[Skimmertalk] Improvement of CPU consumption by RTTY Skimserv

[Jeff Blaine]

Pete there are.  Not sure how they correlate.  Much better than a 
general CPU benchmark though.

The AMD 2700X at stock frequency, as a benchmark, runs about 1100-1200 
decoders before RTTY SS starts to throttle.  That should work for the 
normal conditions.  I seem to need about 400 decoders per open & active 
band and another 50-150 for each of the other bands with some activity.  
In peak solar condx, even this horsepower is going to be marginal if 
leaving SS set on 5 bands x 192.  Trimming back on the number of bands 
of course mitigates any shortcomings though.

73/jeff/ac0c
alpha-charlie-zero-charlie
www.ac0c.com


On 1/5/20 8:58 AM, N4ZR wrote:
>
> Jeff, what I was trying to point out is simply that judgments made 
> several years ago, that RTTY Skimserv would stop even the strongest 
> i7, perhaps need to be rethunk.
>
> Is there an available benchmark for FPU capability?
>
> 73, Pete N4ZR
> Check out the Reverse Beacon Network
> at<http://reversebeacon.net>, now
> spotting RTTY activity worldwide.
> For spots, please use your favorite
> "retail" DX cluster.
> On 1/5/2020 12:09 AM, Jeff Blaine wrote:
>>
>> The normal cpu benchmarks are worthless as predictors.  Look for 
>> something that is a reflection of the FPU capability.
>>
>> That also explains why the Fx series were so poor.  They have 2 ALU 
>> for 1 shared FPU.  All of the Intels (and the newer Amd) are 1:1 ALU:FPU.
>>
>> 73/jeff/ac0c
>> alpha-charlie-zero-charlie
>> www.ac0c.com
>>
>>
>> On 1/4/20 8:49 PM, N4ZR wrote:
>>>
>>> Before the RU started, I had CW Skimserv running on one receiver of 
>>> the RP-16, and RTTY Skimserv running on the other, both with 8 bands 
>>> x 192 KHz  I forgot to rearrange the RTTY bands when the contest 
>>> started, although I did select the right Contest.ini contest.  What 
>>> I wanted to comment on is the much improved CPU performance I'm 
>>> seeing, compared with the last time I ran RTTY Skimserv in a 
>>> contest.  Currently I have 339 decoders open on RTTY, and 444 spots 
>>> in the last half hour. It's running 63-67% CPU, with occasional 
>>> excursions into the 70s.  Meanwhile, CW is showing 410 decoders, 
>>> with 340 spots in 30 minutes, and 6-8 percent CPU.
>>>
>>> I don't know how to account for this improvement.  It is a new 
>>> computer, running Windows 10 pro.  The processor is an  Intel Core 
>>> i5-8400 CPU @ 2.80GHz, 2808 Mhz, 6 Core(s), 6 Logical Processor(s).  
>>> The machine has 16 GB of RAM and a 512 GB solid state drive, and the 
>>> CPU has a passmark of almost 12,000.  My old Skimmer machine has an 
>>> AMD FX-8350 with a passmark of almost 9,000, but there really is no 
>>> comparison so far as RTTY performance.
>>>
>>> I'm wondering if others have had similar experience with the i5 or 
>>> similar, newer CPUs.  Perhaps they are better adapted to running the 
>>> RTTY Skimserv code for some reason.
>>>
>>> A postscript - I just cut the RTTY Skimserv back to 6 bands x 
>>> 192KHz, and it is now drawing 55-61 percent CPU with 401 decoders 
>>> and 425 spots in the last 30 minutes.
>>>
>>> -- 
>>>
>>> 73, Pete N4ZR
>>> Check out the Reverse Beacon Network
>>> at<http://reversebeacon.net>, now
>>> spotting RTTY activity worldwide.
>>> For spots, please use your favorite
>>> "retail" DX cluster.
>>>
>>> _______________________________________________
>>> Skimmertalk mailing list
>>> Skimmertalk at contesting.com
>>> http://lists.contesting.com/mailman/listinfo/skimmertalk
>>
>> _______________________________________________
>> Skimmertalk mailing list
>> Skimmertalk at contesting.com
>> http://lists.contesting.com/mailman/listinfo/skimmertalk
>
> _______________________________________________
> Skimmertalk mailing list
> Skimmertalk at contesting.com
> http://lists.contesting.com/mailman/listinfo/skimmertalk
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.contesting.com/pipermail/skimmertalk/attachments/20200105/6c24583f/attachment-0001.html>