Etern4l

My long and painful quest for the ultimate non-LM TIM concluded with PTM7958 SP. I think it has been running without a hitch for unprecedented several months now.

Thanks. I guess another difference between AIOs and many custom loops is that the former tend to use opaque black tubing, presumably to optimise durability. Clearly this would limit bio growth, but apparently the other benefit is that clear tubes can leach some plasticizer compounds into the loop. The recommendation here is to use black EPDM tubing.

13900K can probably be pushed into that territory, given that it "achieves" 330W+ stock, and again: at 1/3 the die size of a Threadripper. It's nuts. Yes, but my question is: do you absolutely need to annually service a custom loop if there is no apparent performance issue? Best practice to prolong pump and rad life I guess.

What would you say if asked to write the promo piece? 🙂 If you tried to share the truth and enumerated the potential concerns you would be out of the door. Additionally, this is not the only model of this kind, and some will likely end up being open sourced soon (if they haven’t been already), meaning they can be run without any scrutiny, adapted etc.

TRX4 has almost 3x the surface area of LGA1700 for a CPU slower (and less power hungry?) than 13900K. Just like Sapphire Rapids, you could probably just run it on air. Why would the annual maintenance be any more of a requirement for a custom loop than it is for an AIO? AIOs come with long guarantees although I don’t think those guarantee no performance degradation. If anything, it would be reasonable/intuitive to assume that the smaller AIOs need to be serviced more often to maintain top performance. Good point about the warranties, just included for completeness’ sake. Didn’t really think @tps3443 would be worried about that lol

Definitely. I would only go with quick disconnects if GPUs were involved.

Cherry picked output examples for sure, but in general a step in this direction has been inevitable. Given how difficult it is to get image generators to produce exactly what you want, let’s wait and see how useful this will prove in practice. The other question would be regarding rendering time and cost. I know a guy high up at a major VFX company, will be able to get an authoritative answer in a few months. What’s worrying is the potential for fake news generation and other illicit uses, where production value criteria are often relaxed.

Are you using a 13900K+ in any scenario that would push it to 90%+ utilisation for extended periods? I can see the following reasons against using a custom loop: 1. Lack of know-how / insufficient understanding of the benefits 2. Fear of leaks 3. Extra setup cost 4. Air+AIO cooling already results in a pleasantly quiet and efficient system 5. Voided GPU warranty 6. Extra setup hassle So in the case of @tps3443 running a 14900K system only 5 and 6 might apply, as he already has a custom loop setup in place. I did not include extra space requirements as a consideration, since an external chiller/rad is optional. The reason my PC is not running a custom loop is a combination of 1-3, 5, 6, and initially 4 (with 12th gen). I wish I did not spend all that time trying to satisfy 4. with 13900K. Would have been much better off switching to a custom CPU loop at least.

Well that looks quite conclusive. Is it not possible to set max performance mode in the AMD GPU control panel?

Yeah, happened again: “Your connection to this site is not fully secure. Certificate valid.” Wild guess is that some content urls are not https, sometimes, for some reason.

Is there conclusive evidence the stuttering has something to do with the GPU, as opposed another system hardware or software component? (did they try another GPU?)

This was showing up in the Brave address bar. Good since yesterday p.m. I think.

The percentages tell their own story, as do the absolute numbers. While the difference between say 1% and 5% of peak user base still being active doesn't look like much - the latter game would have retained 5x larger chunk of peak player base, in relative terms. On the absolute side: 7k players just 6 months on or so looks pretty low for a game Todd Howard bragged sold over 3M copies, and where other games targeting the same demographic have 2-4x larger active user base years on. Let us also keep the details around the type of the game in mind: virtually endless open world to explore, supposedly huge replayability etc. If this was a short campaign single-player only FPS for $35 then fair enough. Angry Joe is rarely wrong I guess 🙂

I bet you would 🙂 Apparently, I’m so excited about this unmissable opportunity that I have stated suffering from Freudian slips and typing “notebooktalk.bet” into the browser 😉

True enough (RDR2 seems to be one exception), however, the devil is in the details. Not sure if it makes sense to compare to the overall population of single-player games, however, we could compare to a few reasonably similar open-world AAA titles and look at the absolute number of active players as well (taking time since release as well): StarFlop: https://steamdb.info/app/1716740/charts/#all Peak 330k, 7k ca. 1/2 year after release CP2077: https://steamdb.info/app/1091500/charts/#all Peak: 1053k, 28k over 3 years after release RDR2: https://steambase.io/games/red-dead-redemption-2 Peak: 77k, 28k over 4 years after release Fallout 4: https://steamcharts.com/app/377160#All Peak: 470k, 15k now - 8 years after release The data provides very clear evidence of SF strongly underperforming its peers in terms of sustained player interest.

Definitely need to get some betting going around that resolution lol AIOs suck for 13900K+ at "stock speeds", you would be back on a custom loop inside a month.

Started seeing this several days ago. SSL cert looks fine. Thoughts?

How much do people expect to pay for a GPU water block? I was checking out the options for my 3090 Ti FE and all I could find so far was a £300+ unit from Extra Kostly Water Blocks (out of stock anyway).

OpenAI CEO Warns That 'Societal Misalignments' Could Make AI Dangerous Didn’t read the whole thing but rings hollow coming from ClosedAI.

In this instance there is a reason. In order to run on small GPUs, the models have to be dumbed down and only Ampere onwards support the required low res modes. Still a good marketing gimmick I guess, given the limitations.

A very niche offering. I wonder how many of those would be expected to sell globally. To be fair they did bump the caches significantly in 14900K, so it’s a bit more than a 13900K “Super.” marketing gimmick.

Ha! Checking AIB's websites for stock is indeed a good idea... even more stock that way. BTW ASUS have no shame whatsoever https://shop.asus.com/us/90yv0id7-m0nm00-rog-matrix-rtx4090-p24g-gaming.html

Looks like JO’s fav topic: Interestingly, and sadly, none of the antitrust bills have actually passed.

Well, it kind of makes sense, right? While 4080S is not a 4090 killer outright, the performance gap between them does not really justify a $800 premium. You would expect the prices to get closer, and that’s what’s happening. A lot now rides on how much 4080S stock NVidia needs to shift, and how many “inventory management” tricks they are willing to pull in order to support the prices. Meanwhile Huang is busy selling GPUs to ’every country’: Every Country Needs Sovereign AI, Nvidia CEO Says

Having issued a fair warning about the existential risks of AI, Sam F. Altman is now going nuts in order to deliver “the dream”: Report: Sam Altman seeking trillions for AI chip fabrication from UAE, others Meanwhile an Alphabet-owned self-driving car was set ablaze by a crowd in none other than San Francisco. The people are waking up…