FEATURED AMD ZEN Discussion (Previous Rumor Thread)

[mackerel]

That's old news M_M and I'm not sure the "quad" is accurate or not. I watched the demo and it was never mentioned aside from them running the same speed ram. I assumed they had the 6900k in dual.

If this is referring to the demo in the New Horizons event, they used dual channel on the Intel system according to the small print they put up at the end which described the configuration. I posted a screenshot of it somewhere earlier in this or similar thread.

Also in my and others testing of the Ryzen Blender demo, it didn't seem much affected by ram bandwidth anyway.


While here, http://pc.watch.impress.co.jp/docs/column/kaigai/1043349.html includes the slides presented at ISSCC. I've not gong through them in detail yet, but it reconfirms my earlier reading they are not attempting to match Intel's FPU potential I always had a feeling AMD would rather you handle FP intensive tasks on GPU instead, which isn't always appropriate.

 

[mackerel]

if you jump just past 29 minutes or so, they discuss AMD saying they did not focus on FPU performance and some reasoning around it. This might contribute a fair amount into the die size claims.

 

[Johan45]

I've been gleaning bits and pieces around the net and wondering if that "X" in the naming stands for AMD's eXtended Frequency Range or XFR. I'm sure that's something they could disable/enable if they wanted to. So the higher "binned" chips at their price will go into higher priced builds read better cooling and overclock higher by themselves?

 

[ATMINSIDE]

My impression was that XFR would be on all CPU's of the Ryzen lineup.

 

[Dolk]

@mackerel

Why does a desktop CPU need a anything past 256bit calculations? As of right now, most FPU calculations for the desktop is ported over to the iGPU or GPU. The internal FPU is for quicker low latency calculations. For graphics, you really need a GPU to do anything these days, or some kind of high speed FPGA. CPUs are only good for single threaded integer and memory instructions, so let them do that. FPU was only brought in because latency was too high between CPU and GPU, but those times have changed.

Intel continues to increase FPU on desktop, because they will use that same core for Workstation/Server with a couple added parts. AMD decided to design two different cores for Server, Desktop this generation.

 

[Johan45]

My impression was that XFR would be on all CPU's of the Ryzen lineup.

That was my first impression as well. Just trying to make sense of their naming.
Funny I just tried 2 of the pricing links and the info is gone. Looks like AMD must have had a problem with the leak. They've been really tight this time and I was surprised to see those links pop up. Too late now though isn't it with the cat already out of the bag

 

[mackerel]

@Mackerel

Why does a desktop CPU need a anything past 256bit calculations? As of right now, most FPU calculations for the desktop is ported over to the iGPU or GPU. The internal FPU is for quicker low latency calculations. For graphics, you really need a GPU to do anything these days, or some kind of high speed FPGA. CPUs are only good for single threaded integer and memory instructions, so let them do that. FPU was only brought in because latency was too high between CPU and GPU, but those times have changed.

Intel continues to increase FPU on desktop, because they will use that same core for Workstation/Server with a couple added parts. AMD decided to design two different cores for Server, Desktop this generation.

I'm not sure I can answer that question, as you'd need to ask programmers what they use and why. I would clarify one area. The actions don't actually work on a single value of 256 bits, but multiple 64 bits. All else being equal, AMD going for half the width means they will do about half the work per clock. There is probably some extra overhead in simulating a 256-bit instruction by repeated 128-bit operations which doesn't help. For my interest areas, 64-bit floating point is king. Consumer GPUs have long been going down a road of crippled 64-bit performance as they focus on 32-bit for mainstream uses.

 

[M@rk]

Dont know if anyone already posted this info, But Here are the prices.
http://www.guru3d.com/news-story/us...cessors-surface-as-well-starts-at-316-59.html

https://www.reddit.com/r/Amd/comments/5sz8le

 

[EarthDog]

We've been chatting about it for a day or so.

 

[Helgaiden]

indeed

 

[M@rk]

We've been chatting about it for a day or so.

Oh, I didnt saw these links posted here though.

 

[Alaric]

I think it was page 71.

 

[EarthDog]

Pages can be different for people. I have mine set at 50 posts per page.. so I am page 30. Post #...

 

[Bluefalcon13]

Pages can be different for people. I have mine set at 50 posts per page.. so I am page 30. Post #...

And my app is set to 10/pg. So it's like 146 for me [emoji14]

 

[Alaric]

Pages can be different for people. I have mine set at 50 posts per page.. so I am page 30. Post #...

Well, that answers a bunch of other questions I haven't gotten around to asking. LOL

 

[ATMINSIDE]

https://www.overclock3d.net/news/cpu_mainboard/amd_ryzen_7_1700x_cpu_benchmarks_leak/1

 

[Soulcatcher668]

Still hedging my bets, but this is stacking up to be an epic win for AMD.

 

[Tír na nÓg]

Bloody Q2! AM4, 1800x, 1080ti, Vega... And as I can't resign myself to sell the Haswell-e in sig, will be on white rice+water diet for a few months! Maybe I can sell one of my kids... Never overclocked yet!

 

[Bluefalcon13]

I'm seriously considering selling a bunch of stuff to get a Mobo, DDR4, cpu, and a 480 ><

 

[Mandrake4565]

https://www.overclock3d.net/news/cpu_mainboard/amd_ryzen_7_1700x_cpu_benchmarks_leak/1

I really need to stop reading this thread. I'm chomping on the bit!!!!