NONMEM Users Network Archive

Hosted by Cognigen

Re: AMD vs Intel

From: Jeroen Elassaiss-Schaap <jeroen>
Date: Tue, 19 Nov 2019 00:08:53 +0100

Hi Leonid,

In that case, have you tried virtualization? Couldn’t help but notic=
e involvement of VMware in Epyc. If you run from virtual, you have an additi=
onal abstraction layer. You still have to deal with the optimizations of VMw=
are unfortunately, but it might improve comparability.

Best,
Jeroen

http://pd-value.com
jeroen
  
+31 6 23118438
-- More value out of your data!

> Op 18 nov. 2019 om 23:54 heeft Leonid Gibiansky <lgibiansky
> het volgende geschreven:
>
> Hi Jeroin,
>
> Thanks for your input, very interesting. As far as the goal is concerned, I=
 am mostly interested to find options that would give identical results on t=
wo platform rather than in speed. So far no luck: 4 combinations of gfortran=
 / Intel compilers on Xeon / AMD processors give 4 sets of results that are c=
lose but not identical.
>
> Related question to the group: have anybody experimented with gfortran opt=
ions (rather than using default provided by Nonmem distribution)? Any recomm=
endations? Same goal: maximum reproducibility across different OSs, parallel=
ization options, and processor types.
>
> Thanks
> Leonid
>
>
>
>
>> On 11/18/2019 5:28 PM, Jeroen Elassaiss-Schaap (PD-value B.V.) wrote:
>> Hi Leonid,
>> "A while" back we compared model development trajectories and results bet=
ween two computational platforms, Itanium and Xeon, see https://www.page-mee=
ting.org/?abstract=1188. The results roughly were: 1/3 equal, 1/3 rounding=
 differences and 1/3 real different results. From discussions with the tec=
hnical knowledgeable people I worked with at the time, I recall that there a=
re three levels/sources for those differences:
>> 1) computational (hardware) platform
>> 2) compilers (+ optimization settings)
>> 3) libraries (floating point handling does matter)
>> Assuming you would like to compare the speed of the platforms wrt NONMEM,=
 my advice would be to test a large series of different models, from simple A=
DVAN1 or 2 to complex ODE, ranging from FO to LAPLACIAN INT NUMERICAL, while=
 keeping compilers and libraries the same. Also small and large datasets, as=
 in some instances you might be testing only the L1/L2/L3 cache strategies a=
nd Turbo settings. And with and without parallelization - as that might dete=
rmine runtime bottlenecks in practice.
>> Just having a peek at Epyc - seems interesting (noticed results w gcc7.4 c=
ompilation). As long as you are able to hold the computation in cache, a big=
 if for the 64-core, there might be an advantage.
>> All in all I am not sure that it is worth the trouble. For any given PK-P=
D model there is a lot you can tune to gain speed, but the optimal settings m=
ight be very different for the next and overrule any platform differences.
>> Hope this helps,
>> Jeroen
>> http://pd-value.com
>> jeroen
>>
>> +31 6 23118438
>> -- More value out of your data!
>>> On 18/11/19 6:34 pm, Leonid Gibiansky wrote:
>>> Thanks Bob and Peter!
>>>
>>> The model is quite stable, but this is LAPLACIAN, so requires second der=
ivatives. At iteration 0, gradients differ by about 50 to 100% between Inte=
l and AMD. This leads to differences in minimization path, and slightly diff=
erent results. Not that different to change the recommended dose, but suffic=
iently different to notice (OF difference of 6 points; 50% more model evalua=
tions to get to convergence).
>>> Thanks
>>> Leonid
>>>
>>>
>>>
>>> On 11/18/2019 12:15 PM, Bonate, Peter wrote:
>>>> Leonid - when you say different. What do you mean? Fixed effect and r=
andom effects? Different OFV?
>>>>
>>>> We did a poster at AAPS a decade or so ago comparing results across dif=
ferent platforms using the same data and model. We got different results on=
 the standard errors (which related to matrix inversion and how those are do=
ne using software-hardware configurations). And with overparameterized mode=
ls we got different error messages - some platforms converged with no proble=
m while some did not converge and gave R matrix singularity.
>>>>
>>>> Did your problems go beyond this?
>>>>
>>>> pete
>>>>
>>>>
>>>>
>>>> Peter Bonate, PhD
>>>> Executive Director
>>>> Pharmacokinetics, Modeling, and Simulation
>>>> Astellas
>>>> 1 Astellas Way, N3.158
>>>> Northbrook, IL 60062
>>>> Peter.bonate
>>>> (224) 205-5855
>>>>
>>>>
>>>>
>>>> Details are irrelevant in terms of decision making - Joe Biden.
>>>>
>>>>
>>>>
>>>>
>>>>
>>>>
>>>> -----Original Message-----
>>>> From: owner-nmusers
half Of Leonid Gibiansky
>>>> Sent: Monday, November 18, 2019 11:05 AM
>>>> To: nmusers <nmusers
>>>> Subject: [NMusers] AMD vs Intel
>>>>
>>>> Dear All,
>>>>
>>>> I am testing the new Epyc processors from AMD (comparing with Intel Xeo=
n), and getting different results. Just wondering whether anybody faced the p=
roblem of differences between AMD and Intel processors and knows how to solv=
e it. I am using Intel compiler but ready to switch to gfortran or anything e=
lse if this would help to get identical results.
>>>> There were reports of Intel slowing the AMD execution in the past, but i=
n my tests, speed is comparable but the results differ.
>>>>
>>>> Thanks
>>>> Leonid
>>>>
>>>>
>>>>
>>>

Received on Mon Nov 18 2019 - 18:08:53 EST

The NONMEM Users Network is maintained by ICON plc. Requests to subscribe to the network should be sent to: nmusers-request@iconplc.com.

Once subscribed, you may contribute to the discussion by emailing: nmusers@globomaxnm.com.