There some data about chess power of Ryzen 7 processor but these are very confused.
Therefore I decided to make tests based on well defined parameters.
My rig has a Ryzen 7 1800x processor with 3700 MHz (DEFAULT) CPU clock speed what is overclocked to 8x4000 MHz.
It has 16 GB RAM with 2133 MHz (DEFAULT) clock speed.
The installed operating systems are Win 7 32bits, Win 7 64bits and Win 10 64bits. For tests I used Win 7 64bits system
The tests what I made as follows.
(Sorry, but I have to brake the text because of uncertain net connection )
Chess power of Ryzen processor
Moderators: hgm, Rebel, chrisw
-
corres
- Posts: 3657
- Joined: Wed Nov 18, 2015 11:41 am
- Location: hungary
-
corres
- Posts: 3657
- Joined: Wed Nov 18, 2015 11:41 am
- Location: hungary
Re: Chess power of Ryzen processor
[quote="corres"]
There some data about chess power of Ryzen 7 processor but these are very confused.
Therefore I decided to make tests based on well defined parameters.
My rig has a Ryzen 7 1800x processor with 3700 MHz (DEFAULT) CPU clock speed what is overclocked to 8x4000 MHz.
It has 16 GB RAM with 2133 MHz (DEFAULT) clock speed.
The installed operating systems are Win 7 32bits, Win 7 64bits and Win 10 64bits. For tests I used Win 7 64bits system
The tests what I made as follows.
(Sorry, but I have to brake the text because of uncertain net connection )
[/quote]
TEST 1: Fritz Chess Benchmark
I used benchmark program of Deep Fritz 14.
-SMT (Simultaneous Multithreading, ~HT)=OFF cores=8 (physical cores)
Relative Performance=36.0 speed= 17400 kn/sec
-SMT=AUTO (=ON) cores=16 (8 physical+8 logical)
Relative Performance=46.0 speed=22200 kn/sec
There some data about chess power of Ryzen 7 processor but these are very confused.
Therefore I decided to make tests based on well defined parameters.
My rig has a Ryzen 7 1800x processor with 3700 MHz (DEFAULT) CPU clock speed what is overclocked to 8x4000 MHz.
It has 16 GB RAM with 2133 MHz (DEFAULT) clock speed.
The installed operating systems are Win 7 32bits, Win 7 64bits and Win 10 64bits. For tests I used Win 7 64bits system
The tests what I made as follows.
(Sorry, but I have to brake the text because of uncertain net connection )
[/quote]
TEST 1: Fritz Chess Benchmark
I used benchmark program of Deep Fritz 14.
-SMT (Simultaneous Multithreading, ~HT)=OFF cores=8 (physical cores)
Relative Performance=36.0 speed= 17400 kn/sec
-SMT=AUTO (=ON) cores=16 (8 physical+8 logical)
Relative Performance=46.0 speed=22200 kn/sec
-
corres
- Posts: 3657
- Joined: Wed Nov 18, 2015 11:41 am
- Location: hungary
Re: Chess power of Ryzen processor
[quote="corres"]
[quote="corres"]
There some data about chess power of Ryzen 7 processor but these are very confused.
Therefore I decided to make tests based on well defined parameters.
My rig has a Ryzen 7 1800x processor with 3700 MHz (DEFAULT) CPU clock speed what is overclocked to 8x4000 MHz.
It has 16 GB RAM with 2133 MHz (DEFAULT) clock speed.
The installed operating systems are Win 7 32bits, Win 7 64bits and Win 10 64bits. For tests I used Win 7 64bits system
The tests what I made as follows.
(Sorry, but I have to brake the text because of uncertain net connection )
[/quote]
TEST 1: Fritz Chess Benchmark
I used benchmark program of Deep Fritz 14.
-SMT (Simultaneous Multithreading, ~HT)=OFF cores=8 (physical cores)
Relative Performance=36.0 speed= 17400 kn/sec
-SMT=AUTO (=ON) cores=16 (8 physical+8 logical)
Relative Performance=46.0 speed=22200 kn/sec
[/quote]
TEST 2: Stockfish speed tests
I used Stockfish Speed Test by Bruce Allenbrand
for Stockfish dev.2017.04.23.
-1 core (Hash=512 MB, Depth=15) SMT=DISABLED speed=2253 Mn/sec
-4 cores (Hash=512 Mb, Depth=16) SMT=DISABLED speed=8584 Mn/sec
-8 cores (Hash=512 MB, Depth=16) SMT=DISABLED speed=16000 Mn/sec
-16 cores (Hash=512 MB,Depth=16) SMT=AUTO (=ENABLED!)
speed=23600 Mn/sec
As you can see the the enhancement of speed is about 8000 Mn/sec
from 4 cores to 8 cores and from 8 cores to 16 cores.
[quote="corres"]
There some data about chess power of Ryzen 7 processor but these are very confused.
Therefore I decided to make tests based on well defined parameters.
My rig has a Ryzen 7 1800x processor with 3700 MHz (DEFAULT) CPU clock speed what is overclocked to 8x4000 MHz.
It has 16 GB RAM with 2133 MHz (DEFAULT) clock speed.
The installed operating systems are Win 7 32bits, Win 7 64bits and Win 10 64bits. For tests I used Win 7 64bits system
The tests what I made as follows.
(Sorry, but I have to brake the text because of uncertain net connection )
[/quote]
TEST 1: Fritz Chess Benchmark
I used benchmark program of Deep Fritz 14.
-SMT (Simultaneous Multithreading, ~HT)=OFF cores=8 (physical cores)
Relative Performance=36.0 speed= 17400 kn/sec
-SMT=AUTO (=ON) cores=16 (8 physical+8 logical)
Relative Performance=46.0 speed=22200 kn/sec
[/quote]
TEST 2: Stockfish speed tests
I used Stockfish Speed Test by Bruce Allenbrand
for Stockfish dev.2017.04.23.
-1 core (Hash=512 MB, Depth=15) SMT=DISABLED speed=2253 Mn/sec
-4 cores (Hash=512 Mb, Depth=16) SMT=DISABLED speed=8584 Mn/sec
-8 cores (Hash=512 MB, Depth=16) SMT=DISABLED speed=16000 Mn/sec
-16 cores (Hash=512 MB,Depth=16) SMT=AUTO (=ENABLED!)
speed=23600 Mn/sec
As you can see the the enhancement of speed is about 8000 Mn/sec
from 4 cores to 8 cores and from 8 cores to 16 cores.
-
corres
- Posts: 3657
- Joined: Wed Nov 18, 2015 11:41 am
- Location: hungary
Re: Chess power of Ryzen processor
[quote="corres"]
[quote="corres"]
[quote="corres"]
There some data about chess power of Ryzen 7 processor but these are very confused.
Therefore I decided to make tests based on well defined parameters.
My rig has a Ryzen 7 1800x processor with 3700 MHz (DEFAULT) CPU clock speed what is overclocked to 8x4000 MHz.
It has 16 GB RAM with 2133 MHz (DEFAULT) clock speed.
The installed operating systems are Win 7 32bits, Win 7 64bits and Win 10 64bits. For tests I used Win 7 64bits system
The tests what I made as follows.
(Sorry, but I have to brake the text because of uncertain net connection )
[/quote]
TEST 1: Fritz Chess Benchmark
I used benchmark program of Deep Fritz 14.
-SMT (Simultaneous Multithreading, ~HT)=OFF cores=8 (physical cores)
Relative Performance=36.0 speed= 17400 kn/sec
-SMT=AUTO (=ON) cores=16 (8 physical+8 logical)
Relative Performance=46.0 speed=22200 kn/sec
[/quote]
TEST 2: Stockfish speed tests
I used Stockfish Speed Test by Bruce Allenbrand
for Stockfish dev.2017.04.23.
-1 core (Hash=512 MB, Depth=15) SMT=DISABLED speed=2253 Mn/sec
-4 cores (Hash=512 Mb, Depth=16) SMT=DISABLED speed=8584 Mn/sec
-8 cores (Hash=512 MB, Depth=16) SMT=DISABLED speed=16000 Mn/sec
-16 cores (Hash=512 MB,Depth=16) SMT=AUTO (=ENABLED!)
speed=23600 Mn/sec
As you can see the the enhancement of speed is about 8000 Mn/sec
from 4 cores to 8 cores and from 8 cores to 16 cores.
[/quote]
TEST 3: CHESS POWER TESTS
For the users of chess engines is more important the chess power of a CPU than the speed produced by the given CPU.
In the case of one core CPU the speed and power are well correlated to each other. But for a CPU with several cores the connection between speed and power depends on the structure of the given engine, the parameters of the given CPU and the operating system used. You should make a lot of engine-engine tests running one of engine on a PC with Ryzen processor and the other engine running on an Intel processor.
But there is a method to gain relative good estimation on the chess power: To measure the solving time of well defined positions.
For making these tests I compiled a database what contains 70 tactical test positions with one characterized solution. To determine the solving time I used from Fritz tools the "Process Test Sets" tool. I run it five times
and the average was calculated.
[quote="corres"]
[quote="corres"]
There some data about chess power of Ryzen 7 processor but these are very confused.
Therefore I decided to make tests based on well defined parameters.
My rig has a Ryzen 7 1800x processor with 3700 MHz (DEFAULT) CPU clock speed what is overclocked to 8x4000 MHz.
It has 16 GB RAM with 2133 MHz (DEFAULT) clock speed.
The installed operating systems are Win 7 32bits, Win 7 64bits and Win 10 64bits. For tests I used Win 7 64bits system
The tests what I made as follows.
(Sorry, but I have to brake the text because of uncertain net connection )
[/quote]
TEST 1: Fritz Chess Benchmark
I used benchmark program of Deep Fritz 14.
-SMT (Simultaneous Multithreading, ~HT)=OFF cores=8 (physical cores)
Relative Performance=36.0 speed= 17400 kn/sec
-SMT=AUTO (=ON) cores=16 (8 physical+8 logical)
Relative Performance=46.0 speed=22200 kn/sec
[/quote]
TEST 2: Stockfish speed tests
I used Stockfish Speed Test by Bruce Allenbrand
for Stockfish dev.2017.04.23.
-1 core (Hash=512 MB, Depth=15) SMT=DISABLED speed=2253 Mn/sec
-4 cores (Hash=512 Mb, Depth=16) SMT=DISABLED speed=8584 Mn/sec
-8 cores (Hash=512 MB, Depth=16) SMT=DISABLED speed=16000 Mn/sec
-16 cores (Hash=512 MB,Depth=16) SMT=AUTO (=ENABLED!)
speed=23600 Mn/sec
As you can see the the enhancement of speed is about 8000 Mn/sec
from 4 cores to 8 cores and from 8 cores to 16 cores.
[/quote]
TEST 3: CHESS POWER TESTS
For the users of chess engines is more important the chess power of a CPU than the speed produced by the given CPU.
In the case of one core CPU the speed and power are well correlated to each other. But for a CPU with several cores the connection between speed and power depends on the structure of the given engine, the parameters of the given CPU and the operating system used. You should make a lot of engine-engine tests running one of engine on a PC with Ryzen processor and the other engine running on an Intel processor.
But there is a method to gain relative good estimation on the chess power: To measure the solving time of well defined positions.
For making these tests I compiled a database what contains 70 tactical test positions with one characterized solution. To determine the solving time I used from Fritz tools the "Process Test Sets" tool. I run it five times
and the average was calculated.
-
corres
- Posts: 3657
- Joined: Wed Nov 18, 2015 11:41 am
- Location: hungary
Re: Chess power of Ryzen processor
[quote="corres"][quote="corres"]
[quote="corres"]
[quote="corres"]
There some data about chess power of Ryzen 7 processor but these are very confused.
Therefore I decided to make tests based on well defined parameters.
My rig has a Ryzen 7 1800x processor with 3700 MHz (DEFAULT) CPU clock speed what is overclocked to 8x4000 MHz.
It has 16 GB RAM with 2133 MHz (DEFAULT) clock speed.
The installed operating systems are Win 7 32bits, Win 7 64bits and Win 10 64bits. For tests I used Win 7 64bits system
The tests what I made as follows.
(Sorry, but I have to brake the text because of uncertain net connection )
[/quote]
TEST 1: Fritz Chess Benchmark
I used benchmark program of Deep Fritz 14.
-SMT (Simultaneous Multithreading, ~HT)=OFF cores=8 (physical cores)
Relative Performance=36.0 speed= 17400 kn/sec
-SMT=AUTO (=ON) cores=16 (8 physical+8 logical)
Relative Performance=46.0 speed=22200 kn/sec
[/quote]
TEST 2: Stockfish speed tests
I used Stockfish Speed Test by Bruce Allenbrand
for Stockfish dev.2017.04.23.
-1 core (Hash=512 MB, Depth=15) SMT=DISABLED speed=2253 Mn/sec
-4 cores (Hash=512 Mb, Depth=16) SMT=DISABLED speed=8584 Mn/sec
-8 cores (Hash=512 MB, Depth=16) SMT=DISABLED speed=16000 Mn/sec
-16 cores (Hash=512 MB,Depth=16) SMT=AUTO (=ENABLED!)
speed=23600 Mn/sec
As you can see the the enhancement of speed is about 8000 Mn/sec
from 4 cores to 8 cores and from 8 cores to 16 cores.
[/quote]
TEST 3: CHESS POWER TESTS
For the users of chess engines is more important the chess power of a CPU than the speed produced by the given CPU.
In the case of one core CPU the speed and power are well correlated to each other. But for a CPU with several cores the connection between speed and power depends on the structure of the given engine, the parameters of the given CPU and the operating system used. You should make a lot of engine-engine tests running one of engine on a PC with Ryzen processor and the other engine running on an Intel processor.
But there is a method to gain relative good estimation on the chess power: To measure the solving time of well defined positions.
For making these tests I compiled a database what contains 70 tactical test positions with one characterized solution. To determine the solving time I used from Fritz tools the "Process Test Sets" tool. I run it five times
and the average was calculated.
[/quote]
(continuation of TEST 3)
Results:
-SMT=DISABLED average solving time=0.42 sec
-SMT=AUTO (=ENABLED) average solving time=0.42 sec
As you can see SMT switch ON or OFF has no effect (or a small effect) on the solving time.
That is while switching SMT ON and OFF the speed changes a lot, the chess power of CPU changes barely.
FOR UNANIMITY, THE ABOVE TESTS WAS MADE ON OVERCLOCKED
(8x4000 MHZ) RYZEN 7 1800x CPU!!!
Because the effect of SMT (and HT too) depend on the power state the given CPU I repeated this test with DEFAULT CPU clock frequency (=3700 MHz)
Results:
-SMT=DISABLED average solving time= 0.52 sec
-SMT=AUTO average solving time=0.43 sec
That is the shortening of solving time is abut 20 %.
So it is obvious that for an overclocked Ryzen CPU you ought to switch of SMT, but a non-overclocked Ryzen CPU you should switch on SMT.
Note:
1. Using SMT the temperature of CPU arises dramatically so you need a very effective cooler!
2,As I mentioned this test only an estimation for chess power of Ryzen 7.
You should make a lot of engine-engine tests to state the exact verdict.
[quote="corres"]
[quote="corres"]
There some data about chess power of Ryzen 7 processor but these are very confused.
Therefore I decided to make tests based on well defined parameters.
My rig has a Ryzen 7 1800x processor with 3700 MHz (DEFAULT) CPU clock speed what is overclocked to 8x4000 MHz.
It has 16 GB RAM with 2133 MHz (DEFAULT) clock speed.
The installed operating systems are Win 7 32bits, Win 7 64bits and Win 10 64bits. For tests I used Win 7 64bits system
The tests what I made as follows.
(Sorry, but I have to brake the text because of uncertain net connection )
[/quote]
TEST 1: Fritz Chess Benchmark
I used benchmark program of Deep Fritz 14.
-SMT (Simultaneous Multithreading, ~HT)=OFF cores=8 (physical cores)
Relative Performance=36.0 speed= 17400 kn/sec
-SMT=AUTO (=ON) cores=16 (8 physical+8 logical)
Relative Performance=46.0 speed=22200 kn/sec
[/quote]
TEST 2: Stockfish speed tests
I used Stockfish Speed Test by Bruce Allenbrand
for Stockfish dev.2017.04.23.
-1 core (Hash=512 MB, Depth=15) SMT=DISABLED speed=2253 Mn/sec
-4 cores (Hash=512 Mb, Depth=16) SMT=DISABLED speed=8584 Mn/sec
-8 cores (Hash=512 MB, Depth=16) SMT=DISABLED speed=16000 Mn/sec
-16 cores (Hash=512 MB,Depth=16) SMT=AUTO (=ENABLED!)
speed=23600 Mn/sec
As you can see the the enhancement of speed is about 8000 Mn/sec
from 4 cores to 8 cores and from 8 cores to 16 cores.
[/quote]
TEST 3: CHESS POWER TESTS
For the users of chess engines is more important the chess power of a CPU than the speed produced by the given CPU.
In the case of one core CPU the speed and power are well correlated to each other. But for a CPU with several cores the connection between speed and power depends on the structure of the given engine, the parameters of the given CPU and the operating system used. You should make a lot of engine-engine tests running one of engine on a PC with Ryzen processor and the other engine running on an Intel processor.
But there is a method to gain relative good estimation on the chess power: To measure the solving time of well defined positions.
For making these tests I compiled a database what contains 70 tactical test positions with one characterized solution. To determine the solving time I used from Fritz tools the "Process Test Sets" tool. I run it five times
and the average was calculated.
[/quote]
(continuation of TEST 3)
Results:
-SMT=DISABLED average solving time=0.42 sec
-SMT=AUTO (=ENABLED) average solving time=0.42 sec
As you can see SMT switch ON or OFF has no effect (or a small effect) on the solving time.
That is while switching SMT ON and OFF the speed changes a lot, the chess power of CPU changes barely.
FOR UNANIMITY, THE ABOVE TESTS WAS MADE ON OVERCLOCKED
(8x4000 MHZ) RYZEN 7 1800x CPU!!!
Because the effect of SMT (and HT too) depend on the power state the given CPU I repeated this test with DEFAULT CPU clock frequency (=3700 MHz)
Results:
-SMT=DISABLED average solving time= 0.52 sec
-SMT=AUTO average solving time=0.43 sec
That is the shortening of solving time is abut 20 %.
So it is obvious that for an overclocked Ryzen CPU you ought to switch of SMT, but a non-overclocked Ryzen CPU you should switch on SMT.
Note:
1. Using SMT the temperature of CPU arises dramatically so you need a very effective cooler!
2,As I mentioned this test only an estimation for chess power of Ryzen 7.
You should make a lot of engine-engine tests to state the exact verdict.
-
corres
- Posts: 3657
- Joined: Wed Nov 18, 2015 11:41 am
- Location: hungary
Re: Chess power of Ryzen processor
[quote="corres"][quote="corres"][quote="corres"]
[quote="corres"]
[quote="corres"]
There some data about chess power of Ryzen 7 processor but these are very confused.
Therefore I decided to make tests based on well defined parameters.
My rig has a Ryzen 7 1800x processor with 3700 MHz (DEFAULT) CPU clock speed what is overclocked to 8x4000 MHz.
It has 16 GB RAM with 2133 MHz (DEFAULT) clock speed.
The installed operating systems are Win 7 32bits, Win 7 64bits and Win 10 64bits. For tests I used Win 7 64bits system
The tests what I made as follows.
(Sorry, but I have to brake the text because of uncertain net connection )
[/quote]
TEST 1: Fritz Chess Benchmark
I used benchmark program of Deep Fritz 14.
-SMT (Simultaneous Multithreading, ~HT)=OFF cores=8 (physical cores)
Relative Performance=36.0 speed= 17400 kn/sec
-SMT=AUTO (=ON) cores=16 (8 physical+8 logical)
Relative Performance=46.0 speed=22200 kn/sec
[/quote]
TEST 2: Stockfish speed tests
I used Stockfish Speed Test by Bruce Allenbrand
for Stockfish dev.2017.04.23.
-1 core (Hash=512 MB, Depth=15) SMT=DISABLED speed=2253 Mn/sec
-4 cores (Hash=512 Mb, Depth=16) SMT=DISABLED speed=8584 Mn/sec
-8 cores (Hash=512 MB, Depth=16) SMT=DISABLED speed=16000 Mn/sec
-16 cores (Hash=512 MB,Depth=16) SMT=AUTO (=ENABLED!)
speed=23600 Mn/sec
As you can see the the enhancement of speed is about 8000 Mn/sec
from 4 cores to 8 cores and from 8 cores to 16 cores.
[/quote]
TEST 3: CHESS POWER TESTS
For the users of chess engines is more important the chess power of a CPU than the speed produced by the given CPU.
In the case of one core CPU the speed and power are well correlated to each other. But for a CPU with several cores the connection between speed and power depends on the structure of the given engine, the parameters of the given CPU and the operating system used. You should make a lot of engine-engine tests running one of engine on a PC with Ryzen processor and the other engine running on an Intel processor.
But there is a method to gain relative good estimation on the chess power: To measure the solving time of well defined positions.
For making these tests I compiled a database what contains 70 tactical test positions with one characterized solution. To determine the solving time I used from Fritz tools the "Process Test Sets" tool. I run it five times
and the average was calculated.
[/quote]
(continuation of TEST 3)
Results:
-SMT=DISABLED average solving time=0.42 sec
-SMT=AUTO (=ENABLED) average solving time=0.42 sec
As you can see SMT switch ON or OFF has no effect (or a small effect) on the solving time.
That is while switching SMT ON and OFF the speed changes a lot, the chess power of CPU changes barely.
FOR UNANIMITY, THE ABOVE TESTS WAS MADE ON OVERCLOCKED
(8x4000 MHZ) RYZEN 7 1800x CPU!!!
Because the effect of SMT (and HT too) depend on the power state the given CPU I repeated this test with DEFAULT CPU clock frequency (=3700 MHz)
Results:
-SMT=DISABLED average solving time= 0.52 sec
-SMT=AUTO average solving time=0.43 sec
That is the shortening of solving time is abut 20 %.
So it is obvious that for an overclocked Ryzen CPU you ought to switch of SMT, but a non-overclocked Ryzen CPU you should switch on SMT.
Note:
1. Using SMT the temperature of CPU arises dramatically so you need a very effective cooler!
2,As I mentioned this test only an estimation for chess power of Ryzen 7.
You should make a lot of engine-engine tests to state the exact verdict.
[/quote]
Sorry, but I have to make some corrections.
The text
"TEST 1: Fritz Chess Benchmark"
is incomplete and inaccurate.
The result reported is related to CPU clock frequency= 3700 MHz (DEFAULT)
only.
So the full and correct results are:
-At CPU clock frequency=3700 MHz (DEFAULT)
--SMT=DISABLED
Relative Performance=36.3 speed=17400 kn/sec
--SMT=AUTO (=ENABLED)
Relative Performance=46.0 speed=22200 kn/sec
-At CPU clock frequency=4000 MHz (OVERCLOCKED)
--SMT=DISABLED
Relative Performance=38.9 speed=18700 kn/sec
--SMT=AUTO
Relative Performance=49.8 speed=23900 kn/sec
[quote="corres"]
[quote="corres"]
There some data about chess power of Ryzen 7 processor but these are very confused.
Therefore I decided to make tests based on well defined parameters.
My rig has a Ryzen 7 1800x processor with 3700 MHz (DEFAULT) CPU clock speed what is overclocked to 8x4000 MHz.
It has 16 GB RAM with 2133 MHz (DEFAULT) clock speed.
The installed operating systems are Win 7 32bits, Win 7 64bits and Win 10 64bits. For tests I used Win 7 64bits system
The tests what I made as follows.
(Sorry, but I have to brake the text because of uncertain net connection )
[/quote]
TEST 1: Fritz Chess Benchmark
I used benchmark program of Deep Fritz 14.
-SMT (Simultaneous Multithreading, ~HT)=OFF cores=8 (physical cores)
Relative Performance=36.0 speed= 17400 kn/sec
-SMT=AUTO (=ON) cores=16 (8 physical+8 logical)
Relative Performance=46.0 speed=22200 kn/sec
[/quote]
TEST 2: Stockfish speed tests
I used Stockfish Speed Test by Bruce Allenbrand
for Stockfish dev.2017.04.23.
-1 core (Hash=512 MB, Depth=15) SMT=DISABLED speed=2253 Mn/sec
-4 cores (Hash=512 Mb, Depth=16) SMT=DISABLED speed=8584 Mn/sec
-8 cores (Hash=512 MB, Depth=16) SMT=DISABLED speed=16000 Mn/sec
-16 cores (Hash=512 MB,Depth=16) SMT=AUTO (=ENABLED!)
speed=23600 Mn/sec
As you can see the the enhancement of speed is about 8000 Mn/sec
from 4 cores to 8 cores and from 8 cores to 16 cores.
[/quote]
TEST 3: CHESS POWER TESTS
For the users of chess engines is more important the chess power of a CPU than the speed produced by the given CPU.
In the case of one core CPU the speed and power are well correlated to each other. But for a CPU with several cores the connection between speed and power depends on the structure of the given engine, the parameters of the given CPU and the operating system used. You should make a lot of engine-engine tests running one of engine on a PC with Ryzen processor and the other engine running on an Intel processor.
But there is a method to gain relative good estimation on the chess power: To measure the solving time of well defined positions.
For making these tests I compiled a database what contains 70 tactical test positions with one characterized solution. To determine the solving time I used from Fritz tools the "Process Test Sets" tool. I run it five times
and the average was calculated.
[/quote]
(continuation of TEST 3)
Results:
-SMT=DISABLED average solving time=0.42 sec
-SMT=AUTO (=ENABLED) average solving time=0.42 sec
As you can see SMT switch ON or OFF has no effect (or a small effect) on the solving time.
That is while switching SMT ON and OFF the speed changes a lot, the chess power of CPU changes barely.
FOR UNANIMITY, THE ABOVE TESTS WAS MADE ON OVERCLOCKED
(8x4000 MHZ) RYZEN 7 1800x CPU!!!
Because the effect of SMT (and HT too) depend on the power state the given CPU I repeated this test with DEFAULT CPU clock frequency (=3700 MHz)
Results:
-SMT=DISABLED average solving time= 0.52 sec
-SMT=AUTO average solving time=0.43 sec
That is the shortening of solving time is abut 20 %.
So it is obvious that for an overclocked Ryzen CPU you ought to switch of SMT, but a non-overclocked Ryzen CPU you should switch on SMT.
Note:
1. Using SMT the temperature of CPU arises dramatically so you need a very effective cooler!
2,As I mentioned this test only an estimation for chess power of Ryzen 7.
You should make a lot of engine-engine tests to state the exact verdict.
[/quote]
Sorry, but I have to make some corrections.
The text
"TEST 1: Fritz Chess Benchmark"
is incomplete and inaccurate.
The result reported is related to CPU clock frequency= 3700 MHz (DEFAULT)
only.
So the full and correct results are:
-At CPU clock frequency=3700 MHz (DEFAULT)
--SMT=DISABLED
Relative Performance=36.3 speed=17400 kn/sec
--SMT=AUTO (=ENABLED)
Relative Performance=46.0 speed=22200 kn/sec
-At CPU clock frequency=4000 MHz (OVERCLOCKED)
--SMT=DISABLED
Relative Performance=38.9 speed=18700 kn/sec
--SMT=AUTO
Relative Performance=49.8 speed=23900 kn/sec
-
Isaac
- Posts: 265
- Joined: Sat Feb 22, 2014 8:37 pm
Re: Chess power of Ryzen processor
Nice, but as far as I've gathered knowledge on this forum, solving puzzles does not necessarily correlates with strength. This is why in order to see which engine is best, you run full games, not solve some positions only.
Thus, I don't think your sentence
But as you pointed out, one should perform engine vs engine tests to reach a verdict.
Thus, I don't think your sentence
is correct. It would be correct if it started with ''So it is not obvious...''.corres wrote:So it is obvious that for an overclocked Ryzen CPU you ought to switch of SMT, but a non-overclocked Ryzen CPU you should switch on SMT.
But as you pointed out, one should perform engine vs engine tests to reach a verdict.
-
corres
- Posts: 3657
- Joined: Wed Nov 18, 2015 11:41 am
- Location: hungary
Re: Chess power of Ryzen processor
[quote="Isaac"]
Nice, but as far as I've gathered knowledge on this forum, solving puzzles does not necessarily correlates with strength. This is why in order to see which engine is best, you run full games, not solve some positions only.
Thus, I don't think your sentence
[quote="corres"]
So it is obvious that for an overclocked Ryzen CPU you ought to switch of SMT, but a non-overclocked Ryzen CPU you should switch on SMT.
[/quote]
is correct. It would be correct if it started with ''So it is not obvious...''.
But as you pointed out, one should perform engine vs engine tests to reach a verdict
.[/quote]
Sorry, but an "f" misses from my text:
The good text is:
So it is obvious that for an overclocked Ryzen CPU you ought to switch OFF SMT, but a...
I wait for tests made by others but until someone do not prove the opposite facts my sentences based on my experiments is established.
I ought to note that the tests on Intel CPU prove to switch OFF HT (~SMT) for overclocked CPU is a better choice, too.
Nice, but as far as I've gathered knowledge on this forum, solving puzzles does not necessarily correlates with strength. This is why in order to see which engine is best, you run full games, not solve some positions only.
Thus, I don't think your sentence
[quote="corres"]
So it is obvious that for an overclocked Ryzen CPU you ought to switch of SMT, but a non-overclocked Ryzen CPU you should switch on SMT.
[/quote]
is correct. It would be correct if it started with ''So it is not obvious...''.
But as you pointed out, one should perform engine vs engine tests to reach a verdict
.[/quote]
Sorry, but an "f" misses from my text:
The good text is:
So it is obvious that for an overclocked Ryzen CPU you ought to switch OFF SMT, but a...
I wait for tests made by others but until someone do not prove the opposite facts my sentences based on my experiments is established.
I ought to note that the tests on Intel CPU prove to switch OFF HT (~SMT) for overclocked CPU is a better choice, too.
-
Milos
- Posts: 4190
- Joined: Wed Nov 25, 2009 1:47 am
Re: Chess power of Ryzen processor
Robert, with LazySMP time to depth even on selected position (which you effectively measure) is really not a reliable metric.corres wrote:I wait for tests made by others but until someone do not prove the opposite facts my sentences based on my experiments is established.
I ought to note that the tests on Intel CPU prove to switch OFF HT (~SMT) for overclocked CPU is a better choice, too.
Reason is that with LazySMP tree become much wider than deeper. Therefore, you don't find good moves much faster but you avoid bad moves better. So scaling time to depth is much slower than scaling of the real engine strength.
Additional problem with HT is that some of the extra nodes make the tree wider but some are just useless nodes that would anyway be cut if it was a single core search.
-
corres
- Posts: 3657
- Joined: Wed Nov 18, 2015 11:41 am
- Location: hungary
Re: Chess power of Ryzen processor
[quote="Milos"]
[quote="corres"]
I wait for tests made by others but until someone do not prove the opposite facts my sentences based on my experiments is established.
I ought to note that the tests on Intel CPU prove to switch OFF HT (~SMT) for overclocked CPU is a better choice, too.
[/quote]
Robert, with LazySMP time to depth even on selected position (which you effectively measure) is really not a reliable metric.
Reason is that with LazySMP tree become much wider than deeper. Therefore, you don't find good moves much faster but you avoid bad moves better. So scaling time to depth is much slower than scaling of the real engine strength.
Additional problem with HT is that some of the extra nodes make the tree wider but some are just useless nodes that would anyway be cut if it was a single core search.
[/quote]
As you know HT/SMT uses the unexploited time of physical cores.
But when you overclocked a CPU its physical cores have not unexploited time and this is independent from the kind of SMP or tests with full games or tests based on solving of positions.
So in the case of CPU overclocked switching ON SMT/HT does not yield a notable enhancement in the power of CPU, this is the fact
[quote="corres"]
I wait for tests made by others but until someone do not prove the opposite facts my sentences based on my experiments is established.
I ought to note that the tests on Intel CPU prove to switch OFF HT (~SMT) for overclocked CPU is a better choice, too.
[/quote]
Robert, with LazySMP time to depth even on selected position (which you effectively measure) is really not a reliable metric.
Reason is that with LazySMP tree become much wider than deeper. Therefore, you don't find good moves much faster but you avoid bad moves better. So scaling time to depth is much slower than scaling of the real engine strength.
Additional problem with HT is that some of the extra nodes make the tree wider but some are just useless nodes that would anyway be cut if it was a single core search.
[/quote]
As you know HT/SMT uses the unexploited time of physical cores.
But when you overclocked a CPU its physical cores have not unexploited time and this is independent from the kind of SMP or tests with full games or tests based on solving of positions.
So in the case of CPU overclocked switching ON SMT/HT does not yield a notable enhancement in the power of CPU, this is the fact