First build: Intel vs AMD in the wake of Vmin shift instability
Posted by automaton11@reddit | buildapc | View on Reddit | 16 comments
I'll keep this short because its another blue vs red post amid an ocean of blue v red posts. I was planning an Intel build with a 13700KF or 14700KF for a workstation that will have to wear many hats: programming, gaming, some AI stuff, and music production.
Plan was for Intel because a lot of music stuff is optimized for Intel. I've read about people having significant problems with AMD and virtual instruments, with AMD chips not supporting iLok and thus excluding the use of huge swaths of music software which will not work for me. There are also DAWs which reportedly have issues on AMD systems (Cakewalk?).
Then this Vmin instability issue comes out. There is a BIOS patch but that doesn't really address the cause of the issue, which sounds like its fundamental to the chip itself. As I understand it, this Vmin shift instability is typically observered in healthy chips that have been subjected to extreme conditions, meaning, to me, that for these chips to demonstrate this issue after normal operation, they must have some systemic terminal disease. How can I spend hundred of dollars on a chip I know is permanently and fundamentally flawed, and will need to be underclocked for the rest of its life to prevent burn out?
As of now I suppose I'll wait for 15 gen intel. Anyone can weigh in with any opinion at all. This is my first build and idk shit about fuck
JTG-92@reddit
You’re assuming that everyone is plagued by the idea of physical oxidation causing degradation, that is a non issue now. That effected a small bunch of the earlier releases of 13th and 14th Gen, that in combination with the current voltage issue would have resulted in those chips failing, being RMA’d and no longer in circulation.
That was the physical and unfixable form of degradation nobody wants and now nobody will have to worry about either. The current issue with voltage is not as widespread as you think, nonetheless you also definitely do not have to gimp your system by Underclocking it.
Do the microcode update, tune your CPU, like anyone should if they own any of the 13th and 14th Gen i7’s and i9’s and you will get a proper useful life cycle out of it. Anyone who is purchasing brand new and follows some very basic rules, that are not a burden, will have no issues. We’re at a point now, where it’s not rocket science, as time goes on toward the end of the year, this will calm down. People who are having issues or have complained about having issues, are all people pre microcode update.
I bought a 13600k 2 or so years ago and it’s been overclocked with all those savage Asus enhancements nobody recommends now and it’s immaculate still. I then bought a 14400, which in all fairness is a 12th Gen but its voltages are totally ridiculous, they are so low, even with all the Asus MCE settings on.
When the microcode was released, which I intentionally waited for, I went and bought a 14900KS and while it’s still early days, I can say that nobody in there right mind would throw this in and leave it like that.
The voltages are disturbing right out of the box, but it is also targeted as an enthusiasts CPU, so they expect you to tinker. What I’ve learnt after mucking around with different settings, is that there is a huge range in terms of what voltage is actually required vs what it will default to.
I think people forget that we’re talking about some of the highest clocked consumer CPU’s to exist so far, higher voltages come with that territory and it should have been expected. The thing is, you can change all of those things with mostly no performance loss, what I learnt too, was these CPU’s actually perform better, the cooler you can keep it, even if reducing voltage leads to being cooler, the performance actually goes up and it scales for ages.
At the end of the day, you do you, and it’s not that I’ve been lucky or anything either, this is majority of owners experiences but people only talk about the negative and not the positive. If you want a 14700k, buy one, from a financial perspective, that is the best model for the money and I knew when I bought my 14900ks, it was just pure desire and like throwing cash away in terms of the poor value. I’m just saying that everyone has gotten so paranoid about all of this, that it completely debilitates their ability to rationally understand the reality and proceed with their purchase. If it’s of any interest to you, 15th Gen won’t clock this high, it may perform slightly better but if you want to grab the last of its kind for raw dramatic power consumption and just extreme characteristics in general, this is your opportunity. That might mean nothing to you, but you should be able to understand the thought process anyway, it’s why I decide to just go all out and get the KS, because it’s literally the very last of its kind.
automaton11@reddit (OP)
I dont think you're on the right track regarding your assuredness that the current Intel issues are not related to physical degradation. The problem has only been described as 'Vmin shift instability.' Vmin is the minimum voltage required to activate the structure, which is itself a series of MOSFETs. If you look at the structure of a given MOSFET, you can see that the threshold voltage Vₜₕ is determined by the thickness of doped regions in transistor. For this threshold voltage to increase, thickness of the gate oxide layer or channel thickness or some other physical characteristic of the structure would have to change.
So when Intel says that the Vmin or minimum voltage required to activate a structure is increasing, this means that the structure itself is *degrading* requiring higher and higher threshold voltage for activation. And if this degradation is occurring at nominal voltages, this indicates a fundamental design flaw, not an issue with microcode or something. The chip basically has parkinson's disease - and normal 'environmental' exposure causes deterioration of essential structures
The other type of Vmin shift could be an issue only with the voltage regulation system. In this case, the logical parts of the chip are *not* deteriorating excessively when exposed to nominal voltage. Instead, it is the regulator that is erroneously providing incorrect voltage, thus over-volting the structures and causing damage. As the damage accumulates, the MOSFETs require higher and higher Vth to initiate leading to Vmin shift.
Either way, the chips degrade on their own. They are not structurally sound, and need to be 'babied' in order to not cook themselves. The microcode patch is not a cure, its giving a cane to a cripple
JTG-92@reddit
To be honest, I can see how you got to that conclusion but I think you’ve missed exactly what they mean by it.
The Vmin shift, is not a physically shift in terms of the cause itself, as in from a manufacturing defect like oxidation.
The Vmin shift is a natural path of degradation, it’s not a direct cause, it’s the consequence of high voltages and high temps.
The best way to describe it, is like someone using cocaine, only to find out they require more to achieve the same high as before. That in itself is a build up of tolerance, causing down regulation of dopamine at the receptors, it’s the whole idea behind that saying “chasing the high” because you will never completely get back to that first time again, at least not with the same amount.
You could view that as degradation at the receptors, where it differs to a CPU, is that eventually dopamine up regulates bringing you closer to baseline again. A CPU on the other hand will keep shifting in one direction, the only thing you have control over, is how fast it moves.
This degradation is apart of every CPU’s life cycle, but in relation to what Intel is specifically meaning by it, is that due to a combination of things and they all relate back to temperatures, amperage and voltage. Whether it’s partly due to motherboard partners, an incorrect algorithm in the microcode or for the rare few who did unluckily end up with one that did suffer physical oxidation, in the end, they have all compounded off one another.
Ultimately the major issue with the algorithm, is that without the update, the actual voltage additional applied could never be accounted for, which means it could never be planned for or compensated for. So when the entire system thinks it’s already running within its max boundaries, it’s in fact overstepping constantly without realising. Safeguards can’t protect against an issue like this without the correct calculations made from the algorithm.
If the algorithm was fine all along, this whole issue may of actually not even happened and if it still managed to, there would have been significantly less reports to an extent that barely anyone would of heard about any issue.
The physical oxidation was corrected last year, whatever was in circulation would have rapidly degraded and been RMA’d to a point where there would be almost none of them still floating around in the wild from what was only a small number to start with. If you take that as now an assumed non issue for 99% of people, you assume everyone making a new purchase, started out with the updated microcode, it’s completely reasonable to believe everything will be fine and even if we wanted to call it a band aid, it’s likely a 10 year band aid, which may not be a 30 odd year lifespan of a CPU but it will have outlived its usefulness to most.
You’ll see how reports slow down, they will have already, the only people who are now in line for an RMA or bitching about the degradation they have already suffered, are all pre microcode. And unfortunately that’s not reversible, it sucks for them but I’m not talking about done deals, I’m talking about new buyers with a completely different start to everyone that came before them.
We can argue that we don’t know if it will be enough yet, but by the time it’s been long enough to know that for sure, we will all be onto 16th Gen and none of it will even matter anymore. But the same argument could be made, that by not buying now, you could waste the next 2 years not enjoying a CPU that very well might be perfectly fine still and you would have missed out on the last of this kind of CPU, for better or worse.
In the end, they refer to Vmin shift, not because there’s is a physical aspect causing and adding to the cause, but because due to the voltages and incorrect algorithm, it causes Vmin shift as a consequence and it’s also a nicer way to say the word degradation to the public.
automaton11@reddit (OP)
I didnt realize that 15th gen is expected to have less raw power than 14th gen. How is that going to pan out for intel? Do they really think people get excited about power consumption efficiency? People building powerhouse workstations dont care about saving 60 bucks a year in power costs, they want something that screams. If this really is the direction Intel is going, its time to sell off stock
JTG-92@reddit
OMG!!! Thank you!!! finally!! Someone who can see it from my perspective hahahaha this whole efficiency thing gives me a headache, the savings in electricity mean so little, it’s not even worth mentioning. The thing is, performance wise, they will still perform better and better but I enjoy watching it crank past 320w and steam roll along like a freight train, it’s exciting.
automaton11@reddit (OP)
I think its like a stewardship thing. Its not about whether you like it or not, its about being green and reducing energy consumption. Which is stupid - the major impetus of global carbon issues is not consumers, its large companies and nation states. Oil companies sold the idea that to be green meant to turn off your LED lights when you're not home and everyone bought it. Meanwhile the entire right hand side of the planet burns coal indiscriminately. Its fucking blame shifting, give me my hairdryer cpu
JTG-92@reddit
Bahahahaha I love it! It’s so true, the world is becoming so excessively politically correct too, we’ve allowed so much freedom and acceptance of speech. And now it’s at a point, where we shoot ourselves in the foot once we have anything to say against things like Greenpeace.
zephyrinthesky28@reddit
If you don't want to go 12th gen, anecdotally it seems like the 13600K/14600K have lower failure rates.
Less performance and not enough data yet to know if microcode updates will prevent new chips from degradation, sure. But Intel does provide a warranty that should last until the middle of their next socket.
That said, I keep seeing the 12700K on sale these days and that CPU is no slouch by any measure.
automaton11@reddit (OP)
I mean if I went with 12th gen id just get an i9
KoldPurchase@reddit
I found this thread:
https://community.amd.com/t5/processors/9900x-cpu-makes-music-production-impossible/td-p/703995
It's a software issues that is affecting ZEN 5 cpus and is being fixed. Most seems fixed with an hotpatch, more on the way.
If you can, I suggest you wait a couple more weeks, even more, and check back on this AMD thread I linked.
It's iLok not supporting the AMD cpus, not the reverse, so it's easily fixable.
Intel... Well, you really don't know if the BIOS will fix the problem. It may or it may not. The good news is the warranty is extended for another 2 years.
automaton11@reddit (OP)
Yeah, thanks for this. I guess my issue with AMD is the fear that I could regularly come up against walls like this, which would be very frustrating after spending so much cash building a machine for expressly that purpose.
I would imagine that, especially in the wake of this intel issue, software companies would be eager to build in cross compatibility. But who knows, esp with legacy stuff
I dont think theres anything else (gaming, vms, programming, stable diffusion, EE and image productivity software) I'll be doing that would suffer any impediment from using AMD over Intel that I know of
KoldPurchase@reddit
It's a niche software.
AMD processors have a reputation of being good for gamers and my guess is, the software maker did not test their softwares on Zen 5 cpus. They had Intels and Zen4 and did not bother testing it on a Zen5 figuring it would work.
Now they know...
The problem I see would be if you decide to upgrade your CPU in the future to, say a theoretical, "Zen 7", maybe that bug will come back... So you will have to check carefully before upgrading the computer.
In as much as working for everything else, I don't see a problem, but I just read through forums, I'm by no means an expert. Programmers tend to favor Intel in the past because it was better at multi-tasking, but a Windows 11 bug was recently fixed that affected some issues, so maybe this has changed.
Gaming is better on AMD, more speed for less power, that's for sure.
mildlyfrostbitten@reddit
don't buy anything later than 12th gen right now. the nature of the problem makes it impossible to definitively say if any fix works without long term testing, and similarly if any new child don't have problems too.
automaton11@reddit (OP)
So assuming I need to go Intel over AMD for the music software support, I'm either buying a 12 gen chip, a locked chip, or waiting until arrow lake has been out for \~4 months. Does that sound about the size of it?
UgotR0BBED@reddit
Do you need an unlocked chip for the productivity software that you mention? Maybe consider the non k/kf sku's?
Alternatively, I don't believe the issue is affecting 12900k, which might be something to consider. This issue arose multiple years into the LGA 1700 lifecycle, I'd hate to see you be an early adopter on the next platform only to have something similar impact a new build down the road.
Intel is a dumpster fire right now unfortunately.
automaton11@reddit (OP)
Well the mentality behind the unlocked chip is simply that this pc will be my main machine for a long time and I want some wiggle room as far as power goes. Considering itll be running a server and some vms, yeah I wanted to be able to play around with clock and tune.
But you're also right in that waiting until 15th gen is released is not really sufficient without a buffer period to ensure the issues are resolved. With intel being so cagey about the actual technical cause of the issue, how can I trust that its actually resolved?
If amd had adequate support for music production stuff this would be an easy call. I really cant believe there are entire software suites that wont run on amd. So the world just relies on Intel to not suck, which it suddenly does. Pretty frustrating and stupid