AMD 880K and DDR4??

[Justinbaileyman]

I thought AMD was going to be releasing the 880K for the Fm2+ Socket? if so when is it coming out? If not how close are we to getting a APU refresh with DDR4?Some one has got to know something so please spill the beans..!!

 

[VirtualLarry]

I don't think FM2+ can support DDR4.

 

[ShintaiDK]

You will see new APUs with the AM4 socket and DDR4 next year.

Note the socket got renamed from FM3 to AM4.

The layout for APUs are this:

 

[Justinbaileyman]

Ok thanks... now does anyone have any news on the 880K? Is it being released for the Fm2+ socket still or not?

 

[dark zero]

Forget FM2+. Is an awful moment from AMD.... When they lost their L3 cache, they became irrelevant.

And they will go to AM4 on full DDR4

 

[Shehriazad]

Forget FM2+. Is an awful moment from AMD.... When they lost their L3 cache, they became irrelevant.

And they will go to AM4 on full DDR4

Well hey now...860K is somewhat of an upgrade to the FX4350 even with the lack of L3. (And an 880K would be even more of an upgrade if neither one is being OCd...but even OC'd the 800Ks would be faster than or worst case equal to the FX4XXX)

And we don't even have to start talking about the iGPU.


Granted it's still a terrible platform in the long run due to the lack of more cores (6Core Kaveri or Godavari would've been interesting) or a future upgrade path.

 

[DrMrLordX]

The 880k has been completely MIA, just like the 870k and the 7890k. If AMD is really launching AM4 in March to accommodate Bristol Ridge, they may just cancel all those Kaveri chips.

 

[jpiniero]

The 880k has been completely MIA, just like the 870k and the 7890k. If AMD is really launching AM4 in March to accommodate Bristol Ridge, they may just cancel all those Kaveri chips.

They are probably unable to get the clock speed high enough on 28 nm compared to 32 nm.

Maybe Bristol Ridge is coming sooner because DDR4 pricing has gotten closer to DDR3 faster than they would have expected.

 

[DrMrLordX]

Maybe. I'm interested in seeing if they can deliver AM4 this early. If so, getting boards onto the market would be a great idea IF they have the punch to handle Zen later.

I am not so sure they are having problems with Kaveri clockspeeds like that, since the default voltages on Kaveri refresh are astronomically high. Nearly any sample can hit 4.3 GHz under those conditions. Certainly they should be able to bin for that.

 

[The Stilt]

As said, there is no DDR4 controller in Kaveri / Godavari, but a DDR3 / GDDR5 hybrid controller instead. Carrizo meanwhile has a hybrid DDR3 / DDR4 controller, however it can be considered as useless. DDR4 on Carrizo would only reduce the performance, since DDR4 has higher latencies and Carrizo cannot support higher than 2400MHz memory clock regardless if DDR3 or DDR4 is used.

 

[hojnikb]

iGPUs like bandwidth more than latency, so ddr4 is more than welcome there.

 

[The Stilt]

The question remains, what bandwidth (DDR3 & DDR4 both limited to 2400MHz)?

 

[SPBHM]

The question remains, what bandwidth (DDR3 & DDR4 both limited to 2400MHz)?

GPUs tend to not suffer much with slightly higher latency

and DDR3 vs DDR4 2400 seem to perform the same on Skylake

http://images.hardwarecanucks.com/image/mac/reviews/Intel/SkylakeDDR3versusDDR4/2.jpg

http://images.hardwarecanucks.com/image/mac/reviews/Intel/SkylakeDDR3versusDDR4/10.jpg

http://www.hardwarecanucks.com/foru...-intel-skylake-ddr3-vs-ddr4-comparison-4.html

 

[AtenRa]

Carrizo's DDR-4 is perfect for the Mobile 15-35w TDP SKUs due to lower voltage and heat than DDR-3 at 2133-2400MHz. But for Desktop as Stilt said its worse for the CPU than DDR-3 due to higher latency (at the same speed, 2400MHz for example).

I hope Bristol Ridge DDR-4 controller will be able to work at more than 3000MHz.

 

[The Stilt]

I hope Bristol Ridge DDR-4 controller will be able to work at more than 3000MHz.

"Bristol Ridge" is the same exact thing basically as Carrizo. Just the package(s) and configuration differs. For that reason it will be limited to max 2400MHz too.

 

[TeknoBug]

AM4 and if not at least FM3(?) with DDR4. An APU with DDR4 would benefit for the iGPU, but hybrid crossfire with low end AMD GPU's would need to catch up to ship with DDR4 rather than DDR3 onboard memory.

 

[Shehriazad]

AM4 and if not at least FM3(?) with DDR4. An APU with DDR4 would benefit for the iGPU, but hybrid crossfire with low end AMD GPU's would need to catch up to ship with DDR4 rather than DDR3 onboard memory.

FM3 is not a thing.


AM4 will have APUs and high end CPUs(whatever that means in Zens' case). They want to unite everything on a single socket. I guess that streamlines the designing process and makes everything a bit cheaper.

So let's assume the first product on AM4 is an XV APU with an upgraded iGPU....the only card that would probably match up in performance rating would probably be the 250X...which does not yet have a 300 series counterpart.

I'd say 1GB/2GB 2(3)50X cards wouldn't be held back terribly on the memory side if the XV APU gets to work with 3000 Mhz. I know that current XV mobile chips only support 2400 Mhz...but we've got to look at the limitations of the tdp and HDL...those won't exist on the Desktop version.



Also...sorry for being so darn off topic ^^;

 

[jpiniero]

I'd say 1GB/2GB 2(3)50X cards wouldn't be held back terribly on the memory side if the XV APU gets to work with 3000 Mhz. I know that current XV mobile chips only support 2400 Mhz...but we've got to look at the limitations of the tdp and HDL...those won't exist on the Desktop version.

I suppose 3000 is possible if you overclock it. Although, DC 3000 is only 48 GB/sec, whereas the 250X is 72 GB/sec.

 

[Shehriazad]

I suppose 3000 is possible if you overclock it. Although, DC 3000 is only 48 GB/sec, whereas the 250X is 72 GB/sec.

But I think we have to agree that for low end cards....Vram speed is not exactly the limiting/bottlenecking factor(especially when the cards have GDDR5)... ^^;


Kaveri had clear gains up to 2400 and minor gains up to like 2800ish. Let's say XV ends up just 15% faster (no idea, total guess) on the iGPU side then 3000 would be perfectly fine.







Back to topic...I still want to get my hands on a 870k or 880k...availability for either one is...well...not exactly a thing.

 

[DrMrLordX]

"Bristol Ridge" is the same exact thing basically as Carrizo. Just the package(s) and configuration differs. For that reason it will be limited to max 2400MHz too.

That DDR4 clockspeed limitation is a crying shame. Zen better have a better memory controller by 2017 when the new APUs come out.

I suppose 3000 is possible if you overclock it. Although, DC 3000 is only 48 GB/sec, whereas the 250X is 72 GB/sec.

Assuming Carrizo/Bristol Ridge have the same stability issues as Kaveri with elevated bclk, you'll want to use the 2133 setting along with a bclk of 141 to get there. Few if any Kaveri systems can sustain a bclk that high. The iGPU tends to freak out in certain bclk ranges, and then you have problems with the PCIe and SATA controllers tracking off bclk as well. With proper bus locks, it would be doable, but AMD has shown no initiative in implementing proper bus locks, external clock generators, and suchlike.

If Bristol Ridge got the same treatment that k-class Skylake CPUs got from the Haifa team, then we'd be in business.

Kaveri had clear gains up to 2400 and minor gains up to like 2800ish. Let's say XV ends up just 15% faster (no idea, total guess) on the iGPU side then 3000 would be perfectly fine.

You are forgetting that the CPU and iGPU share memory bandwidth. It is a much bigger bottleneck for an iGPU than it is for a low-end dGPU.

Back to topic...I still want to get my hands on a 870k or 880k...availability for either one is...well...not exactly a thing.

Most people do. I have to think that AMD has probably scrapped plans for FM2+ updates, opting instead to roll out AM4 in March. There may be equivalent XV processors for AM4, which would be interesting.