• Competitor rules

    Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.

AMD Zen 2 (Ryzen 3000) - *** NO COMPETITOR HINTING ***

4K8KW10 said:
Do we expect 40-50% (15% IPC + 25% frequency) performance uplift going from the 2700X to a potential 8-core 3700X?! :eek:
Click to expand...

As I said previously, going to an 8c chip elimates the ccx penalty on ram, also if they uncouple the memory from IF it shouod make memory support more robust.

Add in the front end fixes, branch prediction fixes and other uArch changes and it could be quite special.

Zen+ is already near Intel's peak performance at lower speed, once you eliminate the speed deficit, and increase the IPC and fix zens weak points (memory, latency, front end) etc it should more than take the battle to Intel

I'm hesitant to say it's going to trounce the 9900k, as I'm trying to temper my own expectations, but I think this might be a bit of a landmark moment in CPU history. This could potentially be the moment where AMD start really smashing into Intel, I won't be surprised if Intel face a really tough next 5 years
 
He mentions something interesting. That the AM4 has the space to put 2 chiplets plus a quartered IO and bring 16 cores to mainstream.

 
You can see that in the picture I made/posted in the EPYC thread, the total area covered is the same size as a single 8 core Naples die with 33% space left over.
 
Well that's my mind made up, I'll be switching to Zen 2 when I replace my current i5 4690K rig,Whilst I've no doubt Intel will still retain the single core performance crown, both in terms of IPC & clockspeed, The difference with these new Zen 2 CPU's will now be close enough as to no longer matter in the stuff I use my PC for.

The only question is whether i go and build a new Zen rig now with a decent X370 mobo & fast 3200 DDR4, but only put a cheap 1500X or 1600 in it, just to tide me over until Zen 2 launches, and then drop in a 3700X, or just wait for the launch next year.

Hmmm, decisions,decisions...
 
Cooper said:
Well that's my mind made up, I'll be switching to Zen 2 when I replace my current i5 4690K rig,Whilst I've no doubt Intel will still retain the single core performance crown, both in terms of IPC & clockspeed, The difference with these new Zen 2 CPU's will now be close enough as to no longer matter in the stuff I use my PC for.

The only question is whether i go and build a new Zen rig now with a decent X370 mobo & fast 3200 DDR4, but only put a cheap 1500X or 1600 in it, just to tide me over until Zen 2 launches, and then drop in a 3700X, or just wait for the launch next year.

Hmmm, decisions,decisions...
Click to expand...

If I didnt have Ryzen already I would be waiting and making sure I had the funds to get the "best" of the new motherboards. :)
 
chroniclard said:
If I didnt have Ryzen already I would be waiting and making sure I had the funds to get the "best" of the new motherboards. :)
Click to expand...
Definitely, AMD have commited to the socket, but there could be another chipset and there is a chance of pcie 4.0 on mainstream.
 
I would imagine AMD will release X570 with Zen2 in mainstream and with PCIE4 onboard the motherboards, its already confirmed Epyc is running with PCIE4 on chip and on the motherboards and 7nm Vega is PCIE4.

Expect to see PCIE 4 GPU's potentially towards the end of next year, and expect to see Zen2 motherboards with PCIE on them, i have a CH6 and a 1700, this is going to my boys to use, i will definitely be going to the 3700X or whatever it will be called with an X570 motherboard etc, and then later on a 7nm AMD GPU.
 
I was torn between waiting or not with x470 and a 2600x just to tide me over till 3700x or 3800x but figure I might as well wait now given pcie4 and a better board.
 
If you haven't bought now and think you will be PCI-E lane constrained on AM4, then waiting is probably best if you don't want an HEDT system. Obviously for those people just building a gaming system with a single graphics card and one PCI-E based SSD then there is no real benefit with PCI-E 4.0, but if you are wanting to run a card 'like' the 2080Ti, it will let you drop it down to 8x PCI-E 4.0 lanes without suffering any loss in performance, and freeing up those 8 lanes for other uses.
 
Journey said:
If you haven't bought now and think you will be PCI-E lane constrained on AM4, then waiting is probably best if you don't want an HEDT system. Obviously for those people just building a gaming system with a single graphics card and one PCI-E based SSD then there is no real benefit with PCI-E 4.0, but if you are wanting to run a card 'like' the 2080Ti, it will let you drop it down to 8x PCI-E 4.0 lanes without suffering any loss in performance, and freeing up those 8 lanes for other uses.
Click to expand...
Even if you have proper pcie4 gpu at full pcie speed, you will be able to get full speeds of your multiple SSDs as well.
 
muziqaz said:
Even if you have proper pcie4 gpu at full pcie speed, you will be able to get full speeds of your multiple SSDs as well.
Click to expand...

I was referring to lane constraint, rather than bandwidth constraint. Although you'll find that the newer PCI-E 4.0 NVMe SSD's that will be arriving in the next 12 months will actually require the equivalent of 6 lanes of PCI-E 3.0 lanes to work at full speed. NAND layer density is rapidly increasing and alongside that so is the performance 96-layer 3D NAND modules are performing 25-40% faster than the 48/72 layer packages currently in use, and 128 layer by the end of 2019 a 4x PCI-E 4.0 U.2/M.2 slot would be required to achieve the full bandwidth of the device.

So unless AMD increase the 20-lanes available on AM4 (+4 at the chipset) you would still be limited to one 16x and one 4x device at full speed connected directly to the CPU, with other peripherals attached to the chipset lanes. 4.0 would however allow the addition of things like 10GbE more easily into the chipset lanes which may help the lane deficit, while not sacrificing performance elsewhere.
 
Journey said:
I was referring to lane constraint, rather than bandwidth constraint. Although you'll find that the newer PCI-E 4.0 NVMe SSD's that will be arriving in the next 12 months will actually require the equivalent of 6 lanes of PCI-E 3.0 lanes to work at full speed. NAND layer density is rapidly increasing and alongside that so is the performance 96-layer 3D NAND modules are performing 25-40% faster than the 48/72 layer packages currently in use, and 128 layer by the end of 2019 a 4x PCI-E 4.0 U.2/M.2 slot would be required to achieve the full bandwidth of the device.

So unless AMD increase the 20-lanes available on AM4 (+4 at the chipset) you would still be limited to one 16x and one 4x device at full speed connected directly to the CPU, with other peripherals attached to the chipset lanes. 4.0 would however allow the addition of things like 10GbE more easily into the chipset lanes which may help the lane deficit, while not sacrificing performance elsewhere.
Click to expand...

Yes, more of the bandwidth release for current NVMe SSDs. I hope AMD will increase lane count as well as SSDs are getting cheaper and cheaper, and people are now running multiple devices on single system.
Also the problem I see with speedier future SSDs is Windows not taking any significant advantage of them, same goes with majority of games.
 
Jamin280672 said:
each chiplet can still be 2 x 4 core CCX's stuck together, just smaller and faster because of the die shrink ? who and where does it say its 1 CCX that they've changed to handle 8 cores 16 threads.
Click to expand...
For Zen1 AMD needed CCX size to be four cores to fit same design also to APUs.
8 core CCX as standard would have needed making separate quad core CCX design for APUs.
Something AMD didn't have resources to do.

Now 7nm would likely allow putting that 8 cores and GPU into one reasonable size die.
And increasing size of CCX to eight cores would make it lot easier to realize performance potential in desktop CPU, with same fast low latency communication between all eight cores.
Though who knows if they're going to do APUs with chiplet/MCM design.


Distracted said:
Look at that rock on her finger!
Click to expand...
Well, she has certainly deserved her salary.
By turning AMD around from company with uncertain future, to one with certainly very good years ahead and solid roadmap and development goals.

GPU side might have suffered from Zen/Ryzen drawing most resources.
But even having the best GPUs wouldn't have turned AMD's course, with lots of sheeples buying Nvidia no matter the situation.
Also any future console deals certainly needed good modern performance CPU.
 
You must log in or register to reply here.
Back
Top Bottom