• 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 Ryzen (Summit Ridge) Benchmarks Thread (use new thread)

Page 83 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.
Status
Not open for further replies.
Yeah I would love to know what it's going to be like, but I understand why we won't know until it's out.

I still have a fond memory of reading the 5870 review on Anandtech, being blown away and ordering it as soon as I was done reading it as it showed up on NewEgg. Probably the best PC component I ever bought.
 
sirmo said:
Yeah I would love to know what it's going to be like, but I understand why we won't know until it's out.

I still have a fond memory of reading the 5870 review on Anandtech, being blown away and ordering it as soon as I was done reading it as it showed up on NewEgg. Probably the best PC component I ever bought.
Click to expand...

Absolutely, I cannot wait for official samples to get sent out and reading all the reviews, hopefully for the better. And if things are as expected, I can't wait to try them out myself 😀
 
sirmo said:
I coded ASM in MS debug back in the early 90s before I learned C.
Click to expand...
Me too... But also Z80 (microcontroller with led numeric displays, at school), 68000 (PC emulator, not an actual computer, but a virtual microcontroller with mapped I/O, at school) and 6502/10 on Vic 20, commodore 64 and plus 4... Obviously without an assembler...Very challenging... Without debugger...
 
cytg111 said:
Not giving the competition more room than necessary to counter move. Would be my guess.
Click to expand...
AMD fans kept saying this at Barcelona and Bulldozer times. So did I.

Many others said its due to the products incompetence.

They were right and we were completely wrong.

Sent from HTC 10
(Opinions are own)
 
KTE said:
AMD fans kept saying this at Barcelona and Bulldozer times. So did I.

Many others said its due to the products incompetence.

They were right and we were completely wrong.

Sent from HTC 10
(Opinions are own)
Click to expand...
But in that same history you had things that could be pointed to from the very first presentations to backup that bad design. This time when push comes to shove the nay sayers haven't delivered anything of note that can be discussed and considered.
 
itsmydamnation said:
But in that same history you had things that could be pointed to from the very first presentations to backup that bad design. This time when push comes to shove the nay sayers haven't delivered anything of note that can be discussed and considered.
Click to expand...

When IPC was not known, they said that was at most SB levels, because misinterpretation of the 40% claim (max, mean, relative to BD and not XV etc.).

After IPC was known, they said that in 95W Zen can't anyway beat INTEL clocks, because of superior INTEL process, forgetting the 19 stages of Zen, leading to a low FO4, in the ballpark of XV, that on a shitty 28nm process and HDL low power library, let XV clock at 4.3GHz. Then came blender test, with low consumption on an ES and the latest rumors of ES at 3.2/3.5GHz, at 95W, versus 3.2/3.7 at 140W...

I don't know latest excuses, besides, "because it's AMD" or "low power process, so low clock". Let's see...
 
bjt2 said:
and the latest rumors of ES at 3.2/3.5GHz, at 95W, versus 3.2/3.7 at 140W...
Click to expand...

It sounds like the 3.2/3.5 is going to be 140W, and that's going to be the "Black Edition" model. AMD of course would have had to change their mind about only doing 95W models.
 
Still one major major factor here.. Is this supposed to be GloFo's first successful launch ever? Those E samples, are they from glofo?
 
sirmo said:
Yeah I would love to know what it's going to be like, but I understand why we won't know until it's out.

I still have a fond memory of reading the 5870 review on Anandtech, being blown away and ordering it as soon as I was done reading it as it showed up on NewEgg. Probably the best PC component I ever bought.
Click to expand...

It actually started with the 4870. The rumour was AMD quit competing for the high-end GPU market, then they dropped the tiny 4870 on to the huge 280, with no fanfare. Nvidia had to slash prices a month after release.

Some say that AMD would be shouting loud if they had something good. They also say that everytime AMD shout out loud in the past they had a bad chip, eg Barcelona or Bulldozer. 🙂

More often that not, AMD keeps their best stuff under wraps.
 
KTE said:
AMD fans kept saying this at Barcelona and Bulldozer times. So did I.

Many others said its due to the products incompetence.

They were right and we were completely wrong.
Click to expand...

Yep.

Hence my rather low expectations of Zen and my tag of #OnceBitten !!!
 
Sven_eng said:
It actually started with the 4870. The rumour was AMD quit competing for the high-end GPU market, then they dropped the tiny 4870 on to the huge 280, with no fanfare. Nvidia had to slash prices a month after release.

Some say that AMD would be shouting loud if they had something good. They also say that everytime AMD shout out loud in the past they had a bad chip, eg Barcelona or Bulldozer. 🙂

More often that not, AMD keeps their best stuff under wraps.
Click to expand...

Yes, I still remember HD4850 forcing NVIDIA slashing price of its whole mid-high end lineup. Not very old story but just happened 8 years ago. Of course this require you have not-bad memory to recall, but unfortunately people are just forgetful, short-memory here.
 
Atari2600 said:
Just as an aside, if your worried about python being too slow, then after building and profiling your code, you can use cython to give the bottlenecks a helping hand:
Click to expand...

For math work, Cython does practically nothing. The point of using Python for math is that Python is only used as glue code, all the actual work is done in numpy. Numpy is very heavily optimized and really fast, and making the glue run faster generally doesn't do anything.
 
Latest versions of matlab increase dramatically the speed of some routines, namely fft, convolutions and other heavy calculus. Moreover has an autoparallelization for matrix calculations since years... All is done in library (linpack if i remember well), has a jit (not interpreted) a compiler (optional), a parallel toolbox (optional) and a GPU toolbox (optional). Without GPU toolbox, you can still use CUDA: the mex files are functions or scripts compiled as dll, translated from matlab to c (if you have the compiler) or manually written (including the right headers and libraries) and then compiled with the supported compiler. To use cuda you add cuda headers and link cuda libraries to the dll and voilà...
 
Tuna-Fish said:
For math work, Cython does practically nothing. The point of using Python for math is that Python is only used as glue code, all the actual work is done in numpy. Numpy is very heavily optimized and really fast, and making the glue run faster generally doesn't do anything.
Click to expand...

Be careful with that assumption, you may find it useful to at least check some of your code if its repeated & large operations:

http://stackoverflow.com/questions/7799977/numpy-vs-cython-speed

http://notes-on-cython.readthedocs.io/en/latest/std_dev.html
 
1D3201A2M88F3_35/32_N

So...

Thats:
1 - Unknown sample type
D - Desktop
32 - 3200 MHz Clock (don't know whether base or peak - given then inclusion of 35/32 later on, we could speculate 3.2 base, 3.5 boost)
01 - Single socket ?
A2 - Unknown, assumed reference to socket type
M - Not sure. Tradition would have it mean 638 pins. We know that can't be right.
8 - 8 Cores
8 - Unknown, traditionally ref to cache size
F3 - Microarchitecture & Revision



Similar OPNs have been observed here:
http://wccftech.com/amd-zen-es-benchmarks-leak-out/

1D2801A2M88E4_32/28_N

They have purported Ashes of the Singularity results there too.

***Assuming that the 1D2801A2M88E4_32/28_N corresponds to a 2.8 GHz base w/3.2 GHz boost, and the FPS of 58/sec vs. an i7-4790 @ 4.0 GHz getting 65.4/sec is valid***

Then Zen may more or less equate to Haswell in some benchmarks, ahead in IPC but behind in overall speed due to the lower clocks - which may or may not come with a lower TDP as well.


All extrapolation and guesswork of course.
 
Atari2600 said:
Then Zen may more or less equate to Haswell in some benchmarks, ahead in IPC but behind in overall speed due to the lower clocks - which may or may not come with a lower TDP as well.
Click to expand...
Not sure how you come to conclusion of higher IPC when 8 cores at 25% lower clock are slower than 4 cores.
 
Atari2600 said:
Duuuuu me! 😳

Although I guess this would suggest it doesn't do well beyond 4 anyway:

http://semiaccurate.com/2016/03/01/investigating-directx-12-cpu-scaling/
Click to expand...

Um...

dx12_cpu_ashes_of_the_singularity_beta_2_average_cpu_frame_rate_high_quality_19x10-100647718-orig.png
 
Atari2600 said:
1D3201A2M88F3_35/32_N

So...

Thats:
1 - Unknown sample type
D - Desktop
32 - 3200 MHz Clock (don't know whether base or peak - given then inclusion of 35/32 later on, we could speculate 3.2 base, 3.5 boost)
01 - Single socket ?
A2 - Unknown, assumed reference to socket type
M - Not sure. Tradition would have it mean 638 pins. We know that can't be right.
8 - 8 Cores
8 - Unknown, traditionally ref to cache size
F3 - Microarchitecture & Revision
Click to expand...

Not bad. Not that long ago I found this useful sheet: http://www.cpu-world.com/info/id/AMD-ES-identification.html
 
Status
Not open for further replies.

TRENDING THREADS

Back
Top