Not really, this current zen architecture have a list of don'ts that will cause a generic compiler to fail for c++ and java. Forcing you to do optimization but will cause compatibility with other systems like intel.

This post has been edited by Bonchi: Apr 26 2021, 11:49 AM

Im refering to compiler error not code generated error and precisely on SIMD crashing. Even with npm node, and python, you'll encounter some hiccups with zen architecture especially when youre doing fullstack. You go stack overflow and youd see tonnes of it and it needs some troubleshooting and a few workarounds. However most cases once the issue is solved then no problems running them on intel based env

Gigabyte bios flashback works by unplugging everything including the ram and cpu except the mobo power. Got a bunch of vid guides and the file name is important.

Using these in daily basis along with custom etl tools or coding backend APIs. copying codes and compiler across from one platform to amd will most likely have issues like crashing or cpu utilization stuck at 100% due to an infinite loop happening somewhere in the micrcode level. And you should very well know that you dont simply update the compiler lol.. especially if you have a very large project. I guess it will take some time, usually it's AMD's responsibility to fix it like last time they had issue on a random number gen.

This post has been edited by Bonchi: Apr 26 2021, 08:32 PM

4800MTS goes brrrrrrr

remember to run memtest for at least 8 passes after that which will take about 8 hours to validate the settings are properly stable.

dont wehhh... not all ram error will cause a crash but instead it will slowly corrupt system files and may suddenly brick your whole OS one day. That's why running memtest86 outside of windows is recommended. when that is stable for 2 passes then maybe can do HKEPC/memtestpro which is more punishing to the ram.

I learn the hard way last time and needed to do a full windows reinstall. PC was stable for months but suddenly just kept crashing, didnt run memtest because was a noob back then. errors reported almost immediately when I run memtest despite being "stable" for months on day to day use.

This post has been edited by Bonchi: May 16 2021, 05:38 PM

Here's one china app for memtest but i feel that it tortures the hardest. https://benchlife.info/runmemtest-4-by-dang-wang-download/

Stacked chips.... Launch prime, load CPU and L3.... gonna be a fire hazard lolol. Could potentially be their "desperate measure" against intel's finally able to get out of 14nm. Maybe alderlake will still be behind. But if intel continue to optimize their 10nm like their 14nm, things will be interesting. The competition finally begins.

From experiencing all the fatal stability and reliability problems first hand, i feel like they dont give much thought on that. Even the 5800x is already overwhelmingly hot for such low powerdraw, cant imagine how much worse it will be by insulating the hot ccx cores with a very hot cache..

I think the only cpu i would be interested is the monolithic cezanne APU.

Wont be surprised if alderlake is behind but since AMD has taken the throne, i stopped being an AMD fan they gave me the impression of being the new intel. I somehow wanna see intel succeed so we can get cheaper high performance cpus.

When intel succeeds then AMD have to release a product to overtake intel by at least 10% the very next gen as well as taking their reliability problems more seriously. Then intel would need to slash price of current gen and then do the same for their next gen by overtaking amd by 10%.

with this, we will see a very big performance increment and radical new architectures being introduced every gen. Just like back in the days when Nvidia and ATI used to trade blows every generation.

Actually the large cache is why the zen3 is doing so well. zen2 had a split 16+16 cache that's why it's much weaker from the latencies. However when you take the cache advantage away like superPI, 11900K will dapao 5950X like no eyes see because intel do have much higher IPC. Sadly intel ran out of die space to fit large cache, maybe we will see better improvement on 10nm superfin. perhaps the cache stacking by AMD is a future countermeasure incase intel really did nullify the cache size advantage AMD had by going 10nm.

actually I wouldnt call it AMD's implementation as it was made by TSMC and AMD just came and use it kinda sad that AMD is taking all the credits. Compared to intel and TSMC who has to spend way more on R&D for the fab processes as compared to chip design.

This post has been edited by Bonchi: Jun 3 2021, 10:14 PM

Link to CCD have a wider range of ratio instead of 1:1 and 1:2. The Cache solves most of the latency issues.

Go see superpi world record. Maximum vs maximum situations. Pretty big lead with much fewer cores.

That's LN2 + max power vs LN2+max power so it is still apples to apples. Speaking of stock on rare occasions that doesnt utilize cache intel does win even on fewer cores. Another occasion altho unfair to compare, AVX512... easily 70% lead doing the same calculations.

It's just facts.. AMD lead is all thanks to that insanely huge cache.

A win is a win and a loss is a loss. An F1 will trash a top fuel dragster on a circuit and a top fuel dragster will trash an F1 on a quarter mile. Ignoring cylinders, fuel amount and weight. It is what it is.

Anyways temp is subjective. Ryzens are running way hotter than the power it is drawing and has almost zero gap to TJmax meanwhile intel has a higher TJmax. 90C on AMD is not fine but it always do while 90C on intel is perfectly ok, however everyone expect to run these CPUs below 70c.

Anand's temp is recorded on a 15year old cooler to stay consistent with a core2quad. A 280mm aio is more than enough actually... and there's AVX offset.

And when comes to power, i really dont understand, are people that poor to only afford a 400w psu?.. keep comparing performance/watt. Inb4 ironically AMD consumes more power on average due to inefficient boosting behavior.

Not a fan of both, especially when I was an AMD fan before but kena backstabbed so bad by the brand to the point i drop my biaseness and just jump to intel after spending all the money upgrading my AMD rig in these bad times.

To understand you need to read up on HT link but iinm, you cant really get a direct answer on this. But loosely it's 42gbps here but i remember it's around 32 on ryzens https://www.techjunkies.nl/2019/10/15/__trashed/

This post has been edited by Bonchi: Jun 5 2021, 06:32 PM

Nop, doesnt work like that. Regardless it will still be the achilles heel. It might be running on IF 1:2 with DDR5 which intel will most likely be running the IMC too.. that's why you see the gear2 setting become available.

42 is threadripper quad channel. So ryzen 3rd gen in the link is 32. Theoretically the bus can support way more but there are bottlenecks.