PaulAlcorn :
mapesdhs :
A gaming-focused review for parts like this is bizarre. For a more sensible writeup, see:
https://www.phoronix.com/scan.php?page=article&item=amd-linux-2990wx&num=1
Gaming is not remotely the relevant market for these CPUs, so implying it is by showing the gaming results first is silly. Hardly a wonder everyone is banging on about the 2950X instead. Do some more relevant tests and the 2990WX will shine, but as with anything, the right tool for the right job. Besides, it's going to be a while before OS variants can adapt to tech like this, ditto optimised BIOS/chipsets and especially applications. Kudos to AMD for actually pushing things along though, unlike Intel which sat on its butt for so long.
This is hardly a gaming-focused review. We tested eight games. Compare that to the 41 application tests. Including SPEC workloads for workstations. The games are listed on the first pages, sure, but that is because we apply the same tests and format to all CPUs.
Threadripper 2 X-Series is specifically marketed to gamers. Yes, gaming is relevant.
It's gaming focused if the games are tested first, and a lot of the app testing isn't really all that relevant to where or how this class of CPU would be most useful. They may be marketed to gamers (what a shock) but that won't be their natural core audience. Just because AMD is marketing a CPU in a particular way doesn't mean that actually makes any sense. I was the biggest SGI fanboy around back in the day, but SGI's PR dept. was sometimes completely whacko, claiming all sorts of stuff that wasn't true (or at best misleading), and they were dealing with tech oodles more expensive. I guarantee the engineers who designed these chips think AMD's PR dept. is crazy.
These are not gaming CPUs. To claim otherwise is to ignore the fact that the stock 2990WX performs comparatively poorly for every single one of the gaming tests (remember PBO means operating out of warranty, if I read the info correctly). Can't have it both ways; if the 2990WX is being touted as a gaming CPU then the data suggests it's ruddy awful, so surely instead it makes more sense to say it just isn't that kind of product at all, AMD's marketing is nuts. Have you looked into science focused tasks that could properly push hw of this kind? That would be far more interesting, especially with ECC. Stuff like FEA, CFD, CQD, etc. The 2950X does better for gaming, but that's no surprise, though again the total cost still ignores the more sensible mainstream chipset/CPU choices.
There's a vast range of tasks which would be more suitable for working out what the 2990WX could be really good for, but few of them were tested here. Production rendering isn't Handbrake. Have you looked at how Alfred works on Centos? A cluster of these things could be very potent indeed. What about v-ray? There are so many other possibilities for which this CPU could be a real winner, but in the end it's AMD's fault if they're aiming the marketing at entirely the wrong audience. The kind of people & companies who'd love this CPU in a workstation are exactly those to whom I provided SGI technical & upgrade advice for 15 years while that tech was still current. Sure, such users would go EPYC if they could afford it, but many can't (especially solo pro types, academics on tight budgets, etc.), and for them this could be the best thing since sliced bread. Academic researchers inparticular - biochem, GIS, physics, fire & explosion studies (built environment), all sorts. I've dealt with hundreds of such people over the years, all too often their workstations are budget constrained compromises of the ideal (in some places, just off the shelf consumer builds).
The more pertinent question is why the heck is AMD pushing them as gaming products, it makes no sense. 5 seconds thinking ought to conlude that buying a mainstream platform with an 8700K or looming 9900K (or 2700X, whatever) and spending the cost difference on a more powerful GPU is way more logical for gaming, and there'd be change enough for a 970 Pro C-drive.
Reviews used to be more forceful in pointing these things out. Nobody should be advising consumers that any of these TR2 CPUs make sense for gaming, they really don't. I'm delighted that AMD is back in the game, but I'm not going to tell a friend to get a TR2 for gaming when a Ryzen or CL is far more sensible, and one can't point to SLI/CF because the utility of those technologies has been severely diluted in recent years, with poor driver support and even hw lockouts from some quarters.
I mean ye gods, the cost of just a 2990WX (1640 UKP in the UK) is enough to build an entire very nice gaming machine.
Add in the cost of the RAM & mbd and that's enough for even a prebuilt i7 8086K rig with a 1080 Ti (code LN84893 on Scan). The 2950X does better as I say, but the platform cost difference is still large enough to mean a mainstream build with a better GPU will always be faster for
gaming. If one tries to justify such a purchase by saying yes but they would be better if one also wanted to do some content creation, well in that case they marketing focus on gamers is no longer there. Fact is, one can build a far more potent
gaming system on the same budget by going mainstream, especially for those playing at higher resolutions (since they won't be so dependent on CPU performance anyway, in which case a non-K CPU is fine and hence an even higher available budget for a better GPU). And remember that even for content creation, these days GPU acceleration has become a lot more relevant, eg. the system I built for the
Learn Engineering channel on youtube is eight times faster than their existing best system due to decent GPU power (CUDA in Blender) that blew away the CPU-based setup they'd been using until then.
Btw, SPEC has been flawed for a long time, and CB OGL has been broken for years.
Ian.
Facebook
Twitter
Instagram