What if budget is not a constraint and I want to squeeze maximum performance (without overclocking) for the kind of work I do: programming, data science, Jupyter Lab , machine learning, watching Blu-Ray movies, 1080p streaming from YouTube, Vevo etc. Which one will be good:
Given the both the 2400G and 2600 are nearly the same price, it would be pointless to do a build using the 2400 and a video card. Off hand I will say with the stuff you are doing get the 2600 w/ 16GB and a better then RX 550 dGPU.
Come to think off it, maybe the Ryzen 2700 with 32 gigs of memory will be a better choice. Don't forgot a nice SSD to go with the build.
Decent SSDs with MLC and TLC NAND and have dram cache will last longer then spinners do. The performance difference is night and day compared with HDDs.
In general, the behavior of GCC is substantially identical to that of Visual Studio. You will understand our astonishment when we saw these results. The performances of the Raven Redge APUs in this test are excessively low to say nothing else, we are more than twice as slow as a Ryzen 3 1200 and even the old A10 do much better.
We have long researched the cause of this counter performance, comparing several times the performance of APU with a Ryzen 3 (and configurations Ryzen 4 + 0) on different motherboards (by activating / deactivating the integrated GPU), different memory configurations, different versions of Windows 10, to no avail. We also did not notice a decrease in frequency that would explain this behavior.
We also contacted AMD who failed to explain the behavior we encountered confirmed that we reproduced our results.
What we can notice is that in general, the compilation is quite sensitive to the number of threads and the pressure on the cache and the memory sub system. Note that here the gap between a 2200G and a 2400G is smaller than usual, suggesting that extra threads benefit little and that we are rather limited cache / memory side.
The fact that the loss of performance is so drastic, and that we did not find in benchs that historically are limited by memory (we think of 7-Zip) makes us think that lack of a bottleneck, the problem may be more related to an algorithmic issue in cache management and / or memory access that would differ slightly from that used for classic Ryzen, having been adapted to take into account the Vega GPU integrated.
While waiting for a better explanation from AMD (who has been investigating the issue for several days), in which case we will update our article, we will stay on these assumptions and will be attentive in the following pages possible other against performance in this way.
Therefore, a better choice to me would be the Ryzen 5 2600 + RX 550 route, though it will end up costing significantly more. You could also have a look at the i3 8100 - for your needs the performance would be quite similar to that of a 2400G, and it doesn't suffer from the same kind of issues under compilation workloads which I've linked above.
I think most SSDs will outlive hard drives. Just look at the average lifespan of a NAS hard drive. Majority of them that are used heavily croak after about 3 years. Most SSDs can handle a heavy continuous read/writes much better than a mechanical HDD.
In general, the behavior of GCC is substantially identical to that of Visual Studio. You will understand our astonishment when we saw these results. The performances of the Raven Redge APUs in this test are excessively low to say nothing else, we are more than twice as slow as a Ryzen 3 1200 and even the old A10 do much better.
We have long researched the cause of this counter performance, comparing several times the performance of APU with a Ryzen 3 (and configurations Ryzen 4 + 0) on different motherboards (by activating / deactivating the integrated GPU), different memory configurations, different versions of Windows 10, to no avail. We also did not notice a decrease in frequency that would explain this behavior.
We also contacted AMD who failed to explain the behavior we encountered confirmed that we reproduced our results.
What we can notice is that in general, the compilation is quite sensitive to the number of threads and the pressure on the cache and the memory sub system. Note that here the gap between a 2200G and a 2400G is smaller than usual, suggesting that extra threads benefit little and that we are rather limited cache / memory side.
The fact that the loss of performance is so drastic, and that we did not find in benchs that historically are limited by memory (we think of 7-Zip) makes us think that lack of a bottleneck, the problem may be more related to an algorithmic issue in cache management and / or memory access that would differ slightly from that used for classic Ryzen, having been adapted to take into account the Vega GPU integrated.
While waiting for a better explanation from AMD (who has been investigating the issue for several days), in which case we will update our article, we will stay on these assumptions and will be attentive in the following pages possible other against performance in this way.
Therefore, a better choice to me would be the Ryzen 5 2600 + RX 550 route, though it will end up costing significantly more. You could also have a look at the i3 8100 - for your needs the performance would be quite similar to that of a 2400G, and it doesn't suffer from the same kind of issues under compilation workloads which I've linked above.
Microsoft Visual C++ compiler is a single threaded application. So yeah, you are right that Ryzen does not perform well on single threaded applications.
I use Linux and LLVM compiler has removed this bottleneck. I don't have to worry about it. But then I got another question: For single threaded performance, dedicated graphics card like RX550 performs better than 2400G ?
RX 560 outperforms Vega 11 by a good margin. But then I will need to change 2 more things: PSU and UPS: Corsair VS550 instead of VS450 and APC BX1100 instead of BX600
If budget is not a constraint at all, I would be looking at a 2700x and an RX Vega 64. Or hell go Threadripper, a 2950X or 2990WX + Vega FE (liquid cooled) will blow you away.
Restricted between the given choices, go with choice #2.
If budget is not a constraint at all, I would be looking at a 2700x and an RX Vega 64. Or hell go Threadripper, a 2950X or 2990WX + Vega FE (liquid cooled) will blow you away.
Restricted between the given choices, go with choice #2.
That is really a good advice DrMrLordX. I always buy one level below than the current market (Ultimate Linux Box Guide). Since Ryzen 7 is best in CPU and VEGA in GPU. I want to get one level below. Hence, Ryzen 5 2600 and GPU ... I am just guessing RX 560 is one level below VEGA
use camel camel camel... And track the 1950X thread ripper, it was 450 bucks not too long ago... Absolutely crushes anything Intel has on the market. Also, if you check the hot deals section you can pick up a water cooled evga 1070 for 330 bucks. A pretty nice build for under a 1200 bucks. Of course you'll need some fast RAM and a SSD.
Microsoft Visual C++ compiler is a single threaded application. So yeah, you are right that Ryzen does not perform well on single threaded applications.
I use Linux and LLVM compiler has removed this bottleneck. I don't have to worry about it. But then I got another question: For single threaded performance, dedicated graphics card like RX550 performs better than 2400G ?
RX 560 outperforms Vega 11 by a good margin. But then I will need to change 2 more things: PSU and UPS: Corsair VS550 instead of VS450 and APC BX1100 instead of BX600
The issue I talked about occurs in GCC, not Visual Studio. I don't know whether it is prevalent in LLVM but keep in mind that there might be pathological cases where cache contention could be an issue with the 2400G since it is a 4C8T processor but with only 4MB of victim L3$.
The single threaded performance would be nearly equal between the 2400G and the 2600. The GPU has no part to play in determining CPU single-threaded performance.
The issue I talked about occurs in GCC, not Visual Studio. I don't know whether it is prevalent in LLVM but keep in mind that there might be pathological cases where cache contention could be an issue with the 2400G since it is a 4C8T processor but with only 4MB of victim L3$.
The single threaded performance would be nearly equal between the 2400G and the 2600. The GPU has no part to play in determining CPU single-threaded performance.
Solid build, I would suggest though either upgrading that Coolermaster 212 to an even beefier cooler, or switching to the 2600x and dropping the aftermarket cooler altogether. I'd lean towards 2600x since it basically does auto-overclocking (paraphrasing). The 2600x cooler is almost as good as a 212 but its bundled in at no cost. Dropping the 212 should cover most the cost gap between the 2 and you'd for sure gain some single thread clock speed turbo headroom. The 2600 (no x) cooler is weak so it makes sense to replace it if you want to stick with 2600.
Solid build, I would suggest though either upgrading that Coolermaster 212 to an even beefier cooler, or switching to the 2600x and dropping the aftermarket cooler altogether. I'd lean towards 2600x since it basically does auto-overclocking (paraphrasing). The 2600x cooler is almost as good as a 212 but its bundled in at no cost. Dropping the 212 should cover most the cost gap between the 2 and you'd for sure gain some single thread clock speed turbo headroom. The 2600 (no x) cooler is weak so it makes sense to replace it if you want to stick with 2600.
This site uses cookies to help personalise content, tailor your experience and to keep you logged in if you register.
By continuing to use this site, you are consenting to our use of cookies.
Facebook
Twitter