Return to

AMD or Intel, what should you buy? What's the best deal? Q3/Q4 2017 | Level One Techs


It wasn’t so much about the encoding performance. She’s been annoyed that she’s dropping frames in ark specifically, and when I looked at it she had pegged cores and 80% usage on her 980Ti. I was surprised to see it as anyone.


Well done and I was there for the lulz, @wendell’s laugh (reminds me of dr Evil, but in a good way :smile:)

My upcoming (second 1950X rig) which I’m going to call the 'Ballin-AF Ripper Workstation’ puts me in the insane category - 1950X + ROG Zenith Extreme + EVGA GTX1080Ti FTW3 + (by Jan/Feb 2018) Radeon Vega 64 as host (Fedora) GPU.

I’m already giddy, just waiting for the parts to arrive.

@wendell @MisteryAngel any news if these current boards will get TB3 support?


I recommended the 1600x to a lot of friends, most of them went 1700(non-x) instead because moar cores, moar better.


I have a friend who is just getting into PC gaming. I recommended the 8600. It does beat the r5 in gaming related bench marks, but he will be ok as move into 6 thread optimized titles. For now he can use his extra threads to watch netflix while gaming or whatever.


Do you mean i5 8400, or the i5 8600k?
The K is unlocked multiplier, but also higher factory clocks (1core 4.3 vs 4.0, all core 4.1 vs 3.8) although if you have a board that supports MCE (Multi-Core Enhancement) or equivalent from another brand you can get the single core turbo on all threads, even on non-k parts.




Why is the Ryzen 1600/x mentioned instead of i5 8400? Once the cheap boards come out buying the locked SKU would be cheaper than buying a 1600 with a decent mobo for oc. The i5 8400 performs competitively in non-gaming workloads and definitely better than OCed Ryzen in gaming.


Because the i5 8400 is not as good of a deal as the r5?
You know you can get decent b350 boards like the msi tomahawk for around $40 on sale at places like microcenter?


just waiting for the growing pains of AMD and there new platforms to settle out. then i am looking at ryzen+ what ever the server chip will be. in all honesty i am happy with they way things are with competition but all of my itches have not been scratched by either team at the moment.


Yep, in the US anyway, its much closer in price here in Australia on the mobo, but the stock cooler is much better on the AMD non-k chips too… which saves another $20-30 off getting a cheap 120mm tower cooler I’d honestly recommend on the intel 6 core non-K, especially if exploiting Multi-Core Enhancement or the like.


Fair point about the motherboards but from where I am i5 8400 and the 1600 have identical (and I mean it) prices. Say, later you can buy both mobos at the same price, the i5 will beat ryzen in gaming and be competitive in other workloads. Ryzen beats intel in benchmarks like 7zip, winrar and intel beats ryzen in handbrake or video encoding in general. A quick look at Anandtech benchmarks suggests that the i5 beats ryzen in gaming, encoding, web tests, and system tests ; ryzen and intel sort of cancel each other out in pc mark 10 ; Ryzen beats intel in 7zip, winrar and rendering tests like pov-ray, luxmark,etc that you should be able to offload to the gpu (the rendering tests, that is).
So isn’t i5 better?


Is this the problems with ryzen gpu passthrough you were referring to? , or was it still related to the IOMMU grouping?

I’m still waiting around to see if they manage to fix it, or if I’ll need to keep using my cruddy old windows box for steam streaming to my Linux one.


I wish I had an AMD chip to test with. :confused:

@wendell, will you test the patch in return for an internet cookie?


Ehh, it still sounds like a bit of a stretch. If the price is exactly equal and you get a free cooler comparable to the 1600 cooler then possibly? Motherboards that have multicore enhancement can also “oc” the non k chips, which might tilt it back toward intel.

However, even assuming all these hypotheticals is the i5 on equal footing with the r5. In that case the am4 platform’s longer-livedness I think would also be worth something in the value/best bang for the buck discussion.

It’s also worth pointing out that the gaming benchmarks where the i5 is better are generally situations where you are not gpu bound. For people gaming in gaming scenarios where they are cpu bound, there is not enough headroom to justify picking the intel r5. I would say for people specifically interested in that workload, they must pick a CPU that has some better margin vs the AMD parts – that is not the i5 parts unfortunately.

Plus making recommendations on hypotheticals is a bit silly. Right now, today, the r5 is better.


When on a budget, you are NOT, I repeat, NOT on anything beyond the 1070 (and that is a stretch).
So all the benchmarks for budget cpus with high end gpus (1080ti or better) are worthless.


Yes, been testing it, ryzen R5/7 is generally fixed, especially when paired with a x370 chipset motherboard.

Threadripper still has some issues: Bus reinit problems, amd vi completion timeouts, etc. I’ve been testing the last few days.

The weird thing about the patch is the fps are limited with the official patch but the performance and “smoothness” is very good.

Changing the svm parameters manually can fix the fps issue but introduces other issues.


Is it safe to assume EPYC has the same issues?

This is V1 of the patch. It’s the first real breakthrough on the issue since it’s happened. I’m optimistic that we’ll get improvements in the next few weeks/months.


yes and no. For things like SR-IOV and actual enterprise stuff, the issues seen on threadripper are unlikely to be encountered even with pcie passthrough.


If only there were more channels that could survive without it, I mean we once had an Internet without it.


I’m toying with building a few openstack compute nodes with EPYC, so that’s why I ask.