• We’re currently investigating an issue related to the forum theme and styling that is impacting page layout and visual formatting. The problem has been identified, and we are actively working on a resolution. There is no impact to user data or functionality, this is strictly a front-end display issue. We’ll post an update once the fix has been deployed. Thanks for your patience while we get this sorted.

AMD launches Ryzen Mobile 7 2700U & 5 2500U with Vega Graphics

Page 16 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.
scannall said:
They use a low power A series CPU to drive their touchbar, and assorted other functions. Increasing the load, or shuffling decode for example to a more powerful chip wouldn't be all that difficult.
Click to expand...
Oh that! I thought you meant something else... I thought they only used that for powering the Touchbar and TouchID(Secure Enclave)
 
For the one vs two CCX debate: Even with one CCX, Raven Ridge's memory latency is bad, or HP provided extremely high latency memory.
memlatency.png
 
Bouowmx said:
For the one vs two CCX debate: Even with one CCX, Raven Ridge's memory latency is bad, or HP provided extremely high latency memory.
memlatency.png
Click to expand...

AMD's Zen has high memory latency. Its a weakness which hurts CPU performance in games and other memory latency sensitive applications. AMD is surely learning from the first gen and improving their design. So it would be interesting to see how much they can improve going forward in future generations.
 
I cannot fathom why they could not have simply bought the Intel version of the Envy X360:

http://store.hp.com/us/en/mdp/laptops/envy-x360-204075--1#!&tab=vao

Its in the HP US store,for the same price as the Ryzen 5 version.

I mean also what was with them,dropping the Intel IGP testing halfway through the review??

Currently these reviews are more a test of the different laptop designs than a test of the chips inside them.
 
Last edited:
Which is a vastly better thing to be testing anyway of course - it isn't like you can user specify and buy which chip goes in which laptop 🙂
 
Bouowmx said:
The Tech Report (Jeff Kampman) on AMD Ryzen 5 2500U (15 W) in HP Envy x360: https://techreport.com/review/32877/amd-ryzen-5-2500u-apu-reviewed
Click to expand...
Well thats the sort of test we will see tons of.
The 15w ryzen r5 compared to an 35w tdp intel processor and a 25w - gpu alone - mx150.
Pretty stupid imo because its not remotely comparable products, and smells pretty bad especially as the intel drops out midway lol, but anyway some of the game results is a bit disapointing.
Could use some driver improvements and a ngg and tiled rendering working for that vega arch.
 
Last edited:
USER8000 said:
I mean also what was with them,dropping the Intel IGP testing halfway through the review??
Click to expand...

I thought the review was pretty clear about that; the Intel IGP couldn't run those games at all.

It is unfortunate that they did not include in power consumption tests or battery life metrics, although I understand why.
 
dahorns said:
I thought the review was pretty clear about that; the Intel IGP couldn't run those games at all.

It is unfortunate that they did not include in power consumption tests or battery life metrics, although I understand why.
Click to expand...
", and that seems to hurt the Ryzen 5 2500U in productivity tasks like web browsing and office work where responsiveness is paramount"

So a ipc deficit of 10% for productivity is blown up to this? I cant tell a difference from a ib to kbl ultrabook even with the latter having higher fmax for normalnofdice work and web browsing. Yet this sneaking "reponsiveness". R5 is running 3.6 and having high sustained freq as shown by cb. Bs all the way. There is plenty cpu grunt.

And then he keeps on comparing to a mx150 just 100 or 200 usd more. And forget its double tdp.

I simply dont understand amd didnt control this launch and made sure the apu was not compared in tons of graphs with 35w cpu and 20w gpu that visual presents the wrong picture.

The dell alienware used in the benchmark had a 35w configured cpu and a 1060 gpu ffs! Its like 100w capable chassis. In that chassis a 35w cpu can do a lot of stuff. Can it be more skewed let me know.

Not that a laptop sold with traditional hd should get any attention...
 
Last edited:
raghu78 said:
AMD's Zen has high memory latency. Its a weakness which hurts CPU performance in games and other memory latency sensitive applications. AMD is surely learning from the first gen and improving their design. So it would be interesting to see how much they can improve going forward in future generations.
Click to expand...
Nice finding! Now we can safely address the problem, it is the Infinity Fabric which adds latency to all the traffic going through it! To reduce the latency AMD need a serious bump in IF clock or it has to completely redesign the IF, which IMHO is impossible with the 12nm Zen+ refresh.
 
Desktop ryzen is hold back in games by the memory latency. And thats typically only for plus 60hz gaming with highest end gpu like 1080 or vega.
For these test of the apu variant they dont mean anything practical for gaming.
Its tdp, memory bandwith (a tad drivers) and third pure gpu size keeping performance back.
 
Bouowmx said:
For the one vs two CCX debate: Even with one CCX, Raven Ridge's memory latency is bad, or HP provided extremely high latency memory.
memlatency.png
Click to expand...

I am wondering about that, the gpu constantly needs to read from the dram (because of the framebuffer) to send video data to the lcd.
Because of that , that the test program perceives a higher latency than in reality.
I have no idea how a testprogram determines the latency. I assume by doing a large amount of memory reads and then averaging the result to one value. But if the cpu access and gpu access are interleaved meaning that the cpu has to wait for the gpu before it can access memory, then the testprogram would read a higher latency.
The apu has to share the memory with the gpu. It is possible the real latency might be more like ryzen. Of course, for real world applications it would not make a difference.
 
krumme said:
So a ipc deficit of 10% for productivity is blown up to this?
Click to expand...

The baseline between >= SKL vs. Zen IPC difference is > 12%.
Extremely low frequency (< 2400MHz) and high latency memory can increase it even further, but not decrease it.

The difference is manageable, unless the Intel parts has significantly higher frequency as well. Exactly what happened with CFL. Just stating the obvious.
 
The Stilt said:
The baseline between >= SKL vs. Zen IPC difference is > 12%.
Extremely low frequency (< 2400MHz) and high latency memory can increase it even further, but not decrease it.

The difference is manageable, unless the Intel parts has significantly higher frequency as well. Exactly what happened with CFL. Just stating the obvious.
Click to expand...

Looking at the cb scores I can not see the cfl 15w have higher freq than ryzen apu. R5 vs i5?

Anyway 10 or 15% difference i have used ultrabooks the last plus 10 years for office work and browsing and the comment about "responsiveness" for 3.6 4c cpu is just plain wrong. What makes a difference from ib forward for that kind of work is ssd iops random 4k read/write, ssd consistency under heavy load eg windows updates. Some newer pci ssd based solution is slower than eg a 840evo. What you want is a good consistently fast reliable ssd. The weak link especially for this 4c stuff is certainly there for office and browsing.

If you eg. encode h265 and not h264 or similar avx2 heavy load yes the i5 certainly makes more sense if you dont use the gpu for gaming. And for gaming with a nv1070 and up class you can use the better ram latency in older dx9 derived games. But imo its all extreme corner cases. For like 99% buying the ryzen is the save bet as the gpu perf is the weak link here too. Certainly if i buy a machine for officework i will get a r5 r7 as i can just give it to one of the kids later or heck even take a round of bf4 720 on low in an emergency. Far safer and more flexible buy. Similar argument to intel for server market.
All asuming battery life is compettitive !
I tend to think that was good enough from hsw ultrabooks and forward.
 
krumme said:
comment about "responsiveness" for 3.6 4c cpu is just plain wrong
Click to expand...
It's plain wrong because responsiveness is not about maximum frequency, but how quickly the processor can go from idle to maximum frequency. Third time I've seen that statement in this thread. See Intel Speed Shift.

krumme said:
take a round of bf4 720 on low in an emergency
Click to expand...
That is the best you can come up with, for using bigger integrated GPU? An AAA multiplayer, competitive FPS at reduced render resolution and 30 fps? ok. :grimacing:
 
gOJDO_n said:
No, it's not working that way! Ryzen has very high RAM latency because the IMC is not connected directly but over the IF.
Click to expand...

Yeah, yeah, i know that the IF is the slowdown factor.
But with an apu, having a gpu which has no dedicated VRAM and thus must share ram bandwidth with the cpu , i find the test interesting that it has exactly the same latency number as an 1800x
as tested in May this year by pcper.com in the worst case situation.


https://www.pcper.com/reviews/Processors/Ryzen-Memory-Latencys-Impact-Weak-1080p-Gaming


sandra01.png


Both 118.1 ns for DDR4-2400.
While Hardware.fr found an average of between 94ns and 100ns also at DDR4-2400. That is what made me curious and confused.
 
Bouowmx said:
It's plain wrong because responsiveness is not about maximum frequency, but how quickly the processor can go from idle to maximum frequency. Third time I've seen that statement in this thread. See Intel Speed Shift.


That is the best you can come up with, for using bigger integrated GPU? An AAA multiplayer, competitive FPS at reduced render resolution and 30 fps? ok. :grimacing:
Click to expand...

Take your Intel marketing speedstep nonsense elsewhere. 3% difference on standardized suites like pcmark we use for office work bm is what it gets. Nothing that makes a slightest difference here. As if people was waiting for the os to handle the voltage regulator.

And yes gaming is what makes a difference. People buy these machines also for occasional gaming and the ryzen is twice as fast. That counts for something. Not some marketing gimmick that only lives for a short blink.

You take speedstep. I take 100% gpu perf. Be my guest. Everyone is free to buy the core line forward on.
 
Bouowmx said:
Using a long-running benchmark (tens of minutes) to represent a centisecond technology. Right. ✓


In your hypothetical Battlefield 4 situation, why is going from unplayable resolution and frame rate to... unplayable resolution and frame rate acceptable?
Click to expand...

This APU is much faster than what it is in my laptop, and I manage to play many games. This APU kicks butt plain and simple for the amount of power it consumes.

I can think of many games that will run at 60fps native resolution for the APU.
 
No doubt that there are and will be titles that can run 60 fps (should be at least 1600x900) on AMD Raven Ridge, with the implication that Intel Gen9 GT2 cannot.

I am still curious on why the shining example is an AAA multiplayer, competitive FPS at 30 fps, and why that is acceptable.
 
Bouowmx said:
I am still curious on why the shining example is an AAA multiplayer, competitive FPS at 30 fps, and why that is acceptable.
Click to expand...

It isn't, but it is a popular game I guess. I played Battlefield 3 at 540p on a laptop to get decent fps. Never again.
 
You must log in or register to reply here.

TRENDING THREADS

Back
Top