GPUs & CPUs & Enthusiast hardware: Questions, Discussion and fanboy slap-fights - Nvidia & AMD & Intel - Separe but Equal. Intel rides in the back of the bus.

I mean its not just the voltage issue. They also have an oxidation issue and intel isn't actually sure if they have the problem figured out yet.

There's more to this than just high factory boosts. Intel really fucked it.

I just find it amusing that this happens the one time I think about trying intel in 15+ years. Apparently its a sign from above.
 
They're getting down to atomically small transistor gate sizes. 13th-gen and 14th-gen are 10nm and 7nm, respectively.
They use the same process node, it's just that Intel renamed it to 7N to be more in line with TSMC's naming standards. The numbers of modern process nodes are purely marketing devices that don't reflect any dimension of the transistors.
 
  • Informative
Reactions: Drain Todger
My NVIDIA 1070 is shitting the bed. Fan not operating at all, 50+ c on idle, and I'm getting TDR failure all the time. I'm cheaper than a Jew, so I don't want to waste a thousand dollars on something brand new. Been looking up GPUs and Amazon's offering a refurbished 1070. Reviews are good but I'm not sure. Anything my kiwis could recommend?
How much is the refurbished 1070? You could probably get a better GPU for the same price in 2024 if you're willing to go used
 
So it's safe to get a 12th gen intel cpu? I'm shopping for a laptop and there's a good deal on one with a 4050 and 12th gen intel cpu.
Yeah I'm running an Intel I7-12700F. Much more stable then 13 and 14th gen. Should be fine for the next few years
 
Maybe it's about time for Intel to explore carbon nanotube based processors or something?
 
  • Optimistic
Reactions: Vecr
They're getting down to atomically small transistor gate sizes. 13th-gen and 14th-gen are 10nm and 7nm
Keep in mind that those measures don't really have any relationship to actual physical dimensions, they're just weird holdovers that we use for naming process nodes. Intel 7nm actually has a gate pitch of 54 nm.

I mean its not just the voltage issue. They also have an oxidation issue and intel isn't actually sure if they have the problem figured out yet.

There's more to this than just high factory boosts. Intel really fucked it.

I just find it amusing that this happens the one time I think about trying intel in 15+ years. Apparently its a sign from above.
It's not even necessarily just the problems themselves, it's Intel's response to them. Customer expectation is that once a CPU is up and running in a system, it should basically be 100% reliable. Intel not only took a long-ass time to respond but now the gist of that response is, "We can try to keep your CPU from dying but it still might just shit itself soz"

If Intel had already identified the affected processors and opened up RMAs so people could get a replacement mailed out in advance while they're still using their system, it'd be one thing, but Intel's position is that you have to wait until your processor dies to get anything out of them. I'm sorry but no one wants to deal with downtime like that, especially not business customers (which are the only thing keeping Intel relevant).
 
Keep in mind that those measures don't really have any relationship to actual physical dimensions, they're just weird holdovers that we use for naming process nodes. Intel 7nm actually has a gate pitch of 54 nm.


It's not even necessarily just the problems themselves, it's Intel's response to them. Customer expectation is that once a CPU is up and running in a system, it should basically be 100% reliable. Intel not only took a long-ass time to respond but now the gist of that response is, "We can try to keep your CPU from dying but it still might just shit itself soz"

If Intel had already identified the affected processors and opened up RMAs so people could get a replacement mailed out in advance while they're still using their system, it'd be one thing, but Intel's position is that you have to wait until your processor dies to get anything out of them. I'm sorry but no one wants to deal with downtime like that, especially not business customers (which are the only thing keeping Intel relevant).
Yeah the whole situation is just fucked from all sides. People unsure if their cpu is screwed and when people contact intel about instability apparently get a variety of solutions that might just be slowing down the inevitable decline.

Intel 13th and 14th gen will probably go into a "worst cpus" list, or at least intel's handling of it.
 
I just want a 3050 6GB for $100. Or you know, less (*rubs hands Semitically*).

Maybe it's about time for Intel to explore carbon nanotube based processors or something?
Nah, they just need their new process nodes and to not push power to the Moon. They can also outsource some "tiles" to TSMC. Lunar Lake is all-TSMC.

I think they could use an answer to AMD's 3D V-Cache that doesn't involve turning the power up to 11. They have Adamantine which sounds like an L4 cache, much better than the old Broadwell+ eDRAM, but it is entirely missing in action so far.
 
A bug in the firmware causing the chip to melt itself does not instill confidence in future processors, regardless of what node they're fabricated on. Time for the whole world to go back to IBM Power, I guess.

The 3060 that performs nearly identical to the 4060 and has 12GB of VRAM is around $200

The 3060 is about 15% slower, runs about 50% hotter, and doesn't support DLSS4 frame generation. Frame generation is a big uplift in newer games.
 
A bug in the firmware causing the chip to melt itself does not instill confidence in future processors, regardless of what node they're fabricated on. Time for the whole world to go back to IBM Power, I guess.
Or you know, the obvious choice, AMD.

But I don't expect this to become a regular occurrence out of Intel. They will be wary and pay more attention to quality and especially voltage issues. They are moving desktop and mobile to chiplets, which could help them.

Has anyone mentioned a class action lawsuit yet?
 
  • Like
Reactions: Brain Problems
Time for the whole world to go back to IBM Power, I guess.
I vote we move to DEC Alpha, as we should have from the start. You don’t want IBM to have a stake in the PC market, lest we get a repeat of the System/360 debacle.
 
Yeah the whole situation is just fucked from all sides. People unsure if their cpu is screwed and when people contact intel about instability apparently get a variety of solutions that might just be slowing down the inevitable decline.

Intel 13th and 14th gen will probably go into a "worst cpus" list, or at least intel's handling of it.
Intel was just testing out their new subscription based hardware. You have to subscribe every year or so by buying a new cpu.
 
I just want a 3050 6GB for $100. Or you know, less (*rubs hands Semitically*).
I'll mail you my used RX 470 for postage + $5 lmao.
You're all wrong - the solution is to move to Itanium.
Elbrus ISA works just fine. It's just vaporware outside the CIS.
Kind of astounding that they are recalling everything for validation. But I'd rather see that than have more bad CPUs going out. I ran into problems with a defective 3600x myself, not eager to repeat the experience.
As someone with a Zen 1 segfault issue I was not very happy that Amd shipped the first batch with them (they RMA'd mine eventually) but such is life.
 
Or you know, the obvious choice, AMD.

But I don't expect this to become a regular occurrence out of Intel. They will be wary and pay more attention to quality and especially voltage issues. They are moving desktop and mobile to chiplets, which could help them.

Has anyone mentioned a class action lawsuit yet?
It has always been a regular occurrence with Intel. Literally the thing that forced Apple to switch to in house CPUs was Intel giving them incredibly buggy CPUs
 
Elbrus ISA works just fine. It's just vaporware outside the CIS.
I've actually used Elbrus computers. They're fine. Most of the performance issues are simply caused by the fact that we were almost always running x86 binaries on them through an emulation layer. The processor depends very much on compiler optimisation (like Itanium), so naturally something compiled for x86 is going to run like utter garbage. When we had software actually compiled for it I'd say it ran about as well as it would have on an x86 chip of comparable vintage. Latest chip was half a decade ago, and even then not on a very recent node, because half the use case is army equipment and they want something verified to not flip bits just because the next town over got nuked or other petty details like that.
It has always been a regular occurrence with Intel. Literally the thing that forced Apple to switch to in house CPUs was Intel giving them incredibly buggy CPUs
More accurately Apple kept asking for low-power processors for laptops, and Intel kept failing to deliver. M1 proved them right on that, x86 is still nowhere near catching up in terms of battery life, and is only barely competitive for performance (on laptops, while plugged in) since all the chips put out way too much heat for any mobile heatsink to cope with. Even on desktop high-end Apple Silicon is managing to compete with much hotter x86. Like the 10-core M4 beats the 13700KF in Passmark, and that's a 253W chip! The M4 sips a leisurely 20W, that's well under what passive cooling can handle.
 
Last edited:
  • Thunk-Provoking
Reactions: Vecr
50% failure rate, lmao.

Screenshot_2024-07-25-20-03-21-323_com.brave.browser-edit.jpg

 
doesn't support DLSS4 frame generation. Frame generation is a big uplift in newer games.
not true. it does. they just disable it for some reason, probably to get people to buy the new shiny. theres a script you can install that basically fixes that but it's game dependent and also through modding
cyberpunk 2077 and alan wake 2 are examples where that's the case
 
  • Informative
Reactions: Brain Problems
Back