Search results

Exactly what I thought. Thanks.

No, in truth I'd like to use an AVX stress test, since my workloads with data augmentation make use of AVX (albeit not 512).

I don't overclock. Besides, that kind of hardware cannot be overclocked (I think), but I appreciate your recommendation. I just want to make sure I have a stable machine to work with, since it's new. Now I did the blend test full steam with AVX-512 activated, then another half hour without...

Hi folks. Got a Xeon 8260M upon a X11SPA-TF here, just finished to build it. It works, at least for now. Now, I always used to test my processors with Prime95, latest version, half an hour of blend + FFT small. But recently, a colleague of mine told me that Prime95 can kill stuff (either the...

Please, elaborate.

It won't be a flop, not at all. The Titan RTX was essentially the 2080ti at 3X the price, an still they sold it very well. You judge the GPU market from the point of view of a gamer, while GPUs are quite indispensable computing devices. Compared to the Titan RTX, The 3090 has the same amount of...

I will, but I do deep learning. NVlink will allow for memory pooling, which is a very good thing in my field. Of course, that assuming I can get my hands over them, which seems to be quite impossible at the moment. Besides, I don't want to buy custom designs. And no, fitting two FE won't be...

Advice needed here. I got *four* 2060 Supers (deep learning research), purchased 6 months ago (370 bucks each), blower design, perfect conditions. How much can I ask for each of them?

I think that the Quadro version of the 3090 will be blower-cooled. Maybe some kind of 3-slot blower shell. Of course they will make you pay 3X the price of a plain 3090, as usual.

I just discovered that b450 allows for 16x/4x. For 8x/8x I need x470.

More to the point about the chipset fan: your personal experience constitutes an useful feedback. Consider, though, the my usage scenario is a bit different.. I have to run the system with 2 gpus (8x/8x, for deep learning), plus another crappy one in the 2.0 4x slot (video output). There will be...

I know. But even outside the base clock envelope, you get pretty good power draws. The 3700X manages to stay within 90W even during stress tests. See for example: https://www.tomshardware.com/reviews/ryzen-9-3900x-7-3700x-review,6214-3.html Zen2 seems to be the real deal these days. I'm just a...

Thanks, that's valuable feedback. I honestly hoped not to get below 8 cores, maybe I'll settle for the 3700x.

Interesting. Would you please post the same measurements for your 3900X? Thanks! Apart from this, I'd really like to avoid x570 and its pesky fan.. I wonder if the same tdp-limiting feature is available for B450/X470...

So, it appears it will be available just for OEMs...? I wonder if the same stands for the 3700 non-x..

Hi. I'm in search of the best tdp/performance ratio, since the computer I'm building will likely function 24/7. I'm still unable to find reliable informations about the 3900 non-X, which should provide very good performance in a tdp of ~65W. It has been speculated that such processor (along...

Socketed GPUs do exist: take for example the Nvidia SXM2 (on ebay, the P100/V100 sxm2 are cheap, too. But good luck for finding cheap sxm2 mobos/servers).

Hi guys. I need to provide sufficient cooling to two 2080ti stacked tightly, in a 4-slot space. This would of course exclude any open-air design for the video cards (which I have yet to buy), whereas a blower design will almost surely thermal-throttle. The most obvious cheap solution would be...

Take any Asrock x99, or any C612 board whatsoever, and read the memory QVL. Search for modules that have the same configuration (e.g. 2Rx4). If they match, then teh board will support your modules, no matter the brand. To be even more safe, choose modules with the same density. The same stands...

Hi guys. I'm in the process of choosing components to put together a pc for 24/7 operations. I need inexpensive components, small form factor, low power draw, and compatibility with Intel MKL numerical library, so my choice would be b360/b365 ITX and the i3-9100F. Now, I'm unable to find any...

Ok, it was worth a try, but it didn't work. I gained some 4-5C, but they still go over the throttling threshold. I will go liquid. h55+g12. I just want to ask you a couple of questions: 1. Do I have to unglue some thermal pad from the VRM/vram chips prior to mounting the g12? Do I have to...

Ok, thanks guys! I'll go with the fan bracket first. Should I still go over 84C, I'll switch to liquid.. I'll let you know!

Roger that, thanks!

Good point. I have 2 intakes and 3 exhaust. But the intakes run at 12V whereas the exhaust run at 5V. This should ensure positive pressure, but as you know we got the two 1080ti, whose FE design do exhaust the air outside the case. And when they are close to 100% rpms, that's a lot of air. I...

Nope! Under load, they run at less than 1750 MHz with the standard power limit (250W). Only if I raise the power limit to 300W each, they run at >1850 MHz. This till they hit the 84C limit. Then they run @14xx Mhz. Note that even so, one card goes straight to 90-91C. Then the fan hits some 88%...

This is truly valuable info. Plus, like you said, using two of them upon tightly spaced cards could even be unfeasible. Thanks! First thing I did. No result. Used arctic MX-4. I even polished the thermal surfaces with that arctic polisher. One thing I forgot to mention: none of the cards is...

Look: The two front 120mm fans blow less than 10cm from the air intake of the cards. They blow at 100% rpms. https://www.dropbox.com/s/0kgis3p6zg6vd7t/111.jpg?dl=0 The cards are tightly spaced, but a FE design should not suffer from that...

Full cover WBs are too expensive (cosidering all the other parts too). Yes I could sell them on ebay and get one 2080ti for the same price, but a single 2080ti won't replace two 1080ti for deep learning. I'll have half the memory and about 20-40% less speed (depending on the network type)...

Yes I thought about it. H80i v2 is compatible. But it costs almost twice than the h55, and all I want is to stay below 84C. I'm not in need to pursue some 5, or even 10 Celsius.

Hi mates. I got two 1080TIs Founder's, one of which hits regularly 90C (sometimes 91) with a room temperature of 20/21C, and its fan at 88-90%. The other runs marginally cooler, but always above 85C. Needless to say, they are loud as hell, and I get no boost (sometimes they even go below 1400...

Mh, just google around, and the h75 2018 doesn't match with the g12. Probably I'll go with a h80i v2 or an h55.

Hi fellas. I got two 1080ti, both FE, one from nvidia and one from evga. I use them for training neural networks. Maybe not all the 1080ti are created equal, since the evga one absorbs more power (even at idle), and hits much higher temps. With a 21C room temp, and a good airflow through the...

I found another good alternative: the nanoxia deep silence 5 rev.B It has 10 slots! WP100 and cougar panzer max are both viable alternatives, BUT the former seems to be a bit overkill (and weighs a ton!), while the second is a tad game-ish.. I'm building a deep learning workstation...

This was exactly what I was in need of! Thanks, and thanks to everyone for the answers, all of them were quite informative ;)

Hi mates. I got an Asus x99-WS/IPMI mainboard. Now I also got to install a dual-slot video card in the *last* slot of the board, the one closest to the lower edge of the mobo. I need a case that allows some space between the edge of the board and the PSU, otherwise the video card wouldn't fit...

I need one card to go at full 16X gen3. On the other hand, I do NOT need any SLI/XF stuff. Two different cards doing different stuff.

I think I need full 16X for doing deep learning, it's very bandwidth intensive. There will be a lot of communication between RAM and VRAM, and that should occur at the best speed. Benchmarks show an average loss of 10/12% in terms of performance on an 8X gen3 slot, and over 50% on a 4X slot...

Hi mates, Please help me with two questions I wasn't able to answer after *a lot* of googling. 1. The vast majority of h370 and b360 boards with two mechanical 16X slots do have the following layout: the upper slot has 16 gen3 lanes by the CPU. The lower one has 4 gen4 lanes by the PCH. Now, I...

Yours was the most money-saving advice I got here! Besides, I didn't think about that. Thanks! :)

Uhmm.. I was thinking: what about using this unit in conjunction with the Fujitsu one on the other thread (with the add2psu). Without the burden of powering up the system, will it be able to feed those two video cards comfortably? Consider it is monorail (you don't have to bother about...