AMD Ryzen 5 Begins Shipping

[wizard69]

Linux performance of the Ryzen 1600 seems very good.
https://www.computerbase.de/2017-04/...endungen_linux
https://www.servethehome.com/amd-ryz...like-this-one/

HardOCP is reporting that tensions arose between AMD and hardware review sites again.
Those sites who received review kits from AMD had to stay silent until today, while others could just buy the CPUs themselves a week earlier without signing anything.


Phenom II allowed core unlocking.

That may be true but it really isnt in AMDs best interest to allow unlocking of defective cores.

Also, AMD only sells Ryzen in symmetric configuration, this means if two cores on one CCX are defective, it will become a 2+2 quadcore, even though it could in principle still work as 2+4 hexacore if the rest of the cores are perfectly good.

So how do you tell which cores are defective? From my persoective it is still gambling. Beyond that AMD is selling a six core variant. The problem with enabling adefective core is that you dont know what the defect is, you could end up with a bug that hits once in a blue moon.

I just dont see the attraction in messing with what could be a defective core simply to save a few bucks.

[schmidtbag]

If in fact unused cores are on the die they are often disabled by blowing fuses or laser cutting traces. AMD cant afford to have their reputation trashed by idiots reenabling defective cores. Frankly im not even sure why people would even want to try this. And yes i know that at times perfectly good core will be disabled to meet demand but then you are gambling.

AMD's reputation was never tarnished last time they permitted this. If anything it helped their reputation, because it was like playing scratch tickets with CPUs - sometimes you'd only earn back the money you spent on the ticket, but every once in a while you'd get something great. People knew they were taking a gamble with this feature, and they were prepared for the idea that if it didn't work out the way they wanted, they wouldn't (or shouldn't have) been left disappointed. As far as I'm aware, most of these disabled cores were just simply predicted to be faulty, but were otherwise perfectly usable. Keep in mind the enabled cores were specific to the BIOS they were configured on; it's not like you could re-sell the CPU advertised and priced at it's higher-core model.

Like I said, AMD likely prevents this core unlocking because it hurts their income, which they desperately need to address. Unfortunately for them, it seems the vast majority of their CPUs, regardless of price point, are limited to roughly 4GHz. In other words, there's no reason to buy any X-series model since the cheaper models can pretty much reach the same speeds. However, we'll obviously see how true that holds up as BIOS and microcode updates continue to be released.

[chithanh]

I don't know any instance where unlocking reflected negatively on AMD's reputation. And there is a long history of allowing unlocking, though more examples exist on the graphics side.

Radeon 9500 could be unlocked to 9700
early Radeon HD 6950 could be unlocked to 6970
Athlon II and Phenom II X2 could be unlocked to X3 (and X4 sometimes)
early Radeon RX480 4 GB reference models could be unlocked to 8 GB.
Radeon RX460 disabled shaders could be unlocked.

If an unlocked product would not work properly, nobody blamed AMD for it.

[artivision]

HMP based Rysen with 2 cores near 3.5Ghz, plus ULP based Vega core near 3 Teraflops, plus at least 8GB HMB = 45Wp laptop core. Preferably universal GPU like, without the need for motherboard and cheap as 150 bucks.

[L_A_G]

Too bad it's two compute modules (or whatever they are called) with two cores disabled per module rather than one compute module with four cores. This will mean some additional latency moving data between the modules.

This is only really an issue when scheduling isn't aware of this. When it is, it's not that a big of a deal avoiding this hazard. Also, being spread across more silicon there's obviously the thermal advantage that allows for better manual overclocking and the dynamic clocking to work better.

[artivision]

Lets just stay with the G4560 for now.

[duby229]

This is only really an issue when scheduling isn't aware of this. When it is, it's not that a big of a deal avoiding this hazard. Also, being spread across more silicon there's obviously the thermal advantage that allows for better manual overclocking and the dynamic clocking to work better.

Actually, it seems inverse, but the more die space the harder it is to keeps clocks in sync.

[msroadkill612]

Actually, it seems inverse, but the more die space the harder it is to keeps clocks in sync.

This is an interesting debate u guys are having. If u r right sir, could this impact on ryzens lowish memory clock issues?

ryzen products reviewed so far have all been dual ccxS - no one has tested a single ccx zen cpu. maybe it does have latency advantages?

its esp pertinent, as u would think? the zen/vega apu will be one cpu ccx and 1 vega gpu glued together like a ryzen dual ccx.

if so, the cpu in the apu will be as u describe, not like the ryzens so far tested.

[Anvil]

will get one of these when they get into Australian shores , im due for a New PC

[suberimakuri]

Single ccx might be interesting as said above, but otherwise I think not worth the money.
I'm keen to see any laptop reviews. I will look at some zen/vega type setup for next laptop later this year (Asus Zenbook if happens?).