• Competitor rules

    Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.

AMD demonstrates Ryzen 9 5900X prototype with 3D V-Cache stack chiplet design

pete910 said:
Not long after "engineer" did! I mean every true engineer must scowl when somebody says "Software Engineer" :cry:
Click to expand...

Some of us software people are (were ?) real engineers :p, doing the stuff that tends to lead to news headlines and dead people when it goes wrong :eek:. Even got all the bits of paper, chartered etc. Problem is, it doesn't pay. Gave up eventually and now just hack code, its easier, has little in the way of consequences and doubled my salary overnight. Meh :(

...but keeping to the original subject, it will be interesting to see if a trend develops by AMD to keep producing, in effect, one off halo products to keep the top one place in the game performance charts. What was that slogan from racing? win on Sunday, sell on Monday
 
softwaresimian said:
...but keeping to the original subject, it will be interesting to see if a trend develops by AMD to keep producing, in effect, one off halo products to keep the top one place in the game performance charts. What was that slogan from racing? win on Sunday, sell on Monday
Click to expand...
Isn't that what the 12900 is, to an extent though? And the 11900 before it? An excessively binned chip pushed far beyond its nominal performance envelope with ridiculous power levels, so that it can claim to be the best?

AMD at least are doing it via an interesting technical means - not just cranking it harder.

Or, actually, maybe this is something actually interesting. Think of it in terms of market segmentation - AMD tried to push the 5800x as the gaming chip on release. The 3D stacking cache is a technical variation in chip design that does legitimately provide differences in performance.

I can't remember the last time you couldn't get better gaming performance by buying further up the stack. There were always (small) gains available. Now... there is a reason to stop at 5800x3D.

From AMD's PoV
-AMD make more money on premium parts
-AMD sell less chiplets to people chasing 12/16 cores for marginal improvements
-AMD have more chiplets left, to sell to everyone else.

Helps with most of their problems.
 
I watched the Robert Hallock interview with PC World and 5800X3D sounded much more impressive, up to 40% faster in gaming than a 5800X in the absolute best case. If the price is good then I might upgrade my 3900X to this as I now almost exclusively game on this PC.

My only concern is downgrading from 12 cores to 8, Hallock said that the 5800X3D sees these huge gains in current and old games due to the way they are designed (they are not very multithreaded and rely on many random accesses), but what about in a few years' time, will the extra cores become much more important?
 
ltron said:
I watched the Robert Hallock interview with PC World and 5800X3D sounded much more impressive, up to 40% faster in gaming than a 5800X in the absolute best case. If the price is good then I might upgrade my 3900X to this as I now almost exclusively game on this PC.

My only concern is downgrading from 12 cores to 8, Hallock said that the 5800X3D sees these huge gains in current and old games due to the way they are designed (they are not very multithreaded and rely on many random accesses), but what about in a few years' time, will the extra cores become much more important?
Click to expand...
I wouldn't worry about that, as long as 8 cores are enough for games (consoles are 8 cores so it is standard) even if it utilizes 12 cores, cache will give you more performance, and games can't be 100% parallelized, only some parts so you will not have 100% gain for every extra core.
 
I guess all those who bought 5950X and 5900X for gaming have been doing it all wrong and should have been disabling 4 cores of each CCX for a total of 8 cores and 64mb of cache.
 
Joxeon said:
How is my suggestion complaining?
Click to expand...
Tonality is a difficult thing to portray on the internet; given your stance on AMD's product stack and direction as of late, it came across as really sarcastic.
Joxeon said:
Would be interesting if someone could test this.
Click to expand...
Now, if you'd led with that then it wouldn't sound sarcastic :p and to answer the question, I imagine a half-switched off 5950X would dip under a 5800X in gaming performance. Yes there's more cache involved, but there's also more latency involved with data requests jumping across the IO die. The cache being two lots of 32MiB rather than a single pool of 64MiB is a prime example of IO jumpy skippy data fetchy.
 
Last edited:
LePhuronn said:
Tonality is a difficult thing to portray on the internet; given your stance on AMD's product stack and direction as of late, it came across as really sarcastic.

Now, if you'd led with that then it wouldn't sound sarcastic :p and to answer the question, I imagine a half-switched off 5950X would dip under a 5800X in gaming performance. Yes there's more cache involved, but there's also more latency involved with data requests jumping across the IO die. The cache being two lots of 32MiB rather than a single pool of 64MiB is a prime example of IO jumpy skippy data fetchy.
Click to expand...

Having switched off half cores before to do just as such it didn't change the performance of any of the game I played myself. With that the 5950x I have doesn't boost particularly well and Humbugs 5800x for most gaming situations holds up better (apart from like Star Citizen and Stellaris) just don't let him know that ;)
 
Curlyriff said:
Having switched off half cores before to do just as such it didn't change the performance of any of the game I played myself. With that the 5950x I have doesn't boost particularly well and Humbugs 5800x for most gaming situations holds up better (apart from like Star Citizen and Stellaris) just don't let him know that ;)
Click to expand...
Did you try switching off half the cores from each CCX though or did just shut off one of the CCXs?
 
Joxeon said:
Did you try switching off half the cores from each CCD though or did just shut off one of the CCDs?
Click to expand...

No half from each. It was last year when we just wanted to compare performance figures. It made no difference to all 16 on or 8 on with 4 on each but it was still loosing out to the 5800x for most part. The 5950x wouldn't boost past about 4.6Ghz, sometimes 4.7Ghz though for me regardless of how many cores where as I know humbug has had his 5800x boosting to 5Ghz St times.

Where it was noticed is mid/late game Stellaris or Star Citizen which both ran better with all 16 cores enabled or all cores on one CCX. Oh and late game total war and civilisation. They were fine until then but system chugged more late game as it was completing one form the AI turns.
 
CPPC along with the core scheduler in the OS should pick the best cores from CCD1 to use for workloads that only require a certain amount of cores, ie games. Most games will use 4-8 cores at most, so there is no need to start disabling cores to improve performance as for the most part it won't.

However, if you are using a 5950X/5900x, you may actually see a slightly higher average clock frequency on your CCD1 cores with CCD2 disabled, and this might allow some extra performance due to the higher clock frequency for the prioritised cores. The extra clock frequency comes from the slightly reduced power usage from having half the cores disabled.

If the game is using 6+ cores though with the 5900x, this might hurt performance more than help it as it has less cores to work with than the 5950X. Not tested it myself, just a theory and SMT might mean that its okay.

What I have tested though is this:

Here is my 5950X with 16 cores/32 threads
sOiNMMu.png

Here is my 5950X with CCD2 disabled, so running 8 cores/16 threads.
jSGzAyW.png

Note the difference in average frequency observed over time.

That was taken using COD Black Ops, 4K max settings on a 6900 XT. Other games may see different behaviour, so this may not apply to everything.
 
You must log in or register to reply here.
Back
Top Bottom