Printable View
Sure it is; today. Like I already said above.
I'm well aware of that particular test and several others like it, but my point, which was apparently missed, was this: "in two, three or more years I'm absolutely expecting that x8 3.0 slot to hold me back significantly. I could be wrong about that of course, but I'm not willing to take the chance".
The latter sentence perhaps in particular.
So i just came across this article on WCCFTech that claims they've confirmed the bug in Windows 10 with the scheduler and also, interestingly, the way Windows 10 incorrectly identifies the amount of cache.
As every review (afaik) benchmarked using Windows 10 does that make the results invalid?
Cross that bridge when you get to it......future-proofing a gaming machine is not worth it.Quote:
Originally Posted by Agrippa
Invalid? No - the reviews are showing the performance you will receive if you run Ryzen under Windows 10 today.Quote:
Originally Posted by Corky34
Does it make Ryzen look worse than it potentially could? Yes, probably. But AMD chose to release Ryzen before the Windows scheduler was patched (and frankly I just plain don't believe that they didn't know about this issue before the hard launch date). I'm sure they made that decision for the right business reasons, but that means they accepted in advanced that reviews would show sub-optimal performance in lightly threaded loads.
Best case scenario for AMD is that these patches, fixes and drivers are out in time for the Ryzen 5 launch. I think most gamers will be interested in the 4C/8T parts as an alternative to i7s at i5 prices, and the reduction to a single CCX plus a fixed scheduler ensuring that lightly threaded loads use the primary cores could make those look very tasty indeed...
I also wonder if AMD had half an eye on the Chinese market when they planned the Ryzen 7 launch? AFAIK Win 10 adoption has been particularly slow in China.... ;)
To be fair though Ryzen's launch was probably set in stone a good while before launch then it seems Microsoft surprised everyone by canceling February's patch Tuesday, I'm not implying it's a conspiracy or anything but for all we know this bug should or was going to be fixed 3 weeks before launch.
True enough, but since I said, or at least heavily implied, that it's not a gaming machine (though it's also used for gaming) I much prefer that all bridges that can be pre-crossed are pre-crossed on the day I build it.Quote:
Originally Posted by shaithis
I dunno - the state of the motherboards are dire too. I am going to tell my mate who is most likely going to get an R7 1700 to hold off a bit. This is why I am perplexed why AMD launched them now - even Jayztwocents who actually gave the CPUs a positive review said he could not recommend buying Ryzen quite yet since the motherboards were having a lot of issues.Quote:
Originally Posted by Corky34
The user csbin over on the AT forum came up with this rumour from Chiphell about an AMD X399:
https://forums.anandtech.com/threads...BC%81.2501129/
Seems a bit unlikely, but it's basically meant to be workstation competitior to Intel's upcoming X299 (LGA2066). Can't get Google Translate to work on that as it says it causes an infinite loop though. Works okay in Chrome though.Quote:
16C32T
8CH DDR4
200W TDP
X399
Intel might be in a spot bother if AMD also starts doing that,and also after seeing this news from MS:Quote:
Originally Posted by kompukare
https://www.theregister.co.uk/2017/0...rver_followup/
8 channel memory would only be possible from a 4 die MCM. The chance of the Naples platform not having a 16C/32T entry level chip is, IMNSHO, zero: 4 dies each with one CCX disabled gives you 16 cores. Clock those up to the levels that would generally produce a 45W TDP (say, 3GHz base clock?) and you'd have a 190W TDP. Previous gen Opterons topped out at 140W for server, but given the FX 9000s, I wouldn't put it past AMD to do a high clocked performance part for the workstation market.Quote:
Originally Posted by kompukare
So basically the only questionable bit is the X399 chipset. And if that exists, it looks like it should therefore support the full range of Naples-based server processors. I can easily see them having two platform names though, one that only supports 1P and one that supports 1P and 2P - it'd be the same socket and processors, just tweaked chipsets (don't Intel do that, with the enthusiast X- chipset only supporting 1P even with Xeons, while the server chipsets supports multi-processor?).
I wonder if AMD are still quietly pursuing their ARM-based Opteron chips... ;)Quote:
Originally Posted by CAT-THE-FIFTH
I think something along that line as a while ago I went to see how expensive dual Xeon is while messing around with uk.pcpartpicker.com and was a bit confused with the options even when only looking at Broadwell-E parts.Quote:
Originally Posted by scaryjim
Just seen LegitReviews benchmark suggesting that memory speed is not making a significant difference in 3D.
http://www.legitreviews.com/ddr4-mem...-cpus_192259/4
Will be interested to see more thorough reviews on this when it's available. If there are better places to spend your money, why not?
At 1080P it does as at qHD the system is GPU limited with a GTX1080.Quote:
Originally Posted by jim
Similar story though on FireStrike, Handbrake and 7-Zip which I doubt are GPU limited.Quote:
Originally Posted by CAT-THE-FIFTH
Its kind of the same with SKL/KL - it was certain games which showed memory bandwidth could help improve FPS,but apparently not much else. At 1080P you are seeing a 20% improvement due to memory bandwidth alone. However,as you will soon find out Deus Ex is more a GPU stress benchmark - there are better games to probably test memory bandwidth. I would say the large open world ones like The Witcher,FO4,etc.Quote:
Originally Posted by jim
Cough,anybody notice this:
https://twitter.com/20thcenturyfox/s...83776914173952
http://i66.tinypic.com/2n67qrr.jpg