Ryzen 5 2400G & Ryzen 3 2200G APU (with Vega GPUs) review kits are being sent out

[GoldenLag]

8 minutes ago, Thermosman said:

I wonder how many vega stream processors they could squeeze onto at 8 core am4 cpu

Dont believe that is possible since they remove one of the 4 core clusters to fit the vega compute units. Though it would be interesting. If there is an r7-2x00G it is probably only a compute unit increase and clock speed increase

[Bananasplit_00]

Put one of these in a mining rig and you got another GPU to utilise

[Sakkura]

1 hour ago, Agost said:

DDR4-2933 as an official ram clock? Is it legit?

Makes sense, APUs are usually bottlenecked by memory bandwidth. They've usually come with higher official memory speed support than other CPUs.

[Dabombinable]

2 minutes ago, Sakkura said:

Makes sense, APUs are usually bottlenecked by memory bandwidth. They've usually come with higher official memory speed support than other CPUs.

Lets just hope that the IMC is as good as that of standalone Ryzen. Lets just say that with previous CMT based APU-it wasn't unusual to see DDR3 1600 being dropped down to DDR3 800 with first-gen-ULV-i5 performance.

[Taf the Ghost]

22 minutes ago, Thermosman said:

I wonder how many vega stream processors they could squeeze onto at 8 core am4 cpu

Currently, the RX Vega parts are a bit larger than the entire Ryzen die, so the answer, currently, is very few.

 

However, life gets REALLY interesting at 7nm. And I also just realized why AMD would put Vega on 7nm as an automatic. A RX480/580 Die is small amount larger than a Ryzen Die. 7nm would allow you to put a full 8c Zen1 die and a RX580 on the same package at roughly the exact same size as a current part.  Now, that'd be horribly, horribly bandwidth starved, so the Zen2 APUs, due late 2019 or early 2020 will probably be 8c and ~16 to 24 CUs. Probably 16 CUs and save some die space, maybe clock in around 180 mm².

[The Benjamins]

20 minutes ago, GoldenLag said:

Dont believe that is possible since they remove one of the 4 core clusters to fit the vega compute units. Though it would be interesting. If there is an r7-2x00G it is probably only a compute unit increase and clock speed increase

The issue is what Die are they using, the Zeppelin Die is 8c16t no GPU, and the APU die is 4c8t 11CU.

To make a Ryzen 7 G part they would need to have another Die design or go MCM and using the Zeppelin die with VegaM

[Thermosman]

2 minutes ago, The Benjamins said:

The issue is what Die are they using, the Zeppelin Die is 8c16t no GPU, and the APU die is 4c8t 11CU.

To make a Ryzen 7 G part they would need to have another Die design or go MCM and using the Zeppelin die with VegaM

Oh, I didnt know that they had a only 4 core die

[GoldenLag]

1 minute ago, Thermosman said:

Oh, I didnt know that they had a only 4 core die

Feels a but empty if they left out the R7 in this linup. Not that it matters much but still. The linup doesnt feel completed. 

[GoldenLag]

I am kinda wishing to see a mac mini of sorts using this Processor. Or any other mini system for that matter. Cooling may be an issue in that formfactor, but dialing down the clocks a bit and it should maybe work

[The Benjamins]

8 minutes ago, GoldenLag said:

Feels a but empty if they left out the R7 in this linup. Not that it matters much but still. The linup doesnt feel completed. 

well hopefully with the success of Ryzen, they will make more dies or options for zen2 in 2019 or zen3 in 2020+

[GoldenLag]

1 minute ago, The Benjamins said:

well hopefully with the success of Ryzen, they will make more dies or options for zen2 in 2019 or zen3 in 2020+

Maybe at that point use Navi to combo up with Zen3 or Zen2. Or even 7nm vega since it is quite a nice chip at lower clocks. 

 

Imagine scalable navi with scalable zen in a TR4 socket formfactor. (Nah thats just whishful thinking)

[TVwazhere]

1 hour ago, Noyu said:

 

MY BAD UPON MORE GOOGLE-FU YOU WERE RIGHT!!

 

Zen (14nm) = Ryzen 1 and the Ryzen APUs  (2200G and 2400G)

Zen+ (12nm) = Ryzen 2

Zen 2 (7nm)

 

 

See, I really still think they should have done:

 

Zen (14nm) = Ryzen 1 and the Ryzen APUs  (2200G, 1300x, 2400G 1500x, 1600, 1600x, 1700, 1700x, 1800x)

Zen+ (12nm) = Ryzen 2 (2250G, 1350X, 2450G, 1550x, 1650, 1650X, 1750, 1750x, 1850x)

Zen 2 (7nm) = True 2 series (2600, 2700, etc.)

 

Right now it looks like Zen 2 will be "Ryzen 3" which will be kinda confusing

 

6 minutes ago, GoldenLag said:

Feels a but empty if they left out the R7 in this linup. Not that it matters much but still. The linup doesnt feel completed. 

As @Energycore pointed out in a status, what's REALLY missing is a Ryzen 5 2600G, which would be a six core 12 thread APU

[GoldenLag]

6 minutes ago, TVwazhere said:

As @Energycore pointed out in a status, what's REALLY missing is a Ryzen 5 2600G, which would be a six core 12 thread APU

Well at that point they might aswell make a 8 core 16 thread Ryzen 7 2700G/2800G. Since you need 2 four core cluster tl make that doable, abd at this point there is only 1 four core cluster

[cj09beira]

 

1 hour ago, mariushm said:

It's more like they're limited by the amount of HBM memory production, and the fact that they made contracts with Apple (their Apple Pro machines) and probably Intel to give them priority to the chips with HBM memory, so there's less quantity available for regular gaming cards (the Vega cards)

And even if they'd have enough quantity, they'd be snatched quickly to be used for mining since they're great for mining Monero or whatnot.

 

and also part is going to ssgs 

33 minutes ago, Taf the Ghost said:

Currently, the RX Vega parts are a bit larger than the entire Ryzen die, so the answer, currently, is very few.

 

However, life gets REALLY interesting at 7nm. And I also just realized why AMD would put Vega on 7nm as an automatic. A RX480/580 Die is small amount larger than a Ryzen Die. 7nm would allow you to put a full 8c Zen1 die and a RX580 on the same package at roughly the exact same size as a current part.  Now, that'd be horribly, horribly bandwidth starved, so the Zen2 APUs, due late 2019 or early 2020 will probably be 8c and ~16 to 24 CUs. Probably 16 CUs and save some die space, maybe clock in around 180 mm².

if hbm2 production increases enough....it would be a dream,

add to that functioning vega features and 7nm clocks it would work very well,

[Noyu]

34 minutes ago, TVwazhere said:

As @Energycore pointed out in a status, what's REALLY missing is a Ryzen 5 2600G, which would be a six core 12 thread APU

Damn if AMD hints towards this I'll hit pause on the 2400G and get that instead ahaha.

 

To be honest I just want an APU that can play like a 1050/Ti.

But I doubt that'll happen within the year.

 

[Trixanity]

37 minutes ago, GoldenLag said:

Feels a but empty if they left out the R7 in this linup. Not that it matters much but still. The linup doesnt feel completed. 

Zen+ fills out the lineup. The APUs slot in between CPUs. The only chance of an R7 APU is a heavily binned one but it doesn't really make much sense.

48 minutes ago, Dabombinable said:

Lets just hope that the IMC is as good as that of standalone Ryzen. Lets just say that with previous CMT based APU-it wasn't unusual to see DDR3 1600 being dropped down to DDR3 800 with first-gen-ULV-i5 performance.

It should be better than Ryzen.

23 minutes ago, TVwazhere said:

See, I really still think they should have done:

 

Zen (14nm) = Ryzen 1 and the Ryzen APUs  (2200G, 1300x, 2400G 1500x, 1600, 1600x, 1700, 1700x, 1800x)

Zen+ (12nm) = Ryzen 2 (2250G, 1350X, 2450G, 1550x, 1650, 1650X, 1750, 1750x, 1850x)

Zen 2 (7nm) = True 2 series (2600, 2700, etc.)

 

Right now it looks like Zen 2 will be "Ryzen 3" which will be kinda confusing

 

As @Energycore pointed out in a status, what's REALLY missing is a Ryzen 5 2600G, which would be a six core 12 thread APU

Why would imaginary products to fit a preconceived notion be better?

The APUs aren't on the same schedule as CPUs so there are additional changes found only on the next generation CPUs. That's why Raven Ridge and Pinnacle Ridge exist in the same product line besides the 2018 release year.

Zen microarchitecture doesn't follow the Ryzen marketing (engineering and marketing don't piss in each other's work space). They're separate and mostly irrelevant for the average consumer. It's only confusing if you conflate the two. Zen3 might end up being Ryzen 5000 series and APUs 6000 for all we know so better get used to it now 

 

There is nothing missing. The die tops out at 4c8t and 11 CU. They'd need an entirely new die to go beyond that. That'd take a while and many millions to do that for a single product which means not going to happen. Also, even if it was 12nm based it'd increase die size significantly.

 

So that means Zen2 APUs on 7nm at the earliest. So you'll be waiting 2 years presumably.

[TVwazhere]

2 minutes ago, Trixanity said:

Why would imaginary products to fit a preconceived notion be better?

The APUs aren't on the same schedule as CPUs so there are additional changes found only on the next generation CPUs. That's why Raven Ridge and Pinnacle Ridge exist in the same product line besides the 2018 release year.

Zen microarchitecture doesn't follow the Ryzen marketing (engineering and marketing don't piss in each other's work space). They're separate and mostly irrelevant for the average consumer. It's only confusing if you conflate the two. Zen3 might end up being Ryzen 5000 series and APUs 6000 for all we know so better get used to it now 

There's a reason Intel denoted 8th gen CPU's with an 8 at the beginning of the SKUs... It's because it's easier for the general consumer to understand. And I guarantee you people will mix Zen 2 with Ryzen+ SKU's...

 

When the Nvidia GTX 1080 launched, there were people who thought it was specifically for 1080p gaming. 

[TVwazhere]

Also (Only an unboxing, but still)

 

 

[DocSwag]

It sounds like the IMC is a fair bit better than the original ryzen cpus and also without the IF between CCXes I think it may be possible that these cpus could compete better with Covfefe lake i3s and maybe even i5s than other ryzen cpus.

 

Of course, only time will tell but I think it's totally possible.

[DocSwag]

1 hour ago, Dabombinable said:

Lets just hope that the IMC is as good as that of standalone Ryzen. Lets just say that with previous CMT based APU-it wasn't unusual to see DDR3 1600 being dropped down to DDR3 800 with first-gen-ULV-i5 performance.

I don't think that's what you want . Compared to intel the IMC in ryzen isn't that good, so I really hope they improved it.

[The Benjamins]

15 minutes ago, TVwazhere said:

Also (Only an unboxing, but still)

 

 

We did get 1 peice of new information, The 2200G and 2400G have x8 PCIe for "graphics cards" vs all the others x16

https://overclock3d.net/reviews/cpu_mainboard/amd_ryzen_3_2200g_and_ryzen_5_2400g_apu_preview/1

[GoldenLag]

7 minutes ago, The Benjamins said:

We did get 1 peice of new information, The 2200G and 2400G have x8 PCIe for "graphics cards" vs all the others x16

https://overclock3d.net/reviews/cpu_mainboard/amd_ryzen_3_2200g_and_ryzen_5_2400g_apu_preview/1

The x8 bandwith isnt challanged yet in games(and a lot of other tasks). So i presume for an average user it is not an major issue. I can see why there is only 8 leftover as opposed to 16. Since they might be dedicating 8 to the iGPU

 

Edit: it doesnt make me super happy to see there is only 8x pci left to gpu...... But it is negilable

Edited February 8, 2018 by GoldenLag

[The Benjamins]

2 minutes ago, GoldenLag said:

The x8 bandwith isnt challanged yet in games(and a lot of other tasks). So i presume for an average user it is not an major issue. I can see why there is only 8 leftover as opposed to 16. Since they might be dedicating 8 to the iGPU

 

Edit: it doesnt make me super happy to see there is only 8x pci left to gpu...... But it is negilable

Ya I think it has X16 total which will be x4 for chipset, x4 for M.2 and x8 for extra PCI device (aka GPU).

 

I don't see a reason to need more then x8, for iGPU or dGPU.

[TVwazhere]

2 minutes ago, GoldenLag said:

The x8 bandwith isnt challanged yet in games(and a lot of other tasks). So i presume for an average user it is not an major issue. I can see why there is only 8 leftover as opposed to 16. Since they might be dedicating 8 to the iGPU

If you disabled the iGPU, would that allow a full x16 connection in the PCI-E slot? Kinda a mute point since GPU's run perfectly fine on x8 connections, this is purely for science and something we will have to wait on reviews for.

[GoldenLag]

3 minutes ago, The Benjamins said:

Ya I think it has X16 total which will be x4 for chipset, x4 for M.2 and x8 for extra PCI device (aka GPU).

 

I don't see a reason to need more then x8, for iGPU or dGPU.

Unless you are using Titan V for deep learning and other shenanigans, but you must be a fool to pair a 2400G and a Titan V