not trusting Intel from that response.
looks like any rules created suddenly forgotten if its abt king, the powah he holts
I hope GN still goes ahead with their analysis on 14th gen.
There is a statement on Intel saying the oxidation issues were real and supposedly "addressed" in 2023.
4@@iikatinggangsengii2471
At this point AMD should do what the phone companies does when you switch from one carrier to another you get like a discount on a new phone, well AMD should give you a discount if you have a proof of purchase of intel 13th and 14th gen cpu's, give them like 15% off, imagine the flood 😂😂
@@siveric32 why would any buyer of a 13th:14th gen want to buy an AMD CPU instead? That doesn’t make any sense
Brings a whole new meaning to "Broken Silicon" with Intel. I never liked Intel(3% gen over gen lolz) or nvidia(80% price hike 30% performance. Feels like they sell software now.) I never thought I would see this. Intel chips dying like this? Sure older AMD processors(FX) can have issues, but this is a whole new level of crazy Intel. As for Nvidia they are an AI company now.(insert Michael Jacksons They Don't Really Care About Us)
Can confirm we switched a bunch of redis instances to graviton lately and it's been a huge improvement.
I love it when Wendell comes to the show.
Damn perfect timing for Wendell!
They raised the VID from 1.5 to 1.7 for the 900k parts. Why would suddenly 1.5-7 volts be okay for a CPU? They just could not produce enough good chips, some getting blasted since they run above 1.5v.
They probably expected it to take longer, but this simply can´t be a coincidence.
Server chips follow Intel guideline so they get blasted with 1.55+ vcore too. So, anyone who did OC their parts and dialed in the voltages to common sense will be far less likely to be affected. 1.45 was always the upper limit, better is below 1.4. I got super lucky my chip runs at 1.34 idle and 1.22 under load. Perfectly stable since launch.
My 14900k has a constant and stable shift between 1.3V-1.45V (max) depending whether I'm putting it under load through doing something intensive like gaming, rendering or some other kind of content creation, then down to 1.3V+ whilst mostly idle on desktop.
So you're an outlier with one of the actual good ones off the production line, or it's coming.
When literal firms who definitely know what they're doing with these chips, in motherboard that can't even OC in some cases, are having obtuse stability issues and failure rates, there is a problem. And not user error.
@@MaddJakd looks Like you don’t understand the problem at all.
Always love to see Wendell.
The humbleness from Wendell never goes unnoticed.
They may not be able to RMA them for two reasons. They dont have a fix that is fully tested, so they wont have any chips at all just yet. Secondly, how much physical effort would it be to replace everything and organise a fix on that scale? It would overwhelm them.
Yep, that's the problem. They have no problem RMAing stuff since that's booked as revenue lost rather than additional costs incurred unlike a mass refund (makes shareholders much less angry despite being more or less equivalent), but they need to have a series of known-good CPUs to replace them with. If Raptor Lake really is fundamentally flawed, Intel themselves understandably don't want to play Raptorian roulette with the replacements either.
@@hammerth1421Raptor Lake isn't the issue, Intel is. Since 8th gen they pushed max CORE VID voltages to the Infinity.
13th gen and 14th gen expose the problem because of the single core turbo boost started to increase since many generations ago.
and the retail where I bought my 13900KS, and basically all retails in my region, don't give even half a fudge about RMA. they even put out a 'store warranty' just so they can have lower pricing. So I'm left with the option to suck it up, sell it to someone gullible, which would be really difficult since the majority would only buy entry level and those who would are knowledgable enough they might already caught up with this drama, OR switch to an x3d system which is practically building a new PC for me.
@@donloder1 Downgrade from an i9 to an i7.
Someone needs to call out "framechasers" the hypocrisy coming out of that guys mouth is unbelievable.
He's abrasive and unapologetic, but i do believe he's claimed voltage its at the core of the issue from the beginning.
He's definitely an acquired taste.
Don't forget FC is the same grifter that has stated that T CPUs don't degrade, yet MLiD on another video has stated those too fail.
He said it was single core boost where the single core on the preferred core is boosting to high and then threw Wendel under the bus
you mean the dude that sells 1k CPUS and then gaslights the consumer on thinking it’s their problem?
It's like "it's not Lupus!" until it is Lupus... 😅
Wendel has interest so widely spread, he's always fascinating to hear.
i love that you follow the empirical approach
All of this testing used to be pretty standard in all tech companies.
Most tech companies have significantly downsized their QA teams.
Double checking things takes time and money.
Until a company like Crowdstrike is required to have an insurance policy that covers trillions of dollars of damage for bringing down the internet, that money is not going to get spent.
Anybody remember the TSMC 7nm degradation scare mongering? I have a been a user of Intel, PowerPC and AMD since the P3, G3 and Duron days and this is the worst case of CPU failure I have seen.
Intel updated their release to admit the 13th gen has oxidation issues. And yes they're very sneaky. Waiting till after Zen 5 review window is up for the "fix" as well as simply not being forthcoming about what other issues compounded the problem. And they should recall the 13900K's with oxidation at least. But they're likely hoping they last till warranty is up before they completely die.
Where have they admitted to oxidation issues? The only claim of oxidation is from Steve. Zero evidence yet. So far most of the evidence points to motherboard sending them too much voltage. 🤷
@@DingleBerryschnappsin their statement they admitted there was some oxidation issues previously but said it has been fixed.
@@DingleBerryschnapps They admitted it was microcode bugs causing the CPU to request for incorrectly high voltages, not just the MB messing with settings and sending high voltages. The problem is on Intel for providing bad microcode.
They also admitted, in a hard to find place (PR communication on Reddit, not the original statement), that some of the 13 gen CPUs were indeed effected by an oxidation via issue that they addressed in 2023. They basically confirmed that the oxidation via issue reported by tech outlets was indeed true, and they fixed it sometime in 2023. Why they put this section on Reddit and not their original statement about voltages is anyone's guess.
38:50 Perfect "Um" synchronization
My HP Omen has been rock solid with 14900k, 4090, XMP 5200, locked cores, performance mode, I hope it stays that way. 64gb, 16 x 4.
They had to present a cause/solution so close to the launch of Zen 5, regardless of whether it stated the actual cause or just another possibility. Research into the cause would never have taken so much time if it was simply an error in the microcode.
Who said it simply an error in the microcode?
All evidence points to motherboards sending too much voltage to the chip.
How many configurations are possible in a desktop system? With the near infinite amount of possible configurations, I'm sure it takes quite a while to figure out the root cause for all of the issues. Especially, if the symptoms for each CPU can be different from chip to chip. There is too much speculation and very little actual evidence in any of these videos.
@@DingleBerryschnapps Did you even watch the video?
At 0:48 An Intel employee stated it was an elevated operating voltage stemming from a microcode algorithm giving incorrect voltage requests to the processor.
So yes, too much voltage 'allegedly' - caused by a faulty Intel microcode 'allegedly'.
The employee also stated an updated microcode to correct the 'alleged' issue is going through validation.
Guess time will tell.
@@gbnq2513 do your Research. It was likely a miscommunication. Instead of proving the maximum safe voltage when required boards sent the maximum possible voltage. That’s why on Intels own boards this never really happened.
It’s similar to how you can set the powerlimit to a max of 4096W. But that will never hit the CPU.
The Mainboard manufacturers weren’t directly responsible for that but in the end the Mainboard did the wrong thing. On mainboards with the right setting this never happened
It doesn't explain Gamers Nexus's oxidation issue and neither does it explain why it took them over 6 months to come out with the bulls*t excuse.
simply use 'its lie' or 'not for you' then all good
I’ve been following this chip problem pretty closely. Huge respected to both you and Wendell.
1. The simplest fix that I can think of right now would be to roll critical updates with windows that nerfs the voltage, just as they did with their security mishaps back in 2019. But that isn't going to be simple either. Introducing a new stepping would not affect existing users, while forcing AIB/retailers for massive RMAs is going to be absolutely ridiculous.
2. As someone who had worked on Silicon chip fabrication, I'd say this is most likely dielectric breakdown. Some of the oxides might have been grown too thin (in bad batch/batches) and slowly breaks down due to high current.
3. I'm certain that there was a huge disconnection between the engineering team and the product team. This is basic knowledge from the 50s, and manufacturing/industrial statistics are always taught in the 2nd year of college. Now, the entire company has to take one for themselves, yet again, in a situation where they really need not to screw up.
Oh dear.
@@EntekCoffee wasn’t really an Intel mistake, more like a miscommunication. The Mainboard manufacturers allowed too high voltages because they got Intels spec wrong (and that’s pretty much on Intel). Instead of providing the maximum safe voltage they provided the maximum theoretical possibility voltage which was too much. That’s why it didn’t appear in Intels validations.
Looking good Tom.
Just clicked, man this is going to be a great one!
Thinking about Medusa, making chiplets work for laptop and desktop may be a chip stacking strategy. You have to go through the L3 MMU before you go to physical ram. Integrate an L3 3D Vcache and an IO die and then stack the chiplets on that. The IO die has the driver circuit that physically drive the external memory bus, All off chip access has to go through that anyway.
Plus this gets you CPU cores closer to the heat spreader.
An addendum at the beginning, also known as a preface lol. Love you, Tom
The worst thing that could come to light is if there was an underlying problem with the Intel process nodes.
prob just simply selling bad chips
Hope that AOC mini-PC has an AMD option..
It's Alder Lake and E-Core only. Unaffected.
1:32:20 I am working in Fusion 360 and Blender. The experience with AMD GPUs improved drastically over last 2-3 years. In Fusion 360 it improved so much that I can get the job done on R5 5500U laptop with 8GB of RAM. And I work on no small projects! When not on the go, 7900XT just flies through everything. HIP support improves, the newer the Blender version, the faster RDNA 3 gets
People may think GPU doesn't matter for CAD but yes, yes it does. Especially in bigger projects. VRAM often is the limiting factor. With complicated geometry, GPU core also has a lot to do when working in viewport. The same is true in Blender. This gets better with a push towards HIP implementations and optimizations. Now 7900XT is getting where it should be based on in-games RT performance, so at the level of 4070, 4070Ti. The time you need to render an image is one thing.
But to get everything set up to the point you actually can render an image or video, and do it fast, efficient and without issues is another thing. For rendering, 4090 is a king, no doubt. OptiX helps it a lot. But when comparing lower end NV GPUs, the performance delta is smaller than you'd expect now. To the point where it becomes somewhat irrelevant for smaller companies and individual users. When working in a viewport, so in 90% of time you actually work as the user, 7900XT is doing surprisingly well too.
Plus, when you compare price competitors, 7900XT is doing amazing. In Europe, for the price of really good 7900XT you often can get basic 4070 Super 12GB. High end model or basic 4070 Ti if you are lucky. So again, not only 7900XT isn't really slower than these anymore. When RTX runs out of memory, suddenly 7900XT is faster and also saves a metric ton of time in post-processing. I love how consistent performance I get from this card. And the funny thing is, if or when I need more performance in Blender, I can just add second 7900XT. Not only rendering performance will roughly double, so they will be faster than 4090, it will be in a similar power budget while costing less (also true for 70 class NV cards) and having almost double the VRAM (not true for NV). Before anyone asks, in software like Blender you don't need SLI or crossfire. Not only software splits the work by itself, it also pools the memory. Meaning you can use all the VRAM you physically have on a GPU. And it can work with different GPUs, no need to match them. Though it comes with some inefficiencies.
I'd be surprised if this isn't a tolerance / threshold issue with regards to die binning or acceptable silicon level defects/variances for a given SKU. ie. they're possibly pushing the limits of what is acceptable silicon quality for something like a 13900k or 14900k, in hopes of optimising their yield.
For context, I've had my 13900k running since November of 2022, at DDR5 7000 - it could handle 7200MT stable, but required additional voltage, which wasn't worth the benefit IMHO. Recently I shuffled those sticks over to the wife's system (12900k), and went to a 48gb kit at DDR5 8000MT. It's been running at 8000MT without a hitch, at even *lower* system voltages than the other kit required... and for the record, I run various long duration torture tests so it is genuinely stable (Linpackxtreme; OCCT; TestMem5 etc.). Why is this important? It suggests that *not* *all* CPU's are affected per some major design flaw, but more likely something relating to fabrication (such as incorrect design tolerances for silicon level defects/quality).
Same. I've had my 13900k running at 5.8p/4.5e/5.0ring with 4x8gb of 4200c15 ddr4 on a cheap MSI Tomahawk since November 2022 with zero issues. I like it more than my 7800x3d system.
@@jjlw2378I've actually been wondering if *earlier* CPU batches were less/not affected, and as demand has increased they've loosened the tolerances to meet demand and/or had the oxidation issue. This could also explain why it's taken a while for this issue to rear it's head en-masse...
@@jjlw2378also, I too was running an MSI board (MEG z690i Unify, ITX)... I've filled out the GN survey - it would be interesting to also see if there's any correlation with board manufacturers.
With the MSI I was typically using the 288w "water cooling" profile, as it's typically been enough for the 13900k. Going to unlimited power profiles maybe gains you 5%-10% more performance in synthetic all-core workloads, for an insane increase in power consumption and temperatures.
Recently I switched to an Asus z790i, which by default had *unlimited* power profile settings. I quickly installed the latest BIOS with the less insane default power profiles... so this too could be part of the reason our CPUs aren't wrecked.
Fixed within a couple of months? Glad i got an AMD CPU :)
Friends don't let friends do Intel
Yet, the Intel apologists are saying that... see it's just a microcode fix... It'll be fixed by August!
August starts next week...
@@jorgevaladas788 The thing is. it's not a final fix. Its just to lower the voltages even more. Lowering just prolongs your CPU and lowers your performance more.
Just in time for all the Zen5 benchmarks to be completed.
@@pkpnyt4711 didn’t age too well
@@lukas_ls what didn't age well, that they just released a patch so it doesn't degrade as much anymore? 50% of chips out there (13th and 14th gen) are all degraded. How do we know, Intel just said that 50% of the chips are not affected by degradation. You call that a fix? Are you daft? If this was any other company there would have been a recall but for some stupid reason fanboys think it's fine having 50% of your total sold chips as defective.
Facebook said they used 16,000 H100s to train the latest llama model. Then gave it away for “free”
I think Intel has always known that they're pushing their CPUs too far, they just couldn't predict how long they'd hold up. You can see it in their ridiculous all core boost behavior, where the only all core bench they manage to really win is CB if you do a single run. If you do a 10min test the 13900k score plummets well below a 7950x and the 14900k is only slightly ahead, closer to tied. Both lose outright in Blender and decompression to the 7950X. Any kind of real all core work is typically going to run past that boost timer, so it's almost like it's designed solely to win CB sprints.
AMD had Opteron A1100-series ARM64 (Cortex-A57) maybe ~7 years ago. It was really an interesting one, especially since majority of new ARM SoC/IP companies didn't have previous real life experience in the datacenter land. AMD on the other hand knew what needs to be done.
1:26:55 if I was asked I'd answer - gaming laptops. Their H and HS CPUs are so good, both showing good performance and fantastic battery life(due to them being the same chips used for u-series CPUs), but they are barely present in laptops, especially high end laptops.
Imagine a Legion 7 with a 175w 4090M GPU and Ryzen 9 7940HS. It would be capable of 4K gaming and would have like 10 hour battery life thanks to Ryzen low power consumption at idle and 99 watt hour battery Legion 7 has.
But for some reason, such combinations pretty much don't exist. And I don't understand why. Thanks for the stream, guys!
FYI if you're buying an AM5 APU that has PCIe 5.0, 8 lanes of 5.0 is basically the same bandwidth as a full channel of SODIMM DDR5 if you were to enable CXL, and then, say, give the APUs 24 lanes like the CPUs, you could then do a full 16 lanes of PCIe that is almost as fast 2 channels of desktop DDR5 5200 with tighter timings.
There was a dev board that was available with that AMD ARM k12 chip floating around during the Zen 1 years. I wish I would have gotten it but it was expensive at the time and software compatibility wasn't there yet except for a homegrown Linux distro.
the problem is boosting 1 core to 6ghz which uses to much voltage 1.5v or more , solution is to lock all cores to 5.5 or 5.6 ghz also voltage 1.35 - 1.4 max
This is what I have read also.
Tom, I remember you saying that everyone at AMD is amazed how well the Radeon division is doing with it being that small compared to NVidia.
I can smell class action
Kinda surprised that Wendell's camera feed looks crispier than Tom's.
Is it just focus? Or does Wendell actually use a better camera than Tom?
This podcast was a slam run!!!
38:52 that synchronized uhm is so funny
Lol I didn't even notice that during the recording
Sounds like an Intel employee lied to the public with that statement. I wonder if Intel will throw him under the bus.
its been months now, If AMD have done what Intel PR its doing everyone would be cooking them , I want to disagree with you on this one, for me its not that Intel don't know what its going on, they do very well, but what they are doing its gaslighting, they want people to speculate, First they blamed Motherboards, then the the TVB bug, now its voltage bug. That makes them look like they don't know what its going on, that they are "investigating" because as you said they cant RMA everyone chip so make time, it could be that anything mentioned before could be part of the issue of a grand scheme. Remember when AMD was blowing up CPU they own it they make it transparent they didnt throw OEM under the bus and they fixed the problem. I wish that Intel as a BIGGER company with more resources could do better, they should.
i shall wait more
Would you estimate that Ryzen 9000 will be free of any such issues, such as CPU instability and AMD's former exploding CPU's and melting motherboards?
@@JynxedKoma No company its perfect, but you didnt see AMD gaslighting everyone, in fact they took things on its hand and did what its right and in the end it fixed the problems, nvidia did the same to some extent with their crappy "user error" connector, meanwhile Intel its denying RMA and not taking responsability at all.
@@eilegz Just to clarify, I'm not weighing in on the former issues AMD might have had. I'm just asking as a first time AMD adopter of AM5 + soon 9950x once it releases in august, if I can be sure the chip won't suffer any instability or heat issues. If at all you have experience with that using previous gen AMD chips + motherboards since this past year. Thanks!
@@JynxedKoma and they just delayed ryzen 9000 for quality reason.
this was great, thanks guys
What will come first, Tom's long hoped for and awaited 'AI Bubble Pop', or the also long awaited 'Year of the Linux Desktop'?
Halflife 3 will be out before either lol😂
I dont believe their statment
I just had to do an inventory recall on Amd Ryzen a few days ago, just a heads up
I'm running a 13700K and have had zero issues. This is scary though
You have to wait 5000+ hours of use which is terrifing
Im glad that you finally invited someone that adresses the Linux side of things , I use linux on my work PC and on my laptop (i only use windows for my gaming rig), so linux is a big deal to me
Wendell @58:00: "...no matter how you cut the cheese..."
😂
... it always stinks?
Is there any information on the 13th and 14th generation systems that use DDR4 if they are affected in the same way as DDR5 systems?
2:13:13 Actually Apple publicly stated that they were open to getting Windows on ARM running on Apple Silicon, it's just up to Microsoft to agree. Microsoft so far only officially supports WoA on Apple Silicon via Parallels virtual machines though.
Hey Tom have you heard anything about the supposed "Bald Eagle Point" Strix Refresh with the extra 16 Mb last level cache?
Intel have gone hard into the silicon lottery concept in the worst way.
if intel issue gets big ,people will loose their trust in them, intel will loose their customers, and i hope AMD won't raise their prices in wake of no competition
lose
That is absolutely not but a smaller percentage of the issues of intels cpus. I would be outraged if i hadnt realized that AMD is the move for my last build. Everyone else is seeing this now. Ryzen will dominate. Even moore!
Only 4 or 5 years? My old laptop gave me 8 years of good usage and is still usable, but feels sluggish with its 4c/8t (but 32GB of DDR3 and 2.5TB of SSD storage was not an issue). I'm hoping my 12900H is going to last as long (hopefully my next laptop can be AMD)
Powertop auto tune... and now my USB keyboard and mouse doesn't work 😂
Major question, do they even have the information to detect whether a specific CPU is indeed faulty -__-
Intel has to stop pushing overvoltage out of the box with an overclock and no powe/current limits with their motherboard partners.
That CORE VID of 1.5 volts is embarrasing.
Mine never goes higher than 1.45V for the 14900k.
Frame chasers claiming everyone is lying about intel 🤣🤣. He's a super Intel fan boy.
a 700 USD card with 32Gb Vram is something i as a gamer would buy. it would age super well.
Intel doesn’t want to be genuine and truthful about this whole thing because of the hole they’d put themselves in by admitting it all.
And to say it was a foundry problem. That would hurt the foundry business they’re trying to drum up.
So both ends of the company are not firing on all cylinders as they’re leading everyone to be.
Products always massively delayed and produced at other fabs currently.
i think it is as simple as INTEL is having a hard time competing with AMD and mistakes are happening... pressure causes mistakes. AMD has been applying non stop pressure.
The problem with Intel started before AMD was really back with the first attempt of the 10nm node, goals set too high and no-one stopped the whole thing and asked for a re-evaluation of the goals.
Load line cannot me Mega OHM. More like milli OHM. Love this chat....
i bought a 14700kf and had the tekken 8 vram error early on, i was using intel extreme utility at the time and then uninstalled it, as well as adjusting nvidia vram amount, i no longer have any issues with tekken 8.
this is how the issue started but is now server related and is the log indicating nvram issue as an error for warframe?
I hear hot stuff and it brings me back to the Chippendale skit on SNL. I bet Wendell would be a shoe in for a Chippendale dancer.