AMD Zen FX - information, benchmarks, experience of users

[KiSUAN]

So, this kinda confirms 1500X will be 2 CCX`s, look at the cache size.

Yes, and I can tell you that image is from AMD

[chew*]

But should not have normaly the not so good samples more?

No by nature the Top end chips are usually the highest LN2 clockers. Deneb was sort of an exception 955..

Thuban was high bin = high ln2 chip

Bulldozer and newer revisions the best cold ( high leak ) were in the higher bins.

We are seeing same with 1800x....

[AlleyViper]

So, this kinda confirms 1500X will be 2 CCX`s, look at the cache size.

Wierd thing, 4C/8T 1500X 2+2 and 4C/8T 1400 4+0/0+4.

[chew*]

or failed cache.

[AlleyViper]

Oh, that's right. Forgot it could still be a 2+2 with half L3 per CCX. That would remove the probability of a situation where a cheaper 4+0 would be faster than a 2+2 upper model at same clock, even with less L3.

[StAndrew]

I don't like RyZen 1.0. I'll wait for 2.0.

[Hell Hound]

I don't like RyZen 1.0. I'll wait for 2.0.

OverWatch

1100T @ 1440p



I don't see a bottleneck w/1100T and 1080



I don't believe I will bottleneck vega @ multi monitor
+ I can vsr to 7680x2160 and destroy older titles

[Hell Hound]

1800X @ ? W/GTX 1080(1080p)

[Apokalipse]

3600 already been done according to rumour.

What matters isn't speed. Its performance at x speed timings with equal cpu speed.

I would bet money on 14-14-14 3200 vs those timings/speed even with the nb bumped by 200.

Thats probably Juan testing at ASUS. They dont necessarily give him uber kits to test with. Maybe high spped stuff but not in the tightest latency bin

Mind you, the "infinity fabric" + L3 cache speed is tied to RAM frequency. So increased memory frequency doesn't only increase memory bandwidth.
It probably helps with inter-CCX latency.

I wish it was able to be clocked independently though.

[cdawall]

Mind you, the "infinity fabric" + L3 cache speed is tied to RAM frequency. So increased memory frequency doesn't only increase memory bandwidth.
It probably helps with inter-CCX latency.

I wish it was able to be clocked independently though.

I think there was rumor of that. I doubt it will be for many many bios's down the road though.

[AlleyViper]

On this Anand's article:

We have confirmation from AMD that there are no silly games going to be played with Ryzen 5. The six-core parts will be a strict 3+3 combination, while the four-core parts will use 2+2. This will be true across all CPUs, ensuring a consistent performance throughout.

[chew*]

On this Anand's article:

I popped that question in pre briefing. My only concern was 4 core parts.

[BenchZowner]

2+2 for the 4c8t parts... shooting themselves in the foot ?

[vario]

2+2 for the 4c8t parts... shooting themselves in the foot ?

Thats the thing, you DO get big ass L3 cache in that instance, 8 full megs for any 2 cores, thats a lot.
BUT if scheduler isnt aware of the whole CCX issue, it can go to waste and in some cases even end up slower :-/ .

[AlleyViper]

I wouldn't be surprised if an Athlon model came later with only one CCX on die, 4 cores and probably even less L3 (or none for Sempron).

[vario]

I wouldn't be surprised if an Athlon model came later with only one CCX on die, 4 cores and probably even less L3 (or none for Sempron).

Athlon models will be probably made from raven ridge with disabled GPU.

[chew*]

2+2 for the 4c8t parts... shooting themselves in the foot ?

While i tend to agree and i think 0+4 would probably perform better.....i must remain impartial so i had to ask if at least to point it out in a review.

My intention however was not to question if "silly games" were being played and just to be able to note what configuration i was reviewing.

[AlleyViper]

Oh, good news for people running old Thermalright coolers like me to soften the upgrade to AM4. There's a new TRUE Universal BTK compatible with the ancient TRUEs, Vx, etc for AM4. Shouldn't end with the same mount pressure as the old ones, but it least it works.

Edit: It might be a mess-up on TR's part. I'll email them to confirm, because it seems to be the previous retaining system that only supports the AM3 holes.

[FlanK3r]

Few speculations: Maybe we will see something better for ultra megatasking than Ryzen 7. I read some ifnormation about monster up to 16C/32T for desktop workstations.

[cdawall]

Few speculations: Maybe we will see something better for ultra megatasking than Ryzen 7. I read some ifnormation about monster up to 16C/32T for desktop workstations.

The clockspeed looks great. It beats out my xeon easily.

[FlanK3r]

Oh yes, its coming ...

https://videocardz.com/newz/amd-engi...samples-update

[chew*]

I do not see that being very usefull for the majority.

Its a server chip its clocked lower and if current imc/memory issues are any indicator...quad channel is going to be a mess for early adopters.

Its not even the same package so it can not drop into a x370. Good thing with current x370 boards thermals.

Gamers need higher clocks not more cores at this point....but the crunchers and work station guys will benefit greatly.

Platform would be lovely for current cpu's but...not realistic. The gains with pci x etc etc are due to double stacked R7 cores. Thats why quad channel.. Etc etc.

Ryzen 2 literally probably means Ryzen *2

Respin will be more exciting than re glueing.

[FlanK3r]

Threadripper is almost here! HEDT AMD, summer 2017!

[BenchZowner]

Now bring us the real-life applications ( except rendering, which is already there on core utilization ) and games to use the cores...

[ZenEffect]

Now bring us the real-life applications ( except rendering, which is already there on core utilization ) and games to use the cores...

The civilization series always scaled with cores since civ v