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 wish my CPU would heat my room during the winter.
My home office is a spare bedroom, and the computers live in the attached walk-in closet together with some 3D printers and other office junk. In the summer I can just crack the window in the closet and dump the heat outside, and in winter I can crack the closet door to vent the heat into the office. It actually works really well, cuz in winter I prefer warmer temperatures than my husband does, so while the rest of the house is chilly, the office is kept nice and comfy by about a kilowatt of computers and servers.
See if you can do something similar? Just placing the computer on a windowsill with blackout curtains in front for example should trap the heat away from the rest of the room, so you can crack the window in summer and open the curtains in winter.
 
I don't think it's a new thing. @The Mass Shooter Ron Soye said multiplayer games have been doing that for a while, but so do MMOs.

My tactic for building a PC from 2008 onwards was to target Xbox 360 specs since 99% of games were made with consoles in mind, so as long as you could equal (preferably beat) console specs, you could play games on medium settings comfortably for a good price.


On an unrelated note, the hot summer weather is starting in the UK. I'm tempted to underclock the CPU and maybe even the GPU so it stays cool and doesn't act as a space heater. I've not come close to maxing out the CPU or GPU outside of a few specific situations like messing with Half-Life Raytracing mod, but that also means I don't know if underclocking will have an effect at all.

Power increases approximately like the square of clock speed, so if you cut the power draw by half, expect the performance to drop by about 1/4 (sometimes more like 1/5). For example, the i9-12900 and i9-12900KS are the same chip. The only difference is the former draws 65W, and the latter draws 125W. Across a range of benchmarks, the KS typically gets about 20%-25% better performance, right in line with its ~2x power draw.

You can find benchmarks online for some of the more popular chips undervolted to low power draw, and that's about what you see. If you can undervolt a GPU, you should see the same thing, at worst a 25% performance drop per 50% power drop, but I imagine you already do that.

There are other things you can do like ensure vsync is always on so you're not gulping down power to draw frames you'll never see.
 
A youtubers I watch did a video testing one of the new RDNA3 iGPUs. Performance is pretty decent. People have mentioned in this thread before but it does look like these iGPUs could take over the budget GPU market, especially if the price is compelling.
I'm happy with my current desktop but if they release some cheapo laptops with a these chips I might pick one up to replace my old thinkpad just so I can play newer vidya while traveling. Gaymen laptops are too expensive and almost universally look ugly as shit.
 
A youtubers I watch did a video testing one of the new RDNA3 iGPUs. Performance is pretty decent. People have mentioned in this thread before but it does look like these iGPUs could take over the budget GPU market, especially if the price is compelling.
I wish those RDNA3 APUs would come to desktop, and I don't mean NUCs and mini-PCs.
It will be interesting if Rembrandt and Phoenix drop onto desktop at the same time. Everything's ready, A620 motherboards exist and DDR5 is getting cheaper.

Why would you want a chip like that in a big, fat tower case? Serious question.
How about a Mini-ITX case with no room for a GPU? I know it exists somewhere.
 
It will be interesting if Rembrandt and Phoenix drop onto desktop at the same time. Everything's ready, A620 motherboards exist and DDR5 is getting cheaper.


How about a Mini-ITX case with no room for a GPU? I know it exists somewhere.
I guess the question is why even go that big? If you don't need a GPU card, why not go for something like this?
 
  • Like
Reactions: Judge Dredd
Why would you want a chip like that in a big, fat tower case? Serious question.
I put a 5600G in my home server/router. Great way to get stupid consumer hardware to boot without wasting a valuable PCIe slot on some silly graphics card. Leaves me room to put in more NICs and storage. Server's now 1x10GbE, 5x2.5GbE, 2x500GB nvme cache, 2x500GB Optane for metadata and small files special storage, 4x8TB HDDs for bulk storage, and twelve more SATA ports waiting to be filled when my four new HDDs arrive, and in the future when those also fill up. If I'd had to waste any of my ports on a GPU, I would have lost either the 10GbE, two of the 2.5GbE, the nvme cache, or the twelve extra SATA ports. iGPUs make a lot of sense here.
 
I guess the question is why even go that big? If you don't need a GPU card, why not go for something like this?
I don't see your point in trying to minimize the socketed desktop APU. They are useful, and a system is potentially cheaper to build than buying the equivalent mini PC.

I put a 5600G in my home server/router. Great way to get stupid consumer hardware to boot without wasting a valuable PCIe slot on some silly graphics card. Leaves me room to put in more NICs and storage. Server's now 1x10GbE, 5x2.5GbE, 2x500GB nvme cache, 2x500GB Optane for metadata and small files special storage, 4x8TB HDDs for bulk storage, and twelve more SATA ports waiting to be filled when my four new HDDs arrive, and in the future when those also fill up. If I'd had to waste any of my ports on a GPU, I would have lost either the 10GbE, two of the 2.5GbE, the nvme cache, or the twelve extra SATA ports. iGPUs make a lot of sense here.
Now that the iGPU is standard with Ryzen 7000, people have to decide if they need more graphics performance.

Although a Phoenix APU could bring another feature, the AI accelerator, to desktop before Ryzen 8000 arrives.
 
I don't see your point in trying to minimize the socketed desktop APU. They are useful, and a system is potentially cheaper to build than buying the equivalent mini PC.

Trying to understand why you want something isn't minimizing those wants. "It's cheaper" is as good a reason as any.
 
im going to build an arcade-media machine to hook up on the tv.
i have already got everything for it except a hdd. im going to use the xeon 1246 v3 and the 590 from my old build but im looking for a micro atx board. anyone has recomendations ?
 
  • Like
Reactions: Judge Dredd
Trying to understand why you want something isn't minimizing those wants. "It's cheaper" is as good a reason as any.
idk, that one is ghastly and I would rather someone walk into my living room seeing me sucking cock than seeing that thing, it would be less embarrassing. Plus, people like building their own things, making it look the way they want with the added benefit of choosing their own brands/components.
 
  • Islamic Content
Reactions: Brain Problems
idk, that one is ghastly and I would rather someone walk into my living room seeing me sucking cock than seeing that thing, it would be less embarrassing. Plus, people like building their own things, making it look the way they want with the added benefit of choosing their own brands/components.
To be fair, less tasteless options exist. These also are small enough that you can just stick them behind your TV with double sided tape even if the TV is on a wall mount.
 
  • Informative
Reactions: Brain Problems
My Ryzen 9 5900X completely shit the bed and started doing that WHEA-Logger Error 18 stuff (cache hierarchy error/bus interconnect error), spontaneously rebooting now and then. Apparently, the defect rate on the early Vermeer chips was very high. I swapped the CPU out with one of the newer revision ones (the B2 Stepping of the 5900X) and swapped the Noctua single-tower cooler for an NZXT Kraken 240mm AIO. It's running a lot cooler than the last one. I mean, 10 to 20 C less. Way lower temps, substantially less noise. And, most importantly of all, no more spontaneous reboots.

I also just got done putting together my new weapon of choice.
  • Case: Fractal Design Meshify 2
  • PSU: EVGA SuperNOVA 1300 GT
  • Mobo: MSI MPG Z790 Carbon
  • CPU: Intel Core i9-13900K
  • Cooler: Corsair iCUE H150i ELITE CAPELLIX 360mm AIO
  • RAM: Corsair Vengeance 192GB (4x48GB) DDR5-5200
  • GPU: MSI GeForce RTX 4090 SUPRIM X 24G
  • Storage: 2x 2TB Samsung 980 Pro M.2 SSDs, 2x 4TB Samsung 870 EVO SATA SSDs (soon to be 4x SATA SSDs, when the rest show up).
The fans are a mixture of Corsair AF120 Elites (on the AIO) and AF140 Elites, complete with eye-searing animated iCUE LEDs. The pump came with a 6-fan controller, but it wasn't enough for the 8 fans I wanted to install, so I split them across the pump's mini-Commander thing and a Commander Pro, with four fans on each one. They're nice and quiet, and it seems to be running pretty cool. The CPU is idling at 35C and doesn't seem to go over 70 under load. I'd heard the 12VHPWR horror stories about the 4000-series cards, and the NVIDIA 4-into-1 adapter was a big fugly tumor of a thing and absolutely unbearable to look at, so I got a Cablemod cable that puts the 4x 8-pins close to the power supply and shaves some bulk off the visible parts behind the window. Honestly, 12VHPWR is kind of pointless. There was no reason why they couldn't have simply engineered these cards with four ordinary 8-pin PCI-e connections along the side. They certainly have the space for it.

Overall, it turned out nice:

NewBuild.jpg
 
My pc is getting just a bit long in the tooth, I bought it back in ‘19 just before the world shutdown. Now I’m thinking of getting a high end rig, hopefully before the depression fully hits.

Anyone with experience with AI art generation who can give me some suggestions for specs to shoot for? I’m just getting into it, and the lack of an Nvidia card is proving a pain.

I’ve got a pretty decent budget, $1500-$3500. Mostly going to depend on the graphics card obviously. Is the RTX 4090 worth the price tag, or is a cheaper 4070 better bang for my buck?
 
Honestly, 12VHPWR is kind of pointless. There was no reason why they couldn't have simply engineered these cards with four ordinary 8-pin PCI-e connections along the side. They certainly have the space for it.
No, we already had a connector that would have worked perfectly for this. The EPS12V has been used on datacentre cards for decades and can supply 336W per connector. The separate PCIe power cable never should have been a thing in the first place, EPS12V already existed, was smaller form factor, and provided more power.
Anyone with experience with AI art generation who can give me some suggestions for specs to shoot for? I’m just getting into it, and the lack of an Nvidia card is proving a pain.
I’m running Stable Diffusion on a 6900XT just fine, Nvidia isn’t mandatory if you don’t want to switch. It’s just a bit of a hassle to set up on AMD, you need to run the rocm-PyTorch docker container and make your own image with Stable Diffusion on it.
 
I’ve got a pretty decent budget, $1500-$3500
Idk, that's almost useless to state. It's like saying "I'm looking for a new car and can spend between 30k-100k".

You state how much you really want to spend and then go buy the best parts for that money. If you really feel like dropping $3500, then sure, go get a 4090. Nothing else comes close. It'll hold up for longer, too.
 
I’m running Stable Diffusion on a 6900XT just fine, Nvidia isn’t mandatory if you don’t want to switch. It’s just a bit of a hassle to set up on AMD, you need to run the rocm-PyTorch docker container and make your own image with Stable Diffusion on it.
Would you be willing to give me a few pointers to get me started on SD with an AMD card? I thought I'd made a good choice with a 20GB VRAM Radeon card and then I realised how difficult it was to do anything AI with it. I've recently been contemplating selling it (at a loss!) to buy an Nvidia card purely so I can do SD and related. 4080 is much less RAM and a rip-off. It would be frustrating to have to swap.

I have tried. I found a tutorial which was supposed to "convert" models somehow. I don't recall it involving Docker at all. And whilst I sort of got it working the results were greatly disappointing and I then broke it with some sort of version issue.

I'm not a technical novice, but I don't really know how to get this working and just a few good links or the concepts would help. The tutorial I read didn't really explain everything and assumed you understood how SD and models all work. So when something went wrong, I couldn't figure it out.
 
Back