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

AMD Ryzen 9 3950X Review

anandtech.com

Post reply on HN
Print view View original

191–200 of 290 posts

Re: AMD Ryzen 9 3950X Review

#191

Post by jsgo »

redisman wrote: post #159
overcast wrote: post #102

Earlier quoted context omitted.

All easily avoidable.

I just build a mid-range gaming PC. CPUs don't have RGB, memory modules don't have RGB, some overpriced motherboards have RGB, HDD/SSD don't have RGB, some overpriced GPUs had them but I was looking for value for money.

on the GPU side, it seems RGB is the common thread. There are ways to disable it, but they aren't particularly great as they really only come into play post-OS loading (I have my GPU in an external enclosure so I see it more than I would in a desktop).

Memory modules do have RGB, but you specifically have to buy them (they're the exception, currently). CPUs don't have RGB, but the fans that are going to be mounted on them may be (Ryzen's CPU fans that come with the CPUs have them iirc). I don't see HDD/SDD going the RGB route considering they tend to be somewhat obfuscated, but the faster storage sticks that are all the rage in gaming PCs now may be a candidate (I say that mostly because of the RGB memory modules mentioned earlier).

It is a trend I wish didn't happen, honestly, but I know it is a case of trying to consolidate and appease two audiences (performance hungry users and enthusiasts) so it is what it is.

Re: AMD Ryzen 9 3950X Review

#192

Post by dragontamer »

neogodless wrote: post #173
qzw wrote: post #163

Earlier quoted context omitted.

64 cores ought to be enough for anybody...

Yeah, that's what... 10kb of RAM per core?

Funny story. The Vega64 has 8GB of RAM and 4MB of L2 cache (last-level cache) across 4096 SIMD-cores. That's 1kB of L2 cache and 2MB RAM per core.

It gets worse: although there are 4096 cores, the Vega64 isn't fully loaded until you stick 4-threads-per-core (Or more precisely, 16-wavefronts per Compute Unit (256-SIMD-cores)). That means you will actually need to run 16384 SIMD-threads before the Vega64 is fully utilized.

That's less than 512kB main-RAM per core, and less than 256 bytes of L2 cache per core. Better hope your threads are sharing a lot of memory...

--------

AMD addresses this problem with NAVI / RDNA: NAVI is fully utilized at 1-thread per core. So you only need 2816 threads on the 5700 XT to fully utilize the system.

Re: AMD Ryzen 9 3950X Review

#193

Post by gravypod »

bootloop wrote: post #155
fock wrote: post #96

Can anyone comment on the IOMMU-grouping on typical boards? From what I've just googled, the CPU "lanes" seem to support ACS so it could indeed be working to replace my IGP+GPU-for-the-gaming-VN-Haswell system with a dual-GPU-Ryzen one (contrary to what I believed previously). Have been eyeing TR for that reason, but as I don't really need this amount of I/O (and cores), I might be well served by AM4.

I am finally running a Ryzen 1700 + dual GPU system so I could get rid of dual-booting: CPU: AMD Ryzen 7 1700 Mainboard: GA-AX370-Gaming K5 (Bios: F25) GPU (Host): Gigabyte GeForce GTX 970 GPU (Guest): Gigabyte GeForce RTX 2060 OS (Host): Arch Linux OS (Guest): Win 10 As said it appears to me IOMMU grouping depends a lot on the board used. For mine it looks like all the chipset external PCIes are in one group and the…

Have you experimented with things like lookingglass? Is that stuff more streamlined now?

Re: AMD Ryzen 9 3950X Review

#194

Post by rarecoil »

velox_io wrote: post #127
agumonkey wrote: post #119

Earlier quoted context omitted.

In a similar vein, some local supermarket catalog showed half laptops based on ryzen, the only intel I saw were ultra cheap atom x5.

Wow, I didn't expect that. Laptops were the only market Intel was still strong, still dominated. Having brought an rx5700 XT recently, the power usage has been just as impressive as the performance. AMD will have some very nice products when they start putting mini RX5000s in their CPUs. Decent performance without discrete GPUs is a real possibility. (Just need add some HBM (High Bandwidth Memory) to the mix and thin…

I bought an RX 5700 XT almost immediately after launch, and I can't say that I've had had much luck with RDNA/Navi. I experienced freezes/crashes in Windows 10 while gaming, and the card's hot, loud, and still unsupported by ROCm for compute on Linux. There are rumours that ROCm won't ever support this generation of Navi and is waiting on Arcturus[1]. The drivers still aren't there and support is limited outside of some gaming use. Reviews also appear relatively mixed on Amazon.

That said, even though perf-per-watt isn't great and that's usually my #1 concern, I still buy primarily AMD GPUs and suffer with the issues, as the NV CUDA/DL monopoly needs a counterweight. I'm just sitting on Vega until Arcturus.

[1] https://www.phoronix.com/scan.php?page=news_item&px=Radeon-R...

Re: AMD Ryzen 9 3950X Review

#196

Post by mjevans »

velox_io wrote: post #121
gameswithgo wrote: post #109

Earlier quoted context omitted.

GCC at least goes very fast on the new Ryzens, due to the huge l3 cache. That may translate to many compilers. Rust (which uses llvm on the backend) also compiles very quick.

For the 4000 series, they're planning on sharing the L3 caches between CCD's/ chiplets rather than per CCX. It will be interesting to see if the more cache availible per core, and more cores being able to access the same cache will improve the performance. The speed AMD has raised caches & core-count lately is insane. When the max for mainstream CPUs was 8MB (and 4 cores) until only a few years ago. (Things seem to h…

It would be interesting if we moved back to the final layer of cache being off-core, likely within the IO die or attached as additional chips within the package. That should decrease the core size, improve yields, and allow for differentiation in L3 size as well.

Re: AMD Ryzen 9 3950X Review

#197

Post by tracker1 »

Been waiting for this for about a year now... pulled the trigger early on an X570 build as my old system (i7-4790K) was acting up and bought early. Been running an r5 3600, but replacing cpu with a 3950X.

Re: AMD Ryzen 9 3950X Review

#198

Post by nsxwolf »

ajuc wrote: post #107
jtbayly wrote: post #66

Earlier quoted context omitted.

Most people don't need a truck, maybe, but the three top-selling vehicles in the US are trucks. And in 2017 only 35% bought cars, with the rest being trucks and SUVs.

US is far from average needs of a car user. Not a representative sample.

Really? We're the #2 consumer of cars, about half of the annual car sales of China.

Re: AMD Ryzen 9 3950X Review

#200

Post by gameswithgo »

mjevans wrote: post #196
velox_io wrote: post #121

Earlier quoted context omitted.

For the 4000 series, they're planning on sharing the L3 caches between CCD's/ chiplets rather than per CCX. It will be interesting to see if the more cache availible per core, and more cores being able to access the same cache will improve the performance. The speed AMD has raised caches & core-count lately is insane. When the max for mainstream CPUs was 8MB (and 4 cores) until only a few years ago. (Things seem to h…

It would be interesting if we moved back to the final layer of cache being off-core, likely within the IO die or attached as additional chips within the package. That should decrease the core size, improve yields, and allow for differentiation in L3 size as well.

but it would add latency :( but maybe an L4 cache, off core, as additional chips!
Post reply on HN