AMD - Zen chitchat

[CAT-THE-FIFTH]

I've realised that doesn't make sense. We've seen the pictures of the package, and it doesn't have an eDRAM chip on it. So it can't be an eDRAM SKU.

The whole point of the eDRAM was that it was too expensive to fab that much cache (64MB/128MB) as part of the CPU. But yeah, no DRAM chip on the package kind of knocks that thought on the head...

Well we know the first ones had it as another chip,but Intel is not really giving out much info about its newer CPUs,so I wonder if they newer parts might integrate it in the actual chips,so its not off package?? There is no reason why they can't use a smaller quantity as it does improve CPU performance too,and they can drop CPU clockspeeds.

OTH,it could just be the software getting confused at identifying a new GPU.

[CAT-THE-FIFTH]

So Ryzen gaming is the 65W TDP mobile version of the Ryzen APU:

https://videocardz.com/74464/amd-pre...ga-11-graphics

The 65W TDP version seems pointless,IMHO OFC, for a laptop unless they have gotten around the bandwidth limitations!!

Edit!!

[scaryjim]

Gotta agree that 65W TDP in a laptop doesn't make a right lot of sense. Perhaps they'll ratify up to 3200MHz SO-DIMMs? If some of the extra TDP went into the memory clocks that'd make more sense. I just can't see the difference between 35W and 65W versions being that big...

Looks like your call on the Ryzen 3 U being 4C/4T + Vega 6 is probably spot on too, CAT! Wonder when we'll see that in shipping devices...

There's a lot of stuff that doesn't quite make sense with the Ryzen Mobile launches. I wonder if they'll have a completely separate segment for the 2C die - a Y-series maybe? Just to confuse matters more...

[kalniel]

I don't know.. a 7th gen i7 HQ + discrete GPU like 1050 has got to be around 65W at least too.

[CAT-THE-FIFTH]

Thats the thing,looking and the Ryzen 2500U,the CPU actually seems perfectly fine at lower TDPs,so it seems like it must be the GPU clockspeeds and the memory clockspeeds which push up the rating to 65W TDP. It seems that the 2400G has all 704 shaders enabled,so it makes sense that the memory speed is higher. So it makes me wonder if AMD is trying to push Nvidia out of the lower end of the GPU market,since I assume it should be cheaper and simpler to use a single 65W TDP SOC,rather than a CPU,GPU and southbridge combination.

However,Ryzen 3 looks more of a winner - looking at the performance of the 2500U in gaming,it looks pretty decent,so if AMD can make sure Ryzen 3 laptops can run in dual channel,its going to be a pretty good entry level CPU,and it also means AMD can bundle a mobile RX460 or RX560 with it too,if required.

[scaryjim]

I don't know.. a 7th gen i7 HQ + discrete GPU like 1050 has got to be around 65W at least too.

It does, but with only 704 shaders Ryzen Gaming isn't going to get anywhere near GTX 1050 performance. I'd guess they'll get it above a GTX 1030, and no doubt CPU performance will increase quite a bit, but I'd personally hope they could beat a 1030 @ 35W, given the performance they can get out of the 15W U versions... unless they need the extra TDP for the memory clocks. If the difference between the 35W version and the 65W versions was 2666MHz memory v 3200MHz memory then I could see it making a reasonable difference...

[CAT-THE-FIFTH]

BTW,41 games tested on the Ryzen 5 2500U,and that includes many new titles.

https://www.youtube.com/watch?v=MdgMv_9A0z8

00:01 - Destiny 2 03:12 - Titanfall 2 05:32 - Rise of the Tomb Raider 08:09 - Microsoft Flight Simulator X Steam Edition (1080p) 10:51 - Project CARS 12:39 - The Elder Scrolls V: Skyrim 14:56 - Micro Machines World Series 16:56 - Redout 18:53 - The Crew 20:53 - Rainbow Six Siege 22:33 - Crysis 3 25:21 - Total War: Warhammer II 26:21 - Overwatch 29:56 - Battlefield 1 Multiplayer 64 32:40 - Fortnite 35:28 - Middle-earth: Shadow Of War 38:30 - Rocket League 40:55 - Dota 2 44:13 - Counter-strike: Global Offensive (CS:GO) 46:10 - Far Cry Primal

https://www.youtube.com/watch?v=IqGnYR_bNIg

00:01 - Wolfenstein II 02:46 - Pro Evolution Soccer 2018 05:36 - Mad Max 08:43 - Metal Gear Solid V 14:39 - TrackMania Turbo 15:49 - CoD: Infinite Warfare 18:23 - Overwatch 20:23 - Assasins Creed: Origins 22:33 - Far Cry Primal

https://www.youtube.com/watch?v=l9MYwQbR5bc

00:01 - Watch Dogs 2 02:12 - Deus Ex: Mankind Divided (DirectX 11) 04:13 - X-Plane 11 08:23 - The Surge 10:28 - Team Fortress 2 12:09 - Sniper: Ghost Warrior 3 13:40 - Prey 16:09 - Mortal Kombat XL 19:18 - Killer Instinct 21:04 - Fallout 4 24:25 - The Witcher 3 32:29 - Ashes of the Singularity: Escalation (DirectX 12)

Also,I had another thought about the Ryzen Gaming SKUs - maybe they are for a NUC-like device??

[CAT-THE-FIFTH]

Going back to the leak a few days ago:

https://i.redd.it/c7fix5jkqy201.jpg



Look at when the Ryzen Gaming SKU is meant to launch - Q2 2018.

So,it does make me wonder why is there such a delay. With Ryzen 3 Mobile I can understand as they need to ramp up production and it is probably using CPUs which can't quite make the Ryzen 5 and Ryzen 7 grade.

Does,it mean to get the clockspeeds up,Ryzen Gaming SKUs need special binning(meaning the desktop ones might not clock that high,which again does not make sense as the U series can run upto 3.5GHZ),or is it since they might be bundling some form of faster onboard memory with it - does the Vega IGP on Raven Ridge support HBM2??

[scaryjim]

I can't see them sticking HBM on a 704 shader IGP, or including an extra memory controller on the Raven Ridge die - it's a niche product, it's not particularly high performance, and it'd be a lot of expense for the HBM, and a lot of silicon area for a memory controller that would only be used in a small minority of cases.

I can see two possibilities:

a) binning for the best yielding/clocking IGPs - don't forget we don't currently have a Vega 11 part on the market, so perhaps they're waiting for yields to improve so they can use all 11 CUs at high clock speeds

b) Ryzen gaming will be a 12nm part. It launches after 2nd gen Ryzen on desktop, which is also meant to be 12nm, and the 12nm process is meant to be more performance oriented...

[Xlucine]

I'm distrustful of the claims about the ryzen gaming chips. For one, why are the numbers lower than the U parts? Consumers are gonna want the 2500 over the 2400. Secondly:

Ryzen 7 and ryzen 5 gaming, no ryzen 3 gaming

[CAT-THE-FIFTH]

Looks like the first entries for the 400 series chipsets have been spotted:

https://videocardz.com/74528/amd-pro...ted-at-pci-sig

[CAT-THE-FIFTH]

I saw this posted on AT forums:

http://forum.ixbt.com/post.cgi?id=at...88:14192:1.jpg

[scaryjim]

I saw this posted on AT forums: ...

Radeon Vega 3 Mobile?

Ryzen 3 Mobile uses the 2C/3CU die?

Anyone know if that CPU score would make sense for 2C/4T?

[scaryjim]

... Anyone know if that CPU score would make sense for 2C/4T?

OK, I did the sensible thing and went looking for comparison scores on passmark. The Ryzen 5 2500U scores around 8k marks, so that 5439 looks a little high ... but of course the case clock there is a fair bit higher than on the Ryzen 5, so it's not utterly ridiculous - it all depends on the actual clock speeds the processor runs the workload at. A 4C/4T set up might make more sense of the score though. I'd tack that one down as "unclear".

Either way, it's a pretty good score for a (probably) 15W chip - the i5 5200U only scores ~ 3500. The 35W Bristol Ridge desktop A10 9700E can only do a shade over 5100 points.

[kompukare]

First review of the Ryzen 5 2500U equipped Acer Swift 3 SF315 from NBC
https://www.notebookcheck.com/Test-A....275900.0.html
(Currently only in German but I'm sure the NBC translators will rectify that soon.)
Looks quite good. Obviously some thermal throttling in (for example) continuous Cinebench runs, and ilde power seems a bit poor.
Gaming wise the Vega 8 keeps up with MX940 but not MX150 and beats Intel HD620 by a huge amount.

[CAT-THE-FIFTH]

Some new Ryzen laptops have also been announced from Acer and Dell:

https://www.anandtech.com/show/12233...-vega-on-7nm/2