3950x got a release date 25 nov!

Changing CPU shouldn't be a problem, changing mobo is what requires reinstall.

 

...but it doesn't require a reinstall.. everything is working fine here.

 

People are just trying to help. You say "everything is working fine", but you also said your 4 year old i5 6600K ran better than your new Ryzen 7 3800X. If you simply moved the boot drive from your own system to your new system without re-installing Windows fresh, then you are probably not getting the full performance out of your new system.

 

but that 6600k was at 4.7ghz.. all cores.. ryzen 3800x on stock cooler..does not maintain even its rated 4.5ghz boost speed in a consistent manner.. https://hexus.net/tech/news/cpu/134345-amd-ryzen-3000-boost-survey-worse-expected-der8auer/ ..not saying it was a bad upgrade 8cores ~4ghz > 4cores 4.7ghz ..losing ~15% fps isn't so bad.. just not much of an upgrade in the gaming department (downgrade more like) ..so if you have an fairly recent less than 5year old intel and overclock it.. no need to upgrade unless you need the extra cores.. I didn't really need it until the old setup broke so had to take the upgrade early than I wanted too.

I have the new watercooling brackets for am4 socket now, so that will be the next step..to see if low temps will allow it to keep higher boost all the time. I don't want the ballache of setting up winshi10 again (only because I tweak so much of it because the default install and settings for everything is rubbish)... and neither do I see an ounce of difference will come from reinstalling that crap.. when all system drivers are working fine including mobo stuff.

The other thing is ddr4 @3000mhz is fine for intel.. while amd chips bawlk at it.. again I don't care to sell 32gb of ddr4 ram at that speed just to buy something that is slightly faster and works better for amd cpu's..to get what a 10% fps increase? not worth it.. still amd cpu's are better price/multi core performance right now for sure.

 

Ah. Stock cooler. Completely understand now. Replace that with a Hyper 212 EVO. Problem solved.

https://www.amazon.com/Cooler-Master-Hyper-Direct-Contact/dp/B005O65JXI

 

I have watercooling, just haven't gotten around to installing it... I do hope it improves overall boost performance.

 

@AndersMalmgren how does baking sound worked as a structure, especially relative to emitter and receiver?

 

It bakes the scene into Voxels then it does some magic that stores this without completly eating all RAM and also in a format thats quick enough to query in realtime.

https://docs.microsoft.com/en-us/azure/cognitive-services/acoustics/what-is-acoustics

 

The stock cooler from AMD are pretty OK, i use it with a 2700x at dayjob which I runt OC at 4.25 ghz without problem. But at home office I offcourse have a custom waterloop

 

Zen 2 behaves differently from Zen+. Heat is centered around certain points on the CPU's heatspreader making it harder to cool with the stock cooler. Hardware reviewers have been recommending custom cooling with the 3700X and higher.

 

But its the same with zen+ thanks to the infinite fabric. Maybe there are some other differences im not aware of though. My CPU block have survived 4 builds including my future 3950x build, pretty good value for money there

 

No, it has nothing to do with the Infinity Fabric. A chiplet design means that the silicon making up the processor is only connected by the heatspreader for purposes of dissipating heat whereas a monolithic design has silicon in addition to the heatspreader. A chiplet can't spread the heat out as evenly and you end up with potential cooling problems.

 

Grabbed the 3900X with RX5700 when AMD first released it few months back. Overall synthetic benchmarks hit through the roof. In terms of gaming, video encoding its beats the intel hands down....very impressive! Hoping that this beast can increase my unity poductivity, by cutting down the long building time.

But. i must caution those interested to seriously reconsider/benchmark before commiting $$$ into the hardware. As have done several handreds of builds since then on several machine i7/i9 versus the AMD 3900x, side-by-side. The results with the AMD 3900x with best money can buy SSD, is only marginally faster then a intel i7/8750H Laptop using custom build scripts. On latest Win10, Unity2018, it took the AMD 30 minutes complete a build, versus 32 minutes on an intel....So who's the lemon?

 

Try bake a large scene on both and you will see a completely different delta

 

Exactly this, on multithreaded workloads AMD has a unbeatable price / performance ratio.

 

Generate Lighting, reimporting the projects to another platform is indeed able to fully utlilise all the AMD 3900X 12 cores...observed difference can be >200% for some projects. Most of the heavy lifting works can be farm out these machines, except for the release daily builds handicaps, which is still puzzling me... why a i7 laptop can be as fast as the top of the line AMD (as of today 3950x is not commercially available). Guess nothing is perfect....as the saying goes "no ones gets fired for recommending intel" ;-)

 

For a large project is more loading the project, ReSharper etc that benefits a better CPU. Though compile time should reduce some for a large project because of the large cache on the 3900. Try the orchard core benchmark

https://www.hanselman.com/blog/Buil...C30ThePartsListForMyNewComputerIronHeart.aspx

 

AMD and Intel have nearly identical single-threaded performance, but what Intel has that AMD doesn't have is a very refined process. Intel has refined 14nm to the point that it's able to hit high clock speeds in mobile devices without excessive heat output so while you won't see the full boost you'll see most of it.

Speaking of process nodes most of this will change when Intel makes the leap to 7nm. Intel has only been able to achieve the clock speeds they have thanks to refining 14nm. Assuming a miracle doesn't happen (AMD wasn't expecting the clock speeds they achieved) Intel will have to start mostly from scratch when it comes to high clock speeds.

 

During that build, how multi-threaded is the overall activity? For example, open task manager, open the Performance tab, select CPU, and then right click on the graph and select Change graph to Logical processors. Then start your build and watch how much CPU loads are spread between threads.

If the loads are mostly confined to only a few threads, then the laptop would actually be able to keep up, since your laptop (Intel i7/8750H) can boost up to 4.1 GHz. That boost speed is almost as high as the boost speed on the 3900X, so it would make sense that it could get fairly similar performance if the workload is not heavily multi-threaded.

 

3900x kills the Intel in IPC though

 

Completely depends on how you define IPC. The 3000 series has the cache as high as it does because of the latency that comes with accessing main memory. If we're going to include the cache in our calculations (and we don't have much of a choice) we need to include memory latency too.

For tasks that can take full advantage of the cache they are faster, but once that's no longer the case the memory latency of the processor kills it compared to Intel.

https://www.techspot.com/article/1876-4ghz-ryzen-3rd-gen-vs-core-i9/

 

Do we have any way to benchmark real-world Unity performance on some of these new chips? Should we come up with something?

 

Blender is a close approximation for lightmapping. Compiling benchmarks should largely hold up for C#. Everything else is largely single threaded benchmarks.

 

You cant trust a guy in Birkenstock and tube socks, but Linus seems to like it

 

Thats a new level of spam bot right there

 

I dunno, I could buy a looot of vietnamese food for $750.

 

If you eat on the street you can probably feed an entire pluton But there are luxury restaurants in Vietnam too. I remebered a really nice one we visited in Hué.

 

Never been to Vietnam, but back when I visited Indonesia for school I quickly realized that even if I was eating at the nicest places I could have afforded to eat out every night with the food budget I had for the rest of the year.

 

I wanna go to the far east real bad! I adore their eastern philosophies.

 

Its good prices all over indochina for westerners, but like everywhere else you can find extreme luxury. Though same extreme luxury will offcourse be even more expensive in a western country.

I remembered we ate at The China House situated in the classic Mandarin Oriental hotel in Bangkok, it was easy to break 750 USD there

 

Jealous man

 

If you are in Bangkok, visit it! Its a really nice experience.

 

To get back on topic. 750 USD is still dirt cheap for a 16 core monster, and I cant wait for it to be available

 

Curious how anyone is going to actually USE all these cores. So much software is still single-threaded. I'd much rather see like a 10GHz CPU with 4-8 cores max.

 

Unity uses all the cores it get for baking light, oclussion culling baking, etc. You can run 16 docker containers without problem etc, etc.

 

You probably already have more than 4-8 threads running on your PC. In fact, your PC likely has dozens of processes running right now. With a 12C or 16C CPU, your system will feel more responsive simply because the threads in each of those processes can be serviced immediately instead of waiting. So even if you don't have a single app maxing out all 12-16 cores, your system still feels more responsive in general.

Beyond that, you can have a bunch of browser tabs open at the same time without your browser feeling laggy. And of course you can really see the benefit once you start up a light bake or a compile. You can build a large project while doing a bunch of other tasks (reading web pages, watching video, etc) without any of it feeling laggy.

In servers, more cores is even more exciting. For example, I really like the AMD Epyc 24C/48T CPU (7401P) for most server situations. The base clock on the 7401P is only 2Ghz and the boost is only 3GHz, but the CPU is screaming fast for doing nearly any server side workloads. It can service 48 threads without blocking and needing to wait.

 

I would really like if unity can get on the wagon though and get vulkan/gfx jobs out of preview!

 

Content creation is very heavily multi-threaded in most cases now.

System Usage (because you're never just running one app on a PC)
https://www.anandtech.com/show/15043/the-amd-ryzen-9-3950x-review-16-cores-on-7nm-with-pcie-40/4

Rendering (primarily 3d modelling but lightmapping would qualify too)
https://www.anandtech.com/show/15043/the-amd-ryzen-9-3950x-review-16-cores-on-7nm-with-pcie-40/5

Encoding (primarily video)
https://www.anandtech.com/show/15043/the-amd-ryzen-9-3950x-review-16-cores-on-7nm-with-pcie-40/6

Web Browsing (because most people don't just use a single tab/window)
https://www.anandtech.com/show/15043/the-amd-ryzen-9-3950x-review-16-cores-on-7nm-with-pcie-40/7

Steve (Gamers Nexus) and Wendell (Level 1) discussing the massive performance improvements to compiling

Spoiler

 

Also when you capture screen for trailers and devlogs its good to have many cores

 

Those numbers get me even more horny on the 3950x. So glad I skipped 3900x

 

Yeah, and also DX12 support.

 

Just pray that AMD can fix availability this time around!

 

vulkan or dx12 I dont care as long as it works multithreaded

I really think this is needed for VR to hit desktop fidelity

 

That's why I'm a mac user.

Spoiler

 

As a dude who does a lot with Unity ML Agents, I'm all about the core counts. I've been watching this for a while, and am pretty excited, considering I'm running an i7 6700k right now.

 

I havent used Unity machine learning, but every single one i have tested uses the GPU, so Unity uses the CPU for this?

 

Still don't see how hardly anyone is going to make use of 16 or 32 cores. I mean, many software apps are single threaded, the ones that are multi threaded that could really take advantage are likely to be big software like 3d rendering, video processing and the like. People saying this will be great for improving web browser tabs.. come on. Really? Browsing is a lightweight activity. What might be cooler to see for me is even faster memory and more caching, and that individual cores are much more powerful. And of course, you gotta pair this with a fast SSD otherwise why bother having a monsterous cpu and a slow hard drive.

 

I saw a huge upgrade in smooth feeling when I went from 4 cores to 8,I suspect similar feel when I double it again

edit: And this is a forum for game devs. We use the computer as workstation.

 

People said the exact same thing about four core processors back in 2006, and yet today no one recommends a system with anything less than that. Keep in mind you're not running your apps in a vacuum but are running them alongside many other apps, drivers, system services (aka daemons), etc.

Additionally most people today multi-task. If I have two applications that can easily use up to eight cores each but only have an eight core processor I can only run one of them at a time, but with a 16 core processor I can easily run both of them at the same time.

How many of these apps do we use on a regular basis though? Thinking through my day-to-day software usage I can only think of a few apps that I use on a regular basis that are explicitly single-threaded only. Little apps like Microsoft's Calculator and Snipping Tool are the only two that I can think of that I use all the time.

Yes, like I mentioned in an earlier post, productivity software. That said modern apps and games are slowly being developed to use more cores. Just because many of them don't need more than four doesn't mean it will stay like that for long.

Consoles affect things due to how many of the games coming to PC are merely ports from the consoles, and we know for a fact that the next generation of consoles will be AMD's Zen architecture with an 8C/16T (core/thread) layout.

While that's similar to the current layout of the PS4/XBone the arch that powered it (AMD's Jaguar) was poorly made and that's why you could run four cores just fine. That won't be the case with Zen. Zen is an absolute monster.

By the way there are games now that while they can run off of low-end processors (4C/8T) will not be completely smooth or have frequent micro-stutters during gameplay with anything less than eight cores. Two very easy examples of this are Battlefield 5 and Red Dead Redemption 2 which expect 8C/8T (console ports).

https://www.techspot.com/review/1754-battlefield-5-cpu-multiplayer-bench/
https://www.pcgamer.com/red-dead-redemption-2-stuttering-fix/

How heavily do you use your browser? Do you visit websites that are known to be heavier than most?

Yes, you will want a computer that isn't unevenly balanced, and that's precisely how I've been building every desktop for the past decade. Once SSDs became affordable I swapped every HDD to be long term storage with only occasional access and nothing more.

Moving large projects between HDDs has acceptable performance but for random accesses a budget SSD decimates the most expensive HDDs on the market. Easily a hundred times faster (WD Black achieves 0.5MB/sec while a budget SSD is 50MB/sec) for random access and a dozen times faster for large sequential transfers (eg a project from storage to SSD).

Thanks to a Black Friday deal I caught a week ago my next system will have 4TB NVMe SSD storage.

 

NVMe is so cheap now, I would never recommend someone to but a sata disk