AMD Zen 2 (Refresh) 3900XT/3800XT/3600XT

[jigger]

I would be interesting if AMD allows for higher TDP.

 

[Grim5]

Wow. If they can do those base clocks at the same TDP, then they must be highly binned.

I'll believe it when I see it.

It would also be a mistake not to set the 3600XT base clock to the same as the 3800XT.

not necessarily binned it's rumoured to be a different node. It's not the same 7nm that other zen 2 uses but actually 7nm Enhanced (same as Zen 3 uses)

 

[Quartz]

Seems a bit odd having a gpu naming convention attached to CPU's, "xt".

I disagree. Consistency in naming conventions makes things easier for consumers.

 

[Gerard]

I disagree. Consistency in naming conventions makes things easier for consumers.

It's the first time they've done this for a cpu, nothing really consistent about it. They only relatively recently brought back "xt" for gpu's though internally they apparently always have that name attached. If it's an ongoing thing fair enough but i feel it might well be a one off due to these being released mainly to take the shine (or flames) off comet lake.

 

[random_matt]

Kind of annoying, only recently bought mine.

 

[Minstadave]

Wow. If they can do those base clocks at the same TDP, then they must be highly binned.

I'll believe it when I see it.

It would also be a mistake not to set the 3600XT base clock to the same as the 3800XT.

It's not binning, it's a new process, N7P.

 

[GregI]

Is there any news on these being the same as the 3300x? i.e. all cores on one chiplet?

 

[Minstadave]

Is there any news on these being the same as the 3300x? i.e. all cores on one chiplet?

The 3600XT and 3800XT are likely to be single chiplet.

The 3900XT can't be.

The 3300X is single CCX. None of these can be.

 

[Kirby Wurm]

It would also be a mistake not to set the 3600XT base clock to the same as the 3800XT.

I guess that's in aid of segmentation if true.

 

[muon]

I guess that's in aid of segmentation if true.

It's already segmented by number of cores.

Otherwise they are just shooting themselves in the foot for comparisons against the 10600K.

 

[Kirby Wurm]

It's already segmented by number of cores.

Otherwise they are just shooting themselves in the foot for comparisons against the 10600K.

Yes but a 3600XT with the same clocks as a 3800XT would perform very similarly in many benchmarks (many games for example) so perhaps they don't want cheaper parts cannibalising sales?

 

[0h_no_not_again]

Apparently these chips are doing 2000MHz FCLK according to 1usmus.

Link

That's an interesting tidbit if combined with the XT variants having a higher TDP.

The key is that the external interface will be a sizeable consumer of power. Wonder if there has been a tweak to IO silicon/fuses to allow higher clocks....

Wild speculation!

 

[alec]

2000 FCLK could noticeably reduce latencies.
Sub 60ms ram on Zen2?

 

[Grim5]

2000 FCLK could noticeably reduce latencies.
Sub 60ms ram on Zen2?

Definetly sub 60ns

 

[NinjaCool]

With this FCLK boost combined with an extra 200-300MHz, I can easily see a 5% performance increase over the stock 3000 models

 

[humbug]

Leaked benchmark:

For reference this is the highest scoring Ryzen in the R20 thead....

11, Score 547: AMD Ryzen R9 3950X at 4.7Ghz, jordand77

https://www.hardwaretimes.com/amd-r...-core-i9-10900-10700k-3900xt-tops-the-10900k/

 

[alec]

Suspiciously repeatable.
I mean 3600XT and 3950X SHOULD boost to same clock and probably score around the same, but exactly 531 on the nose?
Probably spreadsheet rather than real benchmarks.

 

[Troezar]

Suspiciously repeatable.
I mean 3600XT and 3950X SHOULD boost to same clock and probably score around the same, but exactly 531 on the nose?
Probably spreadsheet rather than real benchmarks.

All say 4.7GHz, maybe that is why?

 

[humbug]

Suspiciously repeatable.
I mean 3600XT and 3950X SHOULD boost to same clock and probably score around the same, but exactly 531 on the nose?
Probably spreadsheet rather than real benchmarks.

Yeah that is Suspicious, Cinebench doesn't work like that, there is slight score variation from one run to the next.

 

[Fyter]

Yeah that is Suspicious, Cinebench doesn't work like that, there is slight score variation from one run to the next.

Is it possible that he/she ran the tests multiple times to get an average? I get variation is single cores in r20 aswell, anywhere betwen 393 to 419 on 5820k @4.5ghz.