Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

AMD Ryzen price and release date revealed

pcworld.com

Post reply on HN
Print view View original

101–110 of 139 posts

Re: AMD Ryzen price and release date revealed

#101

Post by vondur »

emn13 wrote: post #82
tyingq wrote: post #30

Earlier quoted context omitted.

I would guess at least some meaningful purchases, if only to keep Intel honest.

These chips look better suited to servers than clients as is. Most software still scales poorly with additional cores (or rather: most workloads still have a few single-core bottlenecks). E.g webbrowsers, compilers, video de and en-coders, all kinds of stuff. So even if you have twice the cores, catching up given the handicap of a single-core perf disadvantage is a pretty harsh battle. E.g. see http://www.anandtech.c…

I have a 5820K (6 core 3.3 ghz base) and using Handbrake to encode a HD .mkv file from MakeMKV, the 6 core machine beat the 7700k clocked at 4.2ghz. Once I overclocked the 5820k to 4Ghz, it was about 10-15% faster than the 7700k which I had then clocked to 4.6Ghz. So I'm looking forward to the 1700x. Not all programs scale as well as Handbrake does with more cores.

Re: AMD Ryzen price and release date revealed

#102

Post by some-guy »

icefo wrote: post #93
toxican wrote: post #43

I'm far too broke to shell out cash for a new Mobo and CPU right now, but I'm excited to see what this does to Intel's prices for older CPUs. I'd love to upgrade my i5-2400 soon-ish.

Unless you are regularly doing CPU intensive tasks you can probably stick with it a few more years. I noticed since I bought a SSD that most of time the hard drive is limiting the CPU. Until my DVD reader burned my desktop motherboard I was using a core2quad and it was still fast enough. No lag whatsoever. Now I'm using a laptop with the first gen I5 and I feel no need to upgrade it (I'm using it as a desktop battery…

I bought an i7 desktop CPU from 2008 and have absolutely no inkling or reason to upgrade. I never thought I would be using the same CPU for almost 10 years, but here we are.

Re: AMD Ryzen price and release date revealed

#103

Post by drzaiusapelord »

kayoone wrote: post #80
mrb wrote: post #10

First (amateur) third-party benchmarks confirm a ~$350 Ryzen beats $1000 Intel processors in CPUMARK, 3DMark Fire Strike Physics, Cinebench: https://www.chiphell.com/thread-1706915-1-1.html AMD had said for years their Zen goal was a 40% IPC gain over their previous microarchitecture, but they ended up with a 52% gain: http://www.anandtech.com/show/11143/amd-launch-ryzen-52-more... Today's launch event by AMD's CEO:…

just be aware that Ryzen is probably operating much closer to it's limit than the Intel counterparts, so overclocking could be limited. If that is not interesting for you anyway, then god speed. Also in terms of gaming performance, Intel will probably remain king, due to single threaded performance and OC potential. Personally, i'd always take 8 cores vs 4 for the same price even if they are 10-20% slower.

Do people even overclock anymore? I think the mass overclock gains of the 2500k days are probably behind us as even intel is getting close to its top end due to its poorly performing architectures. I thought the whole idea behind turbo is that its an official overclock and once the chip gets too hot it moves back to 'stock' speeds, then back up, etc. All you need to worry about is keeping the stock cooler or not.

Re: AMD Ryzen price and release date revealed

#104

Post by kllrnohj »

masklinn wrote: post #74
dstaley wrote: post #52

Earlier quoted context omitted.

The linked benchmarks are putting the $350 1700X about on par with the $1100 6900K when it comes to both single-threaded and multi-threaded performance. Both of those processors are 8-core/16-thread chips. Also the 1700K is a 95W TDP compared to the 6900K's 140W TDP.

Again, the 6900K is a two years old µarch, and not one which was considered a great release in the first place. The 1700X will not be competing against the 6900K because nobody gives a rat's ass about the 6900K, the 1700X will be competing against the 7700K and its ilk.

Kaby Lake and Skylake are indistinguishable in terms of IPC.

Kaby Lake only brought GPU & video decoding improvements to the table, that's it. CPU performance is otherwise identical to Skylake.

Re: AMD Ryzen price and release date revealed

#106

Post by emn13 »

tyingq wrote: post #89
emn13 wrote: post #82

Earlier quoted context omitted.

These chips look better suited to servers than clients as is. Most software still scales poorly with additional cores (or rather: most workloads still have a few single-core bottlenecks). E.g webbrowsers, compilers, video de and en-coders, all kinds of stuff. So even if you have twice the cores, catching up given the handicap of a single-core perf disadvantage is a pretty harsh battle. E.g. see http://www.anandtech.c…

In the comparison you linked, the 4 core CPU is clocked at 4.2GHz, and the the 8 core at 3.2GHz. If you pick a comparison where the clock speed isn't so far apart, things look different: http://www.anandtech.com/bench/product/1832?vs=1729 Of course, fewer cores usually means a higher available clock speed. Just pointing out it's not just "code isn't parallel enough" that drives the disparity.

Sure, the 8-core runs at a lower clock speed and and old arch - that's a major (but unavoidable) part of the problem, as I said.

To be specific: you can't get the equivalent of a 7700k with eight cores - the 6900k is (AFAIK) the next best thing, and by the looks of it, ryzen for all it's revolutionary improvements wrt bulldozer, is not all that different from a 6900k. AMD's own numbers place it at an identical single core perf.

Re: AMD Ryzen price and release date revealed

#107

Post by kllrnohj »

kayoone wrote: post #80
mrb wrote: post #10

First (amateur) third-party benchmarks confirm a ~$350 Ryzen beats $1000 Intel processors in CPUMARK, 3DMark Fire Strike Physics, Cinebench: https://www.chiphell.com/thread-1706915-1-1.html AMD had said for years their Zen goal was a 40% IPC gain over their previous microarchitecture, but they ended up with a 52% gain: http://www.anandtech.com/show/11143/amd-launch-ryzen-52-more... Today's launch event by AMD's CEO:…

just be aware that Ryzen is probably operating much closer to it's limit than the Intel counterparts, so overclocking could be limited. If that is not interesting for you anyway, then god speed. Also in terms of gaming performance, Intel will probably remain king, due to single threaded performance and OC potential. Personally, i'd always take 8 cores vs 4 for the same price even if they are 10-20% slower.

> just be aware that Ryzen is probably operating much closer to it's limit than the Intel counterparts, so overclocking could be limited.

Overclocking is also limited on the Intel counterparts, so that's not really a big deal. Most kaby lakes won't do 5ghz, and good ones hit the wall at 5ghz. That's not much of an increase over the stock 4.2/4.5ghz for the 7700K

Also AMD is claiming a lower TDP than Intel, so that would suggest Ryzen has more OC headroom than Intel has. Although of course this is all wait-and-see.

Re: AMD Ryzen price and release date revealed

#108

Post by emn13 »

vondur wrote: post #101
emn13 wrote: post #82

Earlier quoted context omitted.

These chips look better suited to servers than clients as is. Most software still scales poorly with additional cores (or rather: most workloads still have a few single-core bottlenecks). E.g webbrowsers, compilers, video de and en-coders, all kinds of stuff. So even if you have twice the cores, catching up given the handicap of a single-core perf disadvantage is a pretty harsh battle. E.g. see http://www.anandtech.c…

I have a 5820K (6 core 3.3 ghz base) and using Handbrake to encode a HD .mkv file from MakeMKV, the 6 core machine beat the 7700k clocked at 4.2ghz. Once I overclocked the 5820k to 4Ghz, it was about 10-15% faster than the 7700k which I had then clocked to 4.6Ghz. So I'm looking forward to the 1700x. Not all programs scale as well as Handbrake does with more cores.

Sure, there are situations in which it's faster. In my (limited) experience with x264 at least the more extreme quality settings didn't scale quite as well, and those are the only ones I use. Regardless: some programs parallelize quite well, others not so much.

If you're lucky enough to primarily be bottlenecked on parallel stuff: go nuts, of course!

In any case, I don't want to suggest I'm not happy with AMD finally being a real competitor, I'm just trying to temper expectations of intel-beating world domination. Most users aren't so lucky. Many people use laptops, and slow websites are more of a problem than slow handbrake sessions. For me, it's visual studio and various build/run cycles. Sure, some tools use more cores with the right settings, but it typically scales really disappointingly after around 2 or so. That overclocked 5820K would probably lose to a stock 7700k in my workload, and that's just annoying.

Re: AMD Ryzen price and release date revealed

#109

Post by Retric »

cma wrote: post #99
RussianCow wrote: post #88

Earlier quoted context omitted.

In my experience, most modern games aren't CPU-bound anymore; more and more work is being pushed onto the GPU with every generation. My brother has a quad-core AMD CPU (don't remember which) with a powerful GPU, and his CPU is very rarely the bottleneck—there are few games he can't max out at 1080. Also, according to AMD's benchmarks, the Ryzen CPUs are very close to Intel's in single-threaded performance.

It is a big deal in VR, at 90Hz, two eyes, you have 3x the draw calls of a game running at 60Hz. There are some optimizations to reduce the per-eye overhead, but it is still a lot. Sony's headset can run at native 120Hz, increasing things more, but not too many games are using that mode.

4k at 60fps is 498 million pixels per second, Rift @ 90FPS is 233 million pixels per second. GPU's can deal with making 2 images without meaningful CPU overhead. The real issue is stutter and latency not CPU or GPU GFLOPS.

Re: AMD Ryzen price and release date revealed

#110

Post by Arelius »

Retric wrote: post #109
cma wrote: post #99

Earlier quoted context omitted.

It is a big deal in VR, at 90Hz, two eyes, you have 3x the draw calls of a game running at 60Hz. There are some optimizations to reduce the per-eye overhead, but it is still a lot. Sony's headset can run at native 120Hz, increasing things more, but not too many games are using that mode.

4k at 60fps is 498 million pixels per second, Rift @ 90FPS is 233 million pixels per second. GPU's can deal with making 2 images without meaningful CPU overhead. The real issue is stutter and latency not CPU or GPU GFLOPS.

Pixels/second aren't a very useful metric when speaking about CPU performance. Your comparison seems to show that CPU performance might be much more important than GPU perf for VR.

Also, you're correct that GPU's can play all sorts of tricks to get multiple projected views for low CPU-overhead. But that is with significant software support. Otherwise VR will have to build twice the amount of framebuffers/frame as a single view game.

Post reply on HN