LinkBack URL
About LinkBacksRead more.New AM4 Picasso APUs have modest frequency boosts and move from 14 to 12nm.
Read more.New AM4 Picasso APUs have modest frequency boosts and move from 14 to 12nm.
One the one hand - it's nice to see an update.
On the other, it's been a long time coming, and it's fairly minor. However there is a rumour of a 3600G to come later this year with Navi and Zen 2.
Still, the higher boost clocks and slightly improved IPC (3%) aren't to be sniffed at.
If they can also give a 100MHz boost on the GPU that would help too. And maybe pricing at $89/$139 ...
As sykobee says, a minor update, but always appreciated.
These parts are really important, though. When building general purpose PCs for sensible prices for the bulk "non gamer" audience, integrated graphics are very important. Even on Intel's higher end parts (i5, i7), the integrated graphics are a real boon for most business customers.
So, it is good to see AMD improving these parts, as they represent some of AMD's most desireable "maintream" parts.
This is made all the more important because both AMD and Nvidia have neglected low-end video cards for years. Almost all sub 70ukp parts are rebadges of 10 year old (roughly) designs - some are even end-of-life for drivers!
Indeed, I can't help feeling there's a Hexus article in there somewhere. AMD/Nvidia assume that low end card users will now be using integrated graphics, but none of the high-core-count CPUs offer that. A business user must either pay for 3D gaming features they may not want, or fall back to cards so anaemic that they damage the overall performance even in 2D tasks.
Indeed, this is a problem I have to grit my teeth with every day - I need to run DX7 era purely-2D software at high resolutions (3840x1440 or 4k). This requires a modern GPU with displayport or HDMI 2.0 for 60fps, with moderate bandwidth. Intel/AMD APUs manage this no problem, but finding a low end GPU with suitable connections and fast enough bus is a pain - I've ended up with a 4GB AMD 480 which is massive overkill.
Erm, Picasso isn't a shrink of Raven Ridge, ISTR RR was Zen whereas Picasso is Zen+. So it is a derivative, warmed over product but not a simple shrink. Not that I think optical shrinking (literally photo reducing the die masks) has worked for a really really long time, instead you have to re-make masks to the new design rules.To recap, Picasso is an optical die shrink of existing Raven Ridge APU parts from 14nm to 12nm
I had to look it up and even them i may have got confused, but WikiChips says it is, although reading over what you've said i think i maybe misunderstand.
Meh, we know the apu already from laptops. What remains unknown is mere details.
The elephant in the room is, we DONT "know" the new desktop am4 mobo being concurrently released.
Lets get real - this is the basic level - 4 core with IGP (the "sub-dgpu segment"?).
For an appropriate modern balance of the two in the huge business & entry level segment/pricepoint, Intel have no competitive answer except past utter dominance of of this conservative segment.
AMD dont have to do much to raw perf. Their real enemy is credibility (intel have helped heaps by having no product anyway). They win hands down already on paper.
What may well make the new APU devastatingly competitive for this segment, is a single very plausible improvement - a much faster link from the system to the X570 chipset on the new am4 platforms.
I am unconvinced this faster link is exclusive to zen2.
It seems more a modification to Fabric on the Zeppelin die to link in faster native Fabric protocol to the new chipset over the same four physical lanes.
The x570 uses the much greater bandwidth to provide 2 new nvme ports (totaling an extra 8GB/s alone), and all the pcie 2 lanes on the former am4 chipset, now become pcie 3 perf.
I don't think that the laptop apuS did not expand io, disproves my theory. The benefits have little benefit to laptops, and dont commercially justify a new chipset/platform.
It may take a while to sink in, but it leaves intel's competitive position in the dust.
RR is a semitone above vanilla zen, it's got precision boost 2 and a much better memory controllerOriginally Posted by DanceswithUnix
If AMD wants higher speed storage, they don't need to touch the chipset. Just give the APUs the full complement of PCIe lanes (which the socket has pins for), and then the motherboard can route the extra 8 lanes to two x4 NVME drives
I'm disappointed that AMD didn't include an APU in the first round of 7nm parts. The improved power efficiency of 7nm could have made them a real contender in the laptop market, perhaps even giving them enough performance at low power to offering something to compete with Intel's Y series parts. What we're getting this year won't be enough to make AMD more than a minor player in laptops, and won't get them into ultraportables or tablets at all.
Not an easy upgrade. The APUs have fewer PCIe lanes because they only have one CCX. In the non-APU Ryzen parts, half the lanes are connected to one CCX and the other half connect to the other one. AMD would have to redesign the CCX itself to include additional lanes, rather than just cutting one CCX out of the die design and plopping down a GPU in the vacated space.
Are you sure? All the block diagrams I can find show the PCIe lanes the other side of the infinity fabric to the CCXs, and even official AMD slides don't show PCIe hardware in the CCX
There are currently 1 users browsing this thread. (0 members and 1 guests)
Posting Permissions
Reply With Quote