- Apr 12, 2018
- 293
- 5
- 81
Sorry for some of the confusions in previous posts. To summarize: I am not a gamer nor a miner. I want to build a high-end research workstation for doing CUDA-GPU computations. I know that for gaming, PCIe 3.0 at x8 is sufficient but for my work, x16 makes a big difference. At first I chose the 7900X since it allows many lanes and supports AVX-512 which could be useful for some scientific computations. However, after hearing from a colleague that there could be a performance hit due to the Meltdown and Spectre patches, I decided not to invest big money on a buggy system. Instead, I planned to build and use a temporary system until Intel releases a CPU (Cascade Lake?) with these bugs removed at hardware level. I heard that it should be out in the 2nd half of this year so I chose to build a less expensive i7-8700K system with one x16 GPU slot. Components prices change almost daily and sometimes they jump up and down by thundered of dollars. One day, I found that at about $350 more, I could build a Threadripper system with more lanes and have two GPUs running at x16x16 concurrently. Meanwhile, Intel just announced that there is a delay in 10nm CPUs to 2019. Looks like there are lots of confusions and unpredictable factors. Here are my questions:
1. Does that mean those CPUs with Meltdown/Spectre bugs fixed at hardware level won't be out until next year?
2, In this situation, is the Threadripper platform the best way to go?
3. For the ROG Zenith Extreme motherboard, If I assign two high-end GPUs (1080Ti or higher up) to do CUDA computations and add a slower GPU to the system only to drive a 4K monitor, will this slower GPU become a bottleneck and slow down the other two high end GPUs?
4. I heard that on some motherboards, if you add some devices such as nmve m.2 SSD to the system, the performance of the GPU will slow down because new devices and existing GPUs compete for lanes and they have to share. How about this motherboard? At most how may nmve m.2 and SATA III SSD can I add without affecting the performance of the GPUs?
5. I read that unlike the other two ASUS X399 motherboards, this motherboard allows the use of nmve m.2 ssd to become RAM. What is the advantage doing so given that DDR4 RAM is much faster than SSD even of nmve m.2 type.
1. Does that mean those CPUs with Meltdown/Spectre bugs fixed at hardware level won't be out until next year?
2, In this situation, is the Threadripper platform the best way to go?
3. For the ROG Zenith Extreme motherboard, If I assign two high-end GPUs (1080Ti or higher up) to do CUDA computations and add a slower GPU to the system only to drive a 4K monitor, will this slower GPU become a bottleneck and slow down the other two high end GPUs?
4. I heard that on some motherboards, if you add some devices such as nmve m.2 SSD to the system, the performance of the GPU will slow down because new devices and existing GPUs compete for lanes and they have to share. How about this motherboard? At most how may nmve m.2 and SATA III SSD can I add without affecting the performance of the GPUs?
5. I read that unlike the other two ASUS X399 motherboards, this motherboard allows the use of nmve m.2 ssd to become RAM. What is the advantage doing so given that DDR4 RAM is much faster than SSD even of nmve m.2 type.