buying a new computer

[Daniel Shawul]

I am waiting to buy a new desktop myself. AMD is going to release its next generation ryzen 9 cpus with 12-core (24 threads) tomorrow.
I think it will be very good hardware for stockfish at the half price of intel's cpus (500$ compared to 1100$). There 16-core version will not come out until septmember though.

If you are also thinking of buying a gpu, NVIDIA is going to upgrade its rtx lineup to their "super" versions on july 9 for the same price.
A 15% improvement on performance but price remains the same.

There will be deals on cpu and gpus because of this so look out for those.

Daniel

Hi Daniel,

It's been awhile since I've posted here, but I've built a few systems and I think you'll find the experience rewarding. Just take
your time and it will work out. I'm glad you're considering the new Ryzen 3000 CPUs, (full disclaimer I'm part of the AMD Ryzen
design team--I'm an analog circuit designer). There's a lot of me in these new CPUs, so I hope they live up to your and everyone's
else expectations.

best,
--tom

Hi Tom,
I am really excited about it after seeing a couple of benchmarks. Also it is definitely the best cpu interms of performance per dollar
so there was no question for me which cpu I was going to get. You should be proud of your work !!

I would have gotten the navi GPUs too had there been optimized OpenCL deep learning libraries. Caffee and maybe Tensorflow
have experimental versions, but i have not had the chance to try them out yet.

regards,
Daniel

[Daniel Shawul]

I am going to try and build the pc myself -- never built one before but watching lots of youtube videos has given me confidence. Fingers crossed I don't toast one of the components since i seem to be highly charged.

The 3900x was out of stock yesterday but managed to buy one today on newegg. Based on pc part picker i need to spend atleast 1500$ before tax if I pair it with an RTX 2070 super. I am going for a cheap X570 motherboard but maybe i should go for a solid X470 without PCIe 4.0 since i probably will not have any use for the extra SSD speed. 16GB 3200 ram.

Daniel

One thing: 32 GB of RAM. You are doing many experiments and maybe don't need long runs like when one is analyzing, but keep in mind that on your GPU and CPU, something like Leela MCTS on infinite mode will fill up the free part of 16 GB RAM (say 12-13 GB) in less than 15-20 minutes. That's an issue, unlike that with the traditional engines. Also, if your OS cache is on HDD, after such runs, other applications will almost freeze.

I have had the worst mcts tree implementation literally until a week ago where it was roughly consuming 30x more RAM than leela's. It could only go for 5 min with 12G of RAM and that was not good enough for TCEC 30+5 tc. It now consumes 30x less so i expect it to go much longer. I chose 16G because TCEC gpu machines uses it but yeah I could add RAM to the remaining 2 slots to make it 32Gb.

[Zenmastur]

Yes, I speculated on the issue that 3000 series 32 core Threadripper is rumored to fix the issue, and then, not having a clear budget, might be a "best buy". But such a system will be in $5000 range, if coupled with other powerful hardware components. There are "best buys" and "hard to get even better buys", and now that Uri wrote about $2500 budget, this one is ruled out.

…..

There is much speculation in all this. Are you sure that "the coolest running and fastest chiplets will be found on the 3950X"?

No I don't think the 3950X will be getting the best chiplets! Those will be reserved for Epyc CPU's since they are higher priced SKU's with much greater margins. But, I am sure the 3950X will be getting better chiplets than the lower tiered CPU's. The question is: How much better will they be?

Are you sure that 3950X with same reasonable cooling can match 3900X on average clocks on all running cores?

I doubt that will be able to, but your assumption that it will be limited to it's base clock seems absurd. I've seen benchmarks where all cores boosted to over 4Ghz for the entire duration of the test which was in excess of 10 minutes. So, I think it's a fair guess that the 3950X won't be limited to just it's base clock while running an AB chess engine. Since there is an easy way to find out what the truth of the matter is, I suggest it's smarter to wait if you need/want the extra computing power and then there will be no need to speculate!

Nobody knows. Also, L2 + L3 cache per core in 3900X is about 30% higher than that per core in 3950X (that is not a speculation), and for a cache intensive software like Stockfish, it does matter. All in all, I would bet that per core performance with SF on all cores will be higher in 3900X versus 3950X, having similar cooling and temperatures. And the reasons to spend $250 more on 3950X compared to 3900X are debatable ($150 seems to be a fairer price addition).

If it were the case that the price for the entire system was $500 for a 3900X and $750 for a 3950X I might agree with you. But that's only the case if you already own a system and it will accept a 3900X or 3950X as an upgrade with the rest of the system unchanged. If you are building it from scratch and the cost is $2000 for a 3900X system and $2250 for an identically configured 3950X system then your analysis of the true value of a 3950X seems flawed. In this example it will need to boost the NPS by 12.5%. If it can't achieve that then I suggest they won't be selling many of them in the chess community. I have good reason to believe it will exceed a 12.5% increase in NPS by a wide margin. So, I'm willing to wait and see.

A good chunk of applications, Leela included, depends more on 1-2 core performance than that on all cores. Yes, one can go back and boost those very fast 1-2 cores on 3950X, keeping the other on base clock, but it's not what we are usually doing when mixing up full CPU and 1-2 core performances. Let's see, I am speculating a bit too.

There is no need to go back and boost anything on the 3950X as precision boost will do that on the fly for you and it will do two cores to it max boost clock and single cores to more than the boost clock for short periods. So you don't have to worry about such things, if set it up properly the MB/CPU will do all that for you. And then when it sees a workload that needs all cores it will automatically drop the clocks to maintain the power target as long as it doesn't bump the thermal limits or CPU core voltage limits. I think you're trying to make it more complicated than it is.

Regards,

Zenmastur

[smatovic]

Someone ever considered a refurbished workstation? I got a HP Z620, I7, 2.8-3.5Ghz, 20 cores, 40 threads, 16Gb ram, 500Gb SSD, 2 x 1 TB harddisk for under 1000 euro with 2 year guarantee.

...unless you have your own power plant and running 24/7, I guess it pays off to
invest in CPUs/GPUs based on smaller fabrication processes that consume less
watts.

***edit***
resp. have more instruction throughput per watt

--
Srdja

"Own power plant" for 100 extra W, seriously??? You can run it for a full month non-stop to accumulate the same amount of energy as one charge of Tesla Model 3 car.

"Own power plant"....it is not uncommon that people have solar plants on their roof, I know some gauchos who own windmills.

Let us assume 30 Euro-Cents per kWh, and running the hardware for 4 years 24/7, makes 1051.4 Euros for additional 100 watt.

But my point was the switch to a smaller fabrication process...

Code: Select all

Year   GPU          Fab          Power    TFLOPs    price
2014   GTX 980TI    TMSC-32nm    250 W    5.6       200 €  on eBay
2018   RTX 2080TI   TMSC-12nm    250 W    11.7      1200 € new

365 days * 24 hours * 0,250 kW * 0,30 Euros * 4 years makes 2628 Euros.

So which GPU would you prefer to buy to run 24/7 in a workstation or data-center?

Somehow I doubt that you would you buy six GTX 980TI for the price of one RTX 2080TI....

--
Srda

[Laskos]

Yes, I speculated on the issue that 3000 series 32 core Threadripper is rumored to fix the issue, and then, not having a clear budget, might be a "best buy". But such a system will be in $5000 range, if coupled with other powerful hardware components. There are "best buys" and "hard to get even better buys", and now that Uri wrote about $2500 budget, this one is ruled out.

…..

There is much speculation in all this. Are you sure that "the coolest running and fastest chiplets will be found on the 3950X"?

No I don't think the 3950X will be getting the best chiplets! Those will be reserved for Epyc CPU's since they are higher priced SKU's with much greater margins. But, I am sure the 3950X will be getting better chiplets than the lower tiered CPU's. The question is: How much better will they be?

Are you sure that 3950X with same reasonable cooling can match 3900X on average clocks on all running cores?

I doubt that will be able to, but your assumption that it will be limited to it's base clock seems absurd. I've seen benchmarks where all cores boosted to over 4Ghz for the entire duration of the test which was in excess of 10 minutes. So, I think it's a fair guess that the 3950X won't be limited to just it's base clock while running an AB chess engine. Since there is an easy way to find out what the truth of the matter is, I suggest it's smarter to wait if you need/want the extra computing power and then there will be no need to speculate!

I have never said that "it will be limited to its base clock". I just said that it will hardly match with the same cooling to the same temps the speeds of all cores average 3900X. Both will be boosted somewhat on all cores, but the plausible assumption would be that 3950X won't overcome the 300MHz gap towards 3900X. I even proposed that it will shrink the gap to 200MHz, having a better Boost numbers, but it will be still below 3900X. Don't misrepresent what I said. But let's see, who knows, maybe 3950X OC-es much better and you are right? I wouldn't bet on that, though.

Nobody knows. Also, L2 + L3 cache per core in 3900X is about 30% higher than that per core in 3950X (that is not a speculation), and for a cache intensive software like Stockfish, it does matter. All in all, I would bet that per core performance with SF on all cores will be higher in 3900X versus 3950X, having similar cooling and temperatures. And the reasons to spend $250 more on 3950X compared to 3900X are debatable ($150 seems to be a fairer price addition).

If it were the case that the price for the entire system was $500 for a 3900X and $750 for a 3950X I might agree with you. But that's only the case if you already own a system and it will accept a 3900X or 3950X as an upgrade with the rest of the system unchanged. If you are building it from scratch and the cost is $2000 for a 3900X system and $2250 for an identically configured 3950X system then your analysis of the true value of a 3950X seems flawed. In this example it will need to boost the NPS by 12.5%. If it can't achieve that then I suggest they won't be selling many of them in the chess community. I have good reason to believe it will exceed a 12.5% increase in NPS by a wide margin. So, I'm willing to wait and see.

Again, can you read my posts properly? I said that all cores SF NPS will be higher on 3950X compared to 3900X by maybe 20-25%, and that would be higher than 12.5%. But I also said that Leela and the single core applications might be slower in the range of 5-10% on 3950X compared to 3900X (due to the cache size per core too), which might offset somewhat the SF advantage. All in all, it is not clear whether $250 are worth paying. For example, for video-gamers almost surely not. Then again, cooling might come $50 more expensive with 3950X, although you dismiss that. Let's see.

A good chunk of applications, Leela included, depends more on 1-2 core performance than that on all cores. Yes, one can go back and boost those very fast 1-2 cores on 3950X, keeping the other on base clock, but it's not what we are usually doing when mixing up full CPU and 1-2 core performances. Let's see, I am speculating a bit too.

There is no need to go back and boost anything on the 3950X as precision boost will do that on the fly for you and it will do two cores to it max boost clock and single cores to more than the boost clock for short periods. So you don't have to worry about such things, if set it up properly the MB/CPU will do all that for you. And then when it sees a workload that needs all cores it will automatically drop the clocks to maintain the power target as long as it doesn't bump the thermal limits or CPU core voltage limits. I think you're trying to make it more complicated than it is.

Regards,

Zenmastur

I have never owned a Ryzen, so I don't know here. I have an Intel i7, I disabled the Turbo Boost and then OC-ed significantly all the cores. That's simple and stable, your solution seems to be too much "flexible", so I am not sure I would use such "dynamic solutions". I need stability and consistency, not some spikes of a half-GHz.

[corres]

The AMD Ryzen series have a serious drawback: They can not utilize the benefit of bmi2 supply.
I am afraid it will be in the case of Ryzen 9 processors too.

[Dann Corbit]

The AMD Ryzen series have a serious drawback: They can not utilize the benefit of bmi2 supply.
I am afraid it will be in the case of Ryzen 9 processors too.

It's irrelevant.
Just build for SSE type support.
The BMI builds do not run fast.
Run the other build.
You still get better nodes per dollar on Ryzen than anything else.

To your point:
102,352,280 for 2x AMD EPYC 7601 Retail with 128threads running the BMI build
but
127,342,246 NPS for 2x AMD EPYC 7601 Retail with 128threads using the pop build
24% faster

So where is the tragedy here?

[ChiefPushesWood]

So where is the tragedy here?

Easy there, Dann. Let's not let facts cloud the conversation!

Chief

[Dann Corbit]

Someone ever considered a refurbished workstation? I got a HP Z620, I7, 2.8-3.5Ghz, 20 cores, 40 threads, 16Gb ram, 500Gb SSD, 2 x 1 TB harddisk for under 1000 euro with 2 year guarantee.

...unless you have your own power plant and running 24/7, I guess it pays off to
invest in CPUs/GPUs based on smaller fabrication processes that consume less
watts.

***edit***
resp. have more instruction throughput per watt

--
Srdja

"Own power plant" for 100 extra W, seriously??? You can run it for a full month non-stop to accumulate the same amount of energy as one charge of Tesla Model 3 car.

"Own power plant"....it is not uncommon that people have solar plants on their roof, I know some gauchos who own windmills.

Let us assume 30 Euro-Cents per kWh, and running the hardware for 4 years 24/7, makes 1051.4 Euros for additional 100 watt.

But my point was the switch to a smaller fabrication process...

Code: Select all

Year   GPU          Fab          Power    TFLOPs    price
2014   GTX 980TI    TMSC-32nm    250 W    5.6       200 €  on eBay
2018   RTX 2080TI   TMSC-12nm    250 W    11.7      1200 € new

365 days * 24 hours * 0,250 kW * 0,30 Euros * 4 years makes 2628 Euros.

So which GPU would you prefer to buy to run 24/7 in a workstation or data-center?

Somehow I doubt that you would you buy six GTX 980TI for the price of one RTX 2080TI....

--
Srda

My computer farm consumes up to 10KW (with everything going full tilt). That's a dollar per hour in the pacific northwest.
I generally run at about half to a third of that, and I have to shut down all of the heavy hitters in the summer (mid-July to mid-September, with occasional shutdowns whenever the temp goes over 90).
I constantly get emails from the power company, asking if I would like help learning how to conserve electricity.
My electric bill has been well over $600 a time or two, and I guess it would have been cheaper in the long run to buy modern equipment.
But I have no complaints. I have a giant rack of stuff in my downstairs that turns electric current into chess analysis. What's not to love?

[Albert Silver]

I think to buy a new computer and I wonder what is the elo of stockfish and lc0 for different prices of hardware.

Is there a table for comparison that give price and elo estimate for both?

Me too - but you want to wait for this to come out in the fall..

https://browser.geekbench.com/v4/cpu/13495867 AMD Myrtle is the project name For the AMD 16 Core Ryzen 9.

The CPU will only be $799, so this will clearly be the best bang for your hard earned cash.

I picked up a 16-core Threadripper last November for $400.