AMD RyZen CPU Architecture for 2017

[lanek]

???

I think he wanted to write " budget "..

 

[xEx]

I think he wanted to write " budget "..

Yes...sorry lol

 

[Davros]

ok
so a quick question for games what cpu is more future proof ryzen or a similarly priced intel cpu
considering a ryzen core is slower but there is more of them ?

 

[xEx]

Ryzen: more threats, the difference in performance in single threat will get outpaced by future game's use of more threats.

 

[Malo]

Yes Ryzen is much more of a threat

 

[xEx]

oh shit I can't even English today...idk why

Samuel don't kill me please!

 

[Cyan]

I had those DIMMs initially. They're dual ranked.

Depending on the board, RAM support is all over the place at the moment. From the outset on my single-rank G.skill, I could do 2933 without issues. Since the latest update, I can't post past 2133 so I'm running older BIOS version.

Something similar happened to me. But in a different way. I HAD to run an older BIOS version, because I basically bricked the PC.

Started experimenting with memory speed and applied the new version of the BIOS, the 2.60 -assemblers had used the 2.50-. When you change the speed, if something is wrong the BIOS try to restart the PC five times before the last try. When it doesn't, it returns to a 2133MHz speed.

So I set it to 2666MHz and it worked really well with a single reset. 3200MHz was never working. But I said.., "let's try one last time".

The PC tried to restart continuously, then it was switched on all the time, fans running and stuff but doing nothing other than getting the fans to work, as in an infinite loop.

So after more than an hour trying, I had to perform a clear CMOS which saved the day, because it was as if the PC was bricked. I set the memory to 2666MHz and its working without a single issue.

 

[Malo]

One of the big issues with memory apparently is the voltage being applied at power up is only 1.2v instead of what's being set in the BIOS. If it gets past that initial cold boot at the set speeds, it applies the correct voltage set (eg. 1.35v). The next AGESA 1.0.0.6 is supposed to address that and board manufacturers have that code now.

 

[xEx]

https://videocardz.com/newz/amd-engineering-samples-update

top of the line 3.1/3.6; 16c/32t AMD Whitehaven.

 

[itsmydamnation]

https://videocardz.com/newz/amd-engineering-samples-update

top of the line 3.1/3.6; 16c/32t AMD Whitehaven.

3.1/3.6 could be a 95 watt TDP part!
the stilts data was ~41watt @ 3.3ghz vmin ( around 1v, 3.1ghz will be around 950mv).
That sounds awesome, hopefully they follow the 1600, 1600x 1700,1700x,1800x model with a low clocked 12 core part and that it is not much more then 1800x (say around 550usd) and scale up form their to a 16core x being around 900.

The awesome thing for AMD about there SOC model is a 12 core @ 550 is more revenue and margin then 2x 1600's @ 220 while being the same SOC, but its also cheaper for my home VM server requirements all costs considerred for me to get 1x12 core over 2x 6 cores ( case,mb,power supply, extra disk) assuming they price like how i hope.

 

[entity279]

The articles says A2 = 95W tdp; and in the part number of the 16C @ 3.1/3.6 model, A8 is written. Which would contradict your claim

 

[Alexko]

3.1/3.6 could be a 95 watt TDP part!
the stilts data was ~41watt @ 3.3ghz vmin ( around 1v, 3.1ghz will be around 950mv).
That sounds awesome, hopefully they follow the 1600, 1600x 1700,1700x,1800x model with a low clocked 12 core part and that it is not much more then 1800x (say around 550usd) and scale up form their to a 16core x being around 900.

The awesome thing for AMD about there SOC model is a 12 core @ 550 is more revenue and margin then 2x 1600's @ 220 while being the same SOC, but its also cheaper for my home VM server requirements all costs considerred for me to get 1x12 core over 2x 6 cores ( case,mb,power supply, extra disk) assuming they price like how i hope.

If it were a 95W part it should have A2 in its product code, but it features A8 instead—a code whose meaning remains unknown at this point. But the 12-core part has the TDP code A9, also unknown, so I'd guess something around 150W for 16-core SKUs and 125W for 12-core variants. We're talking about a reasonably large amount of silicon here (therefore a large die area) so cooling shouldn't be too much of an issue.

 

[Cyan]

Ryzen is performing spectacularly. If you see in futuremark, the quickest CPUs of the world, in the first place there is a CPU that costs 1500€, but there comes the surprise, in the second place is a 470€ CPU followed by the CPUs of its family. Guess which ones? :smile2:

https://www.futuremark.com/hardware/cpu

in the gpu department there aren't many surprises, but it's good to see the RX 480 placed quite high.

 

[xEx]

Ryzen is performing spectacularly. If you see in futuremark, the quickest CPUs of the world, in the first place there is a CPU that costs 1500€, but there comes the surprise, in the second place is a 470€ CPU followed by the CPUs of its family. Guess which ones? :smile2:

https://www.futuremark.com/hardware/cpu

in the gpu department there aren't many surprises, but it's good to see the RX 480 placed quite high.

3.77 times more expensive(not counting platform) for 5% more performance(or higher score I should say)

 

[pharma]

Really don't know how they get the comparison price amounts. Definitely did not pay $2265 for my 5960x.

 

[Ethatron]

Really don't know how they get the comparison price amounts. Definitely did not pay $2265 for my 5960x.

Quick, eBay, now!

 

[pharma]

Love my 24/7 4.4Ghz too much!
Edit: Just checked ebay and they have some for under $600. Wonder if the benchmark records highest price ever paid?

 

[lanek]

Dont know if allready posted.... AMD CPU roadmap leaked ... Naples 32cores for 2017, 16cores snowy owl, 48 cores 7nm due for 2018.

https://videocardz.com/69428/amd-snowy-owl-naples-starship-grey-hawk-river-hawk-great-horned-owl

I find the quality on some "slide" a little bit questionnable, so i dont know if everything are true...

 

[Deleted member 13524]

Keep in mind these slides are from February 2016.

Naples:
- 16 - 32 cores
- LGA Socket SP3
- 64MB L3 cache
- 8 channel DDR4
- 128 PCIe-lanes
- 32 lanes for SATA or NVME
- 120 - 180W
- Looks like 4 Summit Ridge / Zeppelin chips in one MCM


Snowy Owl (basically half a Naples)
- 4 - 16 cores
- BGA SP4
- 32MB L3 cache
- 64 PCIe lanes
- 16 lanes for SATA or NVME
- 35 - 100W
- The version with more than 8 cores seems to be two Summer Ridge / Zeppelin chips in one MCM



Great Horned Owl SoC (big Raven Ridge)
- 4 CPU cores
- 8MB L3 cache
- Dual-channel DDR4 up to 3200MT/s (50GB/s)
- 11 NCUs
- 16 PCIe lanes
- 4 dedicated SATA/NVMe lanes
- 12-45W on FP5 BGA
- 45-65W on Socket AM4


Banded Kestrel (small Raven Ridge)
- 2 CPU cores
- single-channel 64bit DDR4 3200MT/s or dual-channel 32bit LPDDR4
- 3 NCU
- 4-15W on FP5 BGA





I'm disappointed that the big Raven Ridge can't use LPDDR4. Using 4266MT/s chips they could reach 68GB/s which those 11 NCUs could very well make use of.

 

[hoom]

Snowy Owl (basically half a Naples)

Pin compatible with Zeppelin
But can that socket support quad channel RAM?

Interesting that Naples is supposed to be SP3 socket & 2016 (whats the ETA now & what happened?), Snowy Owl = SP4, Zeppelin = SP4r2 (I'm presuming revision2?) = AM4?

APU with quad core + 11 CUs.
No sign of any HBM