AMD RyZen CPU Architecture for 2017

I'm sure I can, and I'm sure I can also find tests where the difference isn't around 10%, you shouldn't take any single benchmark, no matter who did it, as gospel
Im not taking words as gospel but there is a ton of info and test that show there is a difference, where it is or whats the cause is not yet clear but the numbers say it is.
 
Original post was deleted:

http://webcache.googleusercontent.c...-four-core-processors-available-april-11.html

In a conference call yesterday AMD announced more Ryzen processors. As you guys know AMD made a promise to release the Ryzen 5 series in Q2 this year. Well, they have been able to meet that time slot and as such we can shed some light on Ryzen 5 processors that are to be released April 11th.

Ryzen 5 will be a handful of 6-core processors as well as 4-core processors, priced attractively I must say. Both the 6 and 4 core models all are SMT thus you may double up the cores into threads 4:8 and 6:12.

AMD will release four new models:

Ryzen 5 1600X
This is a six-core part with 12-threads. The base frequency will be 3.6 GHz with a nice 4.0 Ghz boost frequency. Add to that an XFR range of roughly 100 MHz. This processor will cost 249 USD.

Ryzen 5 1600
This is a six-core part with 12-threads. The base frequency will be 3.2 GHz with a nice 3.6 GHz boost aka turbo frequency. Add to that an XFR range of roughly 50 MHz. This processor will cost 219 USD.

Ryzen 5 1500X
This is a four-core part with 8-threads. The base frequency will be 3.5 GHz with a 3.7 GHz boost aka turbo frequency. Add to that an XFR range of roughly 100 MHz. This processor will cost 189 USD.

Ryzen 5 1400
The 1400 is again a four-core part with 8-threads. The base frequency will be 3.2 GHz with a 3.4 GHz boost aka turbo frequency. Add to that an XFR range of roughly 50 MHz. This processor will cost 169 USD.

As stated before these processors are priced competitive alright. The Flagship units will have a 95 Watt TDP, the lower SKUs settle for 65 Watts. These value are indicative that the processors are all 8-core models with a CCX or two cores disabled. AMD is going to make SKUs available with a stock coolers, these will be the Wraith Stealth for the Ryzen 5 1400 and the Wraith Spire for the 1600 and 1600X. These stock coolers do not have LED lighting enabled.

The four Ryzen 5 processors will be available starting April 11th. Ryzen series 3 you will see in the 2nd Half of 2017.
 
Yeah but almost 20% is crazy. I think Ryzen is the paradise for ram makers.

Don't know about that, but it shows the inter-CCX fabric in under-provisioned with bandwidth. Latency would be roughly similar for both memory speeds.

AMD probably linked the fabric to the DDR/2 clock because the individual CCXs can be running at different speeds (the L3 runs at the fastest core speed). Maybe they intended to run at DDR clockspeeds, but had to halve it to support faster RAM The situation is pretty bad, when you add multiranked DIMMs or multiple DIMMS for increased RAM capacity, memory operating frequence falls as does CPU performance.

IMO, they need to make the whole un-core an independent clock domain and ensure the inter-CCX fabric has enough bandwidth to support sustained performance. Either double the link widths, or double the number of links; The Infinity fabric has a mesh topology which degenerates to a single link when there are just two nodes, maybe use two instead ?

Cheers
 
Regarding memory speeds, AMD will release new microcode in May which will allow higher memory clocks (multipliers)
 
A Chinese website has roadmaps containing Pinnacle Ridge and Raven Ridge.

X3buZW5.png


TntLlXz.png


There's no sign of the rumored DDR4 + HBM2 Raven Ridge.
 
If true, the number of cores for Pinnacle Ridge and the lack of a new APU in 2018 are disappointing.
You wanted more than 8 cores for the next version of Ryzen 7, just a year after it's released?
 
So no AM4 Raven Ridge this year? ie laptop only?

I would expect the packaging part to be pretty minor, and for a quite large addressable market..
But maybe Raven RIdge BGA is pretty close to 2018 anyway, I just had an idea of Q3
 
You wanted more than 8 cores for the next version of Ryzen 7, just a year after it's released?

Yes. The 1800X enjoys a comfortable lead over the i7-7700K in non-gaming workloads, but not a huge one. Intel is about to introduce 6-core processors with Coffee Lake and I would expect those to close that gap, and perhaps even overcome Ryzen. There will probably be Skylake-E SKUs with 8, 10, and perhaps 12 cores as well.

So I think AMD needs 12 cores to maintain Ryzen's edge, which shouldn't be difficult on a 14nm process that is already pretty mature now, let alone next year, especially when Ryzen is apparently something like 215mm². The other option would be for Pinnacle Ridge to have at least 10% higher IPC compared to Summit Ridge, slightly higher clock speeds, and fewer issues with its interconnect and memory subsystems. This is possible, but adding more cores would be easier, and not incompatible with any of this.

Alternatively, AMD could make relatively affordable—$1000?—versions of Naples with 16 cores, essentially offering something comparable to Intel's HEDT platform, but with much more raw power.

On a somewhat unrelated note, the R7 1700 seems to be really, really power efficient:

getgraphimg.php

http://www.hardware.fr/articles/939-2/consommation-efficacite-energetique.html

I think the last time AMD was clearly more power-efficient than Intel at anything might have been in 2006…
 
couldn't they just increase single core performance and out pace intel that way ?

Well they use one test (x264 ) performance, i can imagine, it can allready change from one test from another ( i have try watch on some other reviews, they completely differ on test choice and so results too. ).. ( And ofc, CPU binn, system used can get a bit part on it. ( when 12V is mesured )

Actually, i think outside some internal change, IPC increase can come mostly from "clock speed" increase.
 
Last edited:
Yes. The 1800X enjoys a comfortable lead over the i7-7700K in non-gaming workloads, but not a huge one. Intel is about to introduce 6-core processors with Coffee Lake and I would expect those to close that gap, and perhaps even overcome Ryzen. There will probably be Skylake-E SKUs with 8, 10, and perhaps 12 cores as well.

So I think AMD needs 12 cores to maintain Ryzen's edge, which shouldn't be difficult on a 14nm process that is already pretty mature now, let alone next year, especially when Ryzen is apparently something like 215mm². The other option would be for Pinnacle Ridge to have at least 10% higher IPC compared to Summit Ridge, slightly higher clock speeds, and fewer issues with its interconnect and memory subsystems. This is possible, but adding more cores would be easier, and not incompatible with any of this.

Alternatively, AMD could make relatively affordable—$1000?—versions of Naples with 16 cores, essentially offering something comparable to Intel's HEDT platform, but with much more raw power.

On a somewhat unrelated note, the R7 1700 seems to be really, really power efficient:


http://www.hardware.fr/articles/939-2/consommation-efficacite-energetique.html

I think the last time AMD was clearly more power-efficient than Intel at anything might have been in 2006…

AMD will fix ryzens bottlenecks and increase performance before adding more core which I think its the right move. Burn resources into adding core without fixing the flaw in the design would be a bad move. Get the design better first, then scale it.
 
Instant system crash when running sequence of FMA3 instructions

One of my internal benchmark applications is insta-hard-freezing on Ryzen.
Ryzen 7 1800X
Asus Prime B350M-A (BIOS 0502)
4 x 8GB Corsair CMK32GX4M4A2400C14 @ 2133 MHz
Nothing is overclocked. Everything is stock.
Windows 10 Anniversary Update

When I run the Haswell binary from here: https://github.com/Mysticial/Flops/t...naries-windows

The entire system usually freezes when it gets to: Single-Precision - 128-bit FMA3 - Fused Multiply Add

Sometimes, it will make it past that, but it usually ends up crashing/freezing later on in the test anyway.
 
Back
Top