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

[Car Won't Crank]

Im not saying noctua doesn't make a good fan...but really. Come on. Some of the prices are retarded.

Also pro tip, as for longevity, you can keep cheap sleeve bearing fans going until the motor burns out (un likely) by taking the little retention clip off on the back of the fan shaft and putting a tiny bit of lube in there. Also can make them quieter.

Noctua is kinda overrated. I've recently bought some Hyte branded fans 3 pack for under 30 usd and they move tons of air. Slapped them in a server case and they're running fine. Time will tell if they're as durable as the cougar fans I got years ago that still run.

 

[Just Some Other Guy]

Noctua is kinda overrated. I've recently bought some Hyte branded fans 3 pack for under 30 usd and they move tons of air. Slapped them in a server case and they're running fine. Time will tell if they're as durable as the cougar fans I got years ago that still run.

I will admit that I did cave and buy one of their 5v 40mm fans for my Pi4 in a 3d printer.

It felt dirty paying $15 for it, but I didn't feel like buck breaking a 12v/24v fan.

 

[Smaug's Smokey Hole 2]

Is replacing the thermal paste on a 980 a bitch to do? Mine is starting to run hot and loud but I have way too many other things I'd rather spend money on than a 4070

That would depend on the model, look for a disassembly your make/model on youtube so you don't yank any cables or break any clips. Tweezers or a thin flat head screwdriver is recommended for removing/putting back in any annoying thin ZIF ribbon cables.

It's also a good idea to clean out the rest of the cooling assembly, the heatsink can be washed under the tap.

 

[blablabla]

heatsink for the second nvme, yes or not

 

[Susanna]

heatsink for the second nvme, yes or not

PCIe3? No need. PCIe4? It won't do anything for intermittent loads, like surfing the internet, but it will help with loading big files (for example games).

 

[blablabla]

PCIe3? No need. PCIe4? It won't do anything for intermittent loads, like surfing the internet, but it will help with loading big files (for example games).

i want to use the second nvme as the games one. the first one use the motherboard heatsink

 

[DavidS877]

i want to use the second nvme as the games one. the first one use the motherboard heatsink

They're cheap. Just make sure you have the clearance.

 

[Just Some Other Guy]

Huh, apparently newegg had a $500 combo with a ryzen 7700x, 32gb of ram, and an asrock x670E mobo. Pretty damn nice.

 

[F-35 Lightning II]

So anyone know if the scheduler is gonna be fixed for the 7950x3d, just picked up a mobo+ram and was thinking about picking one up

 

[Second Sun]

So anyone know if the scheduler is gonna be fixed for the 7950x3d, just picked up a mobo+ram and was thinking about picking one up

It's pretty much all dependent on the specific piece of software. There's not much AMD or Microsoft can do if something has threads spread across the 2 chiplets that hasn't already been done for multi-CPU/NUMA configs.

 

[Hikaru86]

What is the best option for an "ultra budget" mini itx build? I don't need anything fancy, just to play emulators and light 3d modeling.

Why i bought a fucking d19 pc case instead of the d21, who is a little bigger and compatible with my old gtx 970? I fucked up my budget and trying to fix it now.

 

[Susanna]

What is the best option for an "ultra budget" mini itx build? I don't need anything fancy, just to play emulators and light 3d modeling.

Why i bought a fucking d19 pc case instead of the d21, who is a little bigger and compatible with my old gtx 970? I fucked up my budget and trying to fix it now.

Intel have a slight lead for emulators, since most of those are single-threaded. Light 3D modelling as in Blender or Solidworks? You'll need anything between a good iGPU and a 1660 Super. If you do want to save money and go iGPU, I'd aim for a Ryzen 5600G (or wait until the zen4 APUs launch in a few months, those are probably going to be amazing), because Intel's iGPUs are very weak, but if you're thinking dedicated graphics card, this is what I'd do. The GPU is a little more than you'd strictly need, but it's not actually particularly more expensive. Much better value unless you find a cheap secondhand GPU. If you do want to look for a secondhand GPU, I'd aim for a GTX1080 or up, but with a length limit of just 188mm you're going to have a hard time finding anything worth buying. Bear in mind that the performance of the next generation is roughly comparable to the past generation one tier up, so GTX1080=RTX2070=RTX3060. It's not a perfect comparison, but it gives some idea what you should be looking for.

 

[Hikaru86]

Intel have a slight lead for emulators, since most of those are single-threaded. Light 3D modelling as in Blender or Solidworks? You'll need anything between a good iGPU and a 1660 Super. If you do want to save money and go iGPU, I'd aim for a Ryzen 5600G (or wait until the zen4 APUs launch in a few months, those are probably going to be amazing), because Intel's iGPUs are very weak, but if you're thinking dedicated graphics card, this is what I'd do. The GPU is a little more than you'd strictly need, but it's not actually particularly more expensive. Much better value unless you find a cheap secondhand GPU. If you do want to look for a secondhand GPU, I'd aim for a GTX1080 or up, but with a length limit of just 188mm you're going to have a hard time finding anything worth buying. Bear in mind that the performance of the next generation is roughly comparable to the past generation one tier up, so GTX1080=RTX2070=RTX3060. It's not a perfect comparison, but it gives some idea what you should be looking for.

The best i found in country are second hand Gigabyte GTX 1070 mini for "maybe buy a RX 6600 or a 1660 super stormx instead?".

But I found some OEM GTX 1660ti in Aliexpress for an okay-ish value. And budget AM4 and LGA 1700 mini itx mobos for "buy it in China or fuck off". Hell, i found a LGA 1155 mini itx kit for pennies and almost buying it and the 1660ti. And wait for AMD 5 kits price to get better.

 

[The Ugly One]

I've been working on a dual core microcontroller. The retards are all "Use the mutex and the locks and the xyz." Or, you know I could just use the fact that a byte is still a byte and is still atomic as long as I force the compiler and the CPU to actually write to and read from the memory location. Then again I come from a time where efficiency mattered and you wrote microcontroller code in assembly since the compiler would make it too big for the flash.

Now, if you'll excuse me I need to adjust the onion on my belt.

I knew a guy who threw out the runtime lib's lock and wrote his own. It just used an atomic integer r/w and had the thread spin in a while() loop until the integer changed value.

The reason is subtle - an actual thread lock suspends the thread until it's unlocked, which incurs OS overhead. This method did not actually suspend the thread, but kept it live while not otherwise doing anything. He got a double digit performance gain in that section of the application IIRC. This was a long time ago, though.

 

[Drael]

So anyone know if the scheduler is gonna be fixed for the 7950x3d, just picked up a mobo+ram and was thinking about picking one up

Wouldn't count on it personally. AMD seems to be dug in on "just use xbox game bar" being the solution and outside of some future BIOS updates maybe helping it'll likely remain as it is. You can pick up a normal 7950x for right at or just under $500 if buying "used" doesn't spook you and it'll have most of the game performance without any of the thread scheduling nonsense. That said, I'm sure the 7950x3d is great for those who want to autistically monitor what cores are being used and when but I'd rather have something that JUST WERKS™ like I'm hoping the 7800x3d will.

 

[Agent Abe Caprine]

I've been buying noctua fans for a long time. A good qualitative fan can stay with you long. I've experimented with many different brands of quiet PC fans and the problem always eventually came down to reliability and all kinds of strange bearing and resonance problems, either out of the box or with age. For all the years I bought noctua, I never had to retire one. I think the only fan manufacturer that's in tie with that particular performance metric in similar circumstances in my personal and completely anecdotal opinion is ebm papst, and I'm not sure the thread is familiar with them. It's also not about the noise in decibels alone, but also at what frequency. I bought one closed-loop AIO system out of curiosity a while ago (I used to have an incredibly complicated watercooling thing in the 00s) and found the high pitched whine it made more annoying than any fan. They're probably not all like that, but still. Also consider that the CPU fan in your computer case also has the job to get some moving air over other mainboard components in that area. If you do full watercooling, that might be missing depending on your cooling layout.

I had to retire one Noctua. It was a small fan I had on a CPU cooler. Still have it. It probably just needs more lube.

 

[Chief Nigwelfacheqee]

Fan sperging; people don't typically buy noctua because of performance, but because they mix top end performance with excellent acoustics. If you don't care about that (or care more about looks) then of course spending the same money on 1 fan as you can on a pack of 3 chinesium ones doesn't make any sense.

Then there's autards like me who spend even more money on fans that perform less in both airflow and acoustics but look sick as fuck (lian li)

 

[The Mass Shooter Ron Soye]

https://www.anandtech.com/show/18781/nvidia-unveils-rtx-ada-lovelace-gpus-for-laptops-desktop-rtx-4000-sff

https://archive.ph/kdACw

RTX 4000 SFF, a new small form factor professional Lovelace GPU:

NVIDIA Professional Visualization Video Card Specification Comparison​
​	RTX 4000 SFF Ada Generation​	RTX A2000​	RTX 6000 Ada Generation​
CUDA Cores​	6144​	3328​	18176​
Tensor Cores​	192​	104​	568​
Boost Clock​	~1.56GHz​	1.2GHz​	1200MHz​
Memory Clock​	16Gbps GDDR6​	12Gbps GDDR6​	20Gbps GDDR6​
Memory Bus Width​	160-bit​	192-bit​	384-bit​
VRAM​	20GB​	6GB / 12GB​	48GB​
ECC​	Partial (DRAM)​	Partial (DRAM)​	Partial (DRAM)​
Single Precision​	19.2 TFLOPS​	8 TFLOPS​	91.1 TFLOPS​
Tensor Performance (FP16, Sparse)​	153.4 TFLOPS​	63.9 TFLOPS​	728.5 TFLOPS​
NVENC/NVDEC​	2 / 2​	1 / 1​	3 / 3​
TDP​	70W​	70W​	300W​
Form Factor​	Half Height Dual Slot (2.7in x 6.6in)​	Half Height Dual Slot (2.7in x 6.6in)​	Full Height Dual Slot (4.4in x 10.5in)​
Cooling​	Active​	Active​	Active​
GPU​	AD104​	GA106​	AD102​
Architecture​	Ada Lovelace​	Ampere​	Ada Lovelace​
Manufacturing Process​	TSMC 4N​	Samsung 8nm​	TSMC 4N​
Launch Price​	$1250​	$450​	$6800​
Launch Date​	03/2023​	10/2021​	01/2023​

 

[Susanna]

I have to question going half height just for SFF. Most SFF enthusiasts, myself included, care more about length and thickness than height. In most cases we’re putting the GPUs behind an ITX motherboard, so as long as it’s not taller than an mITX motherboard is, height doesn’t really matter that much. Kudos for the blower fan though, that’ll make the air intake much less sensitive to the cramped conditions.

Half height cards are very much a thing in the data centre, server, and workstation market, so I guess they’re just making a consumer version of the A2000? What’s the market for a non-quadro workstation card though? Seems like a silly product. They should have made it full height, that way they could have fitted a larger, less noisy blower fan. That’d be much more popular, unfortunately a lot of SFF people put their computers on top of their desks so the noise level of a small blower card can be pretty significant.

 

[Generic Retard]

What’s the market for a non-quadro workstation card though? Seems like a silly product.

Probably putting a bunch of them into 2 HE servers for computation rather than display stuff?