Warframe devs report 80% of game crashes happen on Intel's Core i9 chips | Hacker News Reader

Warframe devs report 80% of game crashes happen on Intel's Core i9 chips

tomshardware.com

50 pointsNarishma22 comments

Open articleView on HN

Warframe dev here; there seems to have been some confusion in the article that has misrepresented what we said here:

https://forums.warframe.com/topic/1405008-instability-on-rec...

We posted an follow-up to clarify this morning:

https://forums.warframe.com/topic/1405596-follow-up-regardin...

> As you can clearly see: Waframe crashes on everything!

made my day

Well I'm in the market for a new gaming pc, and because I dont want to do some meta analysis of reports of which intel chip will crash on me, I'm going with AMD.

A good choice even without this issue.

It feels like intel simply pushed the clock rates too close to the edge to keep up with competition. These cpus are very poor over clockers. No margins at all. Add a hot summer day (northern hemisphere), some SSE2 instructions that seem to really push cpus to the edge and a less than perfect heat sink mount and you’ve got a crash.

Laughs in 7945HX and RTX 4090 Mobile

I guess I feel better about how much heat this thing puts out and the occasional crash I get from it when it's hot...

I don't think it's the clock rates. If it was, Intel would've published information about how consumers can work around this by now, or at least given that info to their server customers. Instead you have server farms with 50% CPU failure rates across hundreds of CPUs.

I believe Intel did just that in May with a bunch of bios updates for various vendors.

Asus for example:

https://www.asus.com/us/support/faq/1053456/

Yeah, and that fixes some issues. But even with those conservative settings, Intel CPUs are failing in droves, especially in the data center (specifically game servers, where the high single core of the Core series makes more sense than Xeon). Wendell from Level1Tech has more details here: https://youtu.be/QzHcrbT5D_Y

Kind of seems like both Motherboard and Chip manufacturers are pushing the higher end processors a bit too close to the limit. As easy as it'd be to point at motherboards pushing all-core boosts and bumping power limits and/or boost durations up 50%+, these 13/14 gen chips also act up (though less often) on Workstation/Server boards.

My recent AMD build (5950x) also had a similar high-end part instability where it would lock up under Linux when downclocking to a very low idle. Replaced the processor but it still needed a small voltage bump to keep stable.

as a counterpoint, my 5950x has been rock solid. I ran VMs under proxmox for a while until i moved those offsite and now the 5950x is my primary computer. I did get a water cooler for the CPU and was very careful about application of thermal paste and such. as best i can remember this machine has never crashed, even under severe loads while in a drywalled shed with a window unit keeping the shed ostensibly at 75F. Think CPU cryptomining, or compiling the linux kernel with --jobs=33 24/7.

I’d be curious if these were predominantly mobile or desktop i9s. I have an i9 laptop that regularly runs at 100 degrees Celsius which is supposedly “safe” for these chips but it doesn’t really feel safe when it touches my skin. I’ve experienced firsthand many laptops with thermal issues and just wonder if Intel is spec’ing these too high to win benchmark wars at the cost of stability.

Although if it’s just Warframe crashing perhaps the software itself isn’t well optimized for Intel’s big/little core setup on the newer 13th and 14th generation chips (performance and efficiency cores - supposed to be handled by the OS but with gaming all bets are off)

it says i7s too in the damn title. It’s 13th and 14th gen

Yes, I had to shorten the title so it would fit the HN limit but the article says 13th and 14th gen i7s represent a significant portion of crashes as well.

Not regular i7's, "Core i7 K-series CPUs also have high crash rates".

Are you sure it’s not becuase their data set likely doesn’t contain very many i7 non-ks

Wendell from level1techs made a detailed video about this recently.

Tldr: it's the high-end chips from 13th and 14th Gen specifically. Also the bit at the end about support costs increasing 10x for those platforms really shows how bad this seems to be.

https://www.youtube.com/watch?v=QzHcrbT5D_Y

This looks like general memory corruption generated by the CPU.

Probably hyper threading related or energy throttling related.

Interesting. 12900k has no issues with either.

Maybe ask Francois of he remembers what he optimized a little bit too much?

my reply to that, were asked of me, "no such thing"

Reply on news.ycombinator.com