Testing Ryzen 9 with SMT On vs. SMT Off

Julio Franco

Posts: 8,821   +1,718
Staff member

VitalyT

Posts: 6,032   +6,400
My guess, disabling SMT presents a great TDP relief on the CPU, allowing it to disable Turbo clocking cap, so when Turbo is activated on any core, without SMT, it is truly at full 100% power.

Another possibility, AMD's version of SMT may require non-optimal sync functions with the secondary/hyper core, impeding the physical's core potential.

Or maybe it is a bit of both :)
 
Last edited:

Puiu

Posts: 5,062   +3,923
TechSpot Elite
My guess, disabling SMT presents a great TDP relief on the CPU, allowing it to disable Turbo clocking cap, so when Turbo is activated on any core, without SMT, it is truly at full 100% power.

Another possibility, AMD's version of SMT may require non-optimal sync functions with the secondary/hyper core, impeding the physical's core potential.

Most likely though, it is a bit of both :)
Neah, it doesn't. It just prevents game critical threads from running on the virtual core, which is slower, when the CPU has the headroom to run it on a true physical core. although as we've seem some games like more threads.
 

Maxiking

Posts: 151   +169
I find this obsession funny. Deal with it, they are still slower in gaming clock to clock than the Intel's architecture released in 2015 and crippled by the security patches.

No patch, windows scheduler update or HT off will change that.
 

Puiu

Posts: 5,062   +3,923
TechSpot Elite
I find this obsession funny. Deal with it, they are still slower in gaming clock to clock than the Intel's architecture released in 2015 and crippled by the security patches.

No patch, windows scheduler update or HT off will change that.
does it need to? AMD already offers incredible value for what you are paying for. besides, patches have shown to deliver big gains in the past. even recently, World War Z improved performance for both Intel and AMD.

Nobody will cry about AMD getting an extra 1-2% from a windows or driver update later on if they manage to do it. You just have to buy what fits best in your budget for your gaming PC (most likely the 2600/2600x or the 3600).
 

Amet Monegro

Posts: 75   +24
I find this obsession funny. Deal with it, they are still slower in gaming clock to clock than the Intel's architecture released in 2015 and crippled by the security patches.

No patch, windows scheduler update or HT off will change that.

slower clock to clock? they not using a fixing clock speed, i9 9900K turbos up to 5.00GHz and R9 3900X turbos up to 4.60GHz (9% lower), also games uses a maximun of 6 threads, lower threads in use = higher clock speed

greetings
 

PetrolHead

Posts: 67   +36
Another possibility, AMD's version of SMT may require non-optimal sync functions with the secondary/hyper core, impeding the physical's core potential.

Neah, it doesn't. It just prevents game critical threads from running on the virtual core, which is slower, when the CPU has the headroom to run it on a true physical core. although as we've seem some games like more threads.

Both of these are a bit of an odd way to describe multi-threading. It's not 1 physical core + 1 logical core, it's two logical cores sharing the resources of one physical core. Multi-threading is basically just a way to try and schedule tasks so that the physical CPU core's resources are utilized better. The reason why switching multi-threading off may speed up processing is very, very roughly due to two things: there is a very slight overhead involved in the scheduling and sometimes the tasks are such that there is no way to effectively run two more than one of them on a single core.

It would be interesting to see what the effect in these same games is for the 9900K. Since the processor architectures are different, it's possible that the results might be qualitatively different; in other words, switching SMT off in some games might help in cases where switching HT off would incur a penalty. However, I would expect the conclusion to remain the same: don't switch multi-threading off.
 

ShagnWagn

Posts: 1,297   +1,085
It's confusing why people are so worried about clock-for-clock performance. If AMD is at 2 GHz and Intel was at 2,000 GHz, but AMD still beat Intel, who cares? That's like saying a Honda at 8k RPM should beat a Mustang at 6k RPM. lol

Although these days, heat and efficiency is another consideration. If AMD has to run at 300 degress F to be faster, then I will not use it.
 

phinn888

Posts: 8   +4
Very comprehensive analysis. Based on this I'd say Techstop deaded the issue leaving SMP enabled is the way to go, especially with how badly the 1% FPS tanks in some games like Battlefield V.
 

Maxiking

Posts: 151   +169
I find this obsession funny. Deal with it, they are still slower in gaming clock to clock than the Intel's architecture released in 2015 and crippled by the security patches.

No patch, windows scheduler update or HT off will change that.

slower clock to clock? they not using a fixing clock speed, i9 9900K turbos up to 5.00GHz and R9 3900X turbos up to 4.60GHz (9% lower), also games uses a maximun of 6 threads, lower threads in use = higher clock speed

greetings

3900x and turbo up to 4.6 Ghz, one just can't use it in the same sentence. It is a myth.. Like Trump's racism or transphobia.

There is a test on this site, 9900k vs 3700x + 3900x clocked to 4.0 Ghz. The result 9900k > 3700x, 3900x in gaming. Not like I care but again. These tests are getting tiresome. It is always something after every damn AMD cpus launch. HT, magical updates, bios updates, RAM speeds.. Then it gets tested and busted.
 
Last edited:

JaredTheDragon

Posts: 685   +441
Manually setting core affinity should effectively do the same thing, right? It would be great if there was a tool that could save per application profiles.

ProcessLasso does just this, and effectively replaces Windows TaskMan. In the CGI industry it's a must, as when rendering natively all cores will hit their max. To keep the system responsive (so you can, say, use Photoshop while Maya is rendering) typically we turn off the first two cores for Maya. Controlling affinity is really easy.
 

Maxiking

Posts: 151   +169
I find this obsession funny. Deal with it, they are still slower in gaming clock to clock than the Intel's architecture released in 2015 and crippled by the security patches.

No patch, windows scheduler update or HT off will change that.
does it need to? AMD already offers incredible value for what you are paying for. besides, patches have shown to deliver big gains in the past. even recently, World War Z improved performance for both Intel and AMD.

Nobody will cry about AMD getting an extra 1-2% from a windows or driver update later on if they manage to do it. You just have to buy what fits best in your budget for your gaming PC (most likely the 2600/2600x or the 3600).

If people cared about incredible value, everyone would be driving Prius. So... How many of them have you seen today?
 

Evernessince

Posts: 5,464   +6,149
I find this obsession funny. Deal with it, they are still slower in gaming clock to clock than the Intel's architecture released in 2015 and crippled by the security patches.

No patch, windows scheduler update or HT off will change that.

Incorrect actually. Skylake was released in 2015, of which Zen 2 has already surpassed clock for clock in gaming. Clock for clock Zen 2 wins in productivity by a massive margin as well.

I really don't know if a comment like this is a slightly against Intel or AMD given the fact that Intel added 0 cores over 10 years and the IPC is similar enough to where a 2013 CPU is close to a 2019 CPU. If I compare Zen 1 to Zen 2, it's already made more improvements in 3 years then Intel has in 10. Enough said.

I find this obsession funny. Deal with it, they are still slower in gaming clock to clock than the Intel's architecture released in 2015 and crippled by the security patches.

No patch, windows scheduler update or HT off will change that.

slower clock to clock? they not using a fixing clock speed, i9 9900K turbos up to 5.00GHz and R9 3900X turbos up to 4.60GHz (9% lower), also games uses a maximun of 6 threads, lower threads in use = higher clock speed

greetings

3900x and turbo up to 4.6 Ghz, one just can't use it in the same sentence. It is a myth.. Like Trump's racism or transphobia.

There is a test on this site, 9900k vs 3700x + 3900x clocked to 4.0 Ghz. The result 9900k > 3700x, 3900x in gaming. Not like I care but again. These tests are getting tiresome. It is always something after every damn AMD cpus launch. HT, magical updates, bios updates, RAM speeds.. Then it gets tested and busted.

A majority of reviewers were able to get 4.6 GHz out of the box. Those that didn't were experiencing a BIOS bug. It happens with the launch of new products, only I don't go around pretending they apply to every 3900X. It has already been fixed in a BIOS update.
 

Puiu

Posts: 5,062   +3,923
TechSpot Elite
Both of these are a bit of an odd way to describe multi-threading. It's not 1 physical core + 1 logical core, it's two logical cores sharing the resources of one physical core. Multi-threading is basically just a way to try and schedule tasks so that the physical CPU core's resources are utilized better. The reason why switching multi-threading off may speed up processing is very, very roughly due to two things: there is a very slight overhead involved in the scheduling and sometimes the tasks are such that there is no way to effectively run two more than one of them on a single core.

It would be interesting to see what the effect in these same games is for the 9900K. Since the processor architectures are different, it's possible that the results might be qualitatively different; in other words, switching SMT off in some games might help in cases where switching HT off would incur a penalty. However, I would expect the conclusion to remain the same: don't switch multi-threading off.
The the thread that is put on the SMT created "core" will play second fiddle to the main thread. It will use resources when available.
And yes, some games do get better FPS after disabling HT on Intel, although it seems that it's more pronounced on AMD. I'm just assuming that the threads get better prioritization for Intel on windows and it seems Linux is handling AMD slightly better (prolly better scheduler?)

2. Your conjecture that the 6700K, 7700K, 8700K, and 9900K didn't provide additional IPC is in fact false by a wide margin: http://hwbench.com/cpus/intel-core-i9-9900k-vs-intel-core-i7-6700k.
Are you confusing IPC with single threaded performance? Otherwise I don't see why you would use that link.
 
Last edited:

Evernessince

Posts: 5,464   +6,149
Are you confusing IPC with single threaded performance? Otherwise I don't see why you would use that link.

No, it's just strangely difficult to find anything comparing IPC of the 6700K, 7700K, or 8700K. In fact TechSpot doesn't even have IPC numbers for it's 6700K and 7700K reviews. I did find that GamersNexus link though that does an good job of illustrating the point.
 

axiomatic13

Posts: 299   +254
Anyone else notice that the only time turning SMT off was a negative was running games that use AVX instructions? It makes sense that virtual cores could aide AVX throughput. Of that group, AC:Odyssey exemplified it the most as it leans heavily on AVX.
 

WorthySK1LLZ

Posts: 6   +1
Something cool that a lot of people don't know, is that disabling HyperThreading, usually allows for more OC headroom.
My 8700k, can only clock to 5.2 while staying under 1.6v. With HT off, I can pull off 5.4 while at 1.52v. Also, if your a high fps gamer like myself, turning off HT helps with those scenarios.
It also helps when troubleshooting. You might be gaming, and can't figure out why your GPU and CPU never reach 100%, hence hard to figure out your bottleneck. Usually this is because of HyperThreading. Example, playing a game and your GPU is showing 70% usage and your CPU is showing 50-60% usage. Turn off HT and now your CPU shows 100% but your FPS is the same, if not higher. This means that you CPU is the bottleneck and the game isn't optimized to take advantage of the additional cores/ HT cores.
Usually I get more frames and more consistency by disabling HT for gaming
 

pit1209

Posts: 156   +275
I find this obsession funny. Deal with it, they are still slower in gaming clock to clock than the Intel's architecture released in 2015 and crippled by the security patches.

No patch, windows scheduler update or HT off will change that.
The only thing that I find funny is how Intel hasn't improved one bit since 2015 while feeding us quad cores while just recently they "decided" to make a mess of their line ups and increase the core count in a hurry, I mean what's an i7 series nowadays? looks like an overpriced i5 to me, maybe you could enlighten us about that instead of criticizing the competition.
 
The only site I found that benchmarks SC2. Thanks. I'd like to understand why Dota 2 is not on the list, though, considering how most of the games are naturally GPU bound, while Dota isn't as much.
 

RiCHeeGee

Posts: 6   +4
A majority of reviewers were able to get 4.6 GHz out of the box. Those that didn't were experiencing a BIOS bug. It happens with the launch of new products, only I don't go around pretending they apply to every 3900X. It has already been fixed in a BIOS update.

This isn't what's being experienced out in the wild, despite bios updates to deal with some early quirks. These chips aren't hitting the advertised boost speeds for most and there is little overclocking overhead BUT this isn't to say that they're bad chips, they are unsurpassed in productivity and they're still perfectly good gaming CPUs but Intel does win in that department.

It won't stop me from buying the 3950X when it arrives in September - upgrading from an overclocked 2700 (non-X) at 4.25ghz 1.425V.
 

Evernessince

Posts: 5,464   +6,149
This isn't what's being experienced out in the wild, despite bios updates to deal with some early quirks. These chips aren't hitting the advertised boost speeds for most and there is little overclocking overhead BUT this isn't to say that they're bad chips, they are unsurpassed in productivity and they're still perfectly good gaming CPUs but Intel does win in that department.

It won't stop me from buying the 3950X when it arrives in September - upgrading from an overclocked 2700 (non-X) at 4.25ghz 1.425V.

By out in the wild I assume you are referring to forum posts / reddit. It would make sense if a majority of the people posting there are looking at how to get max boost clocks, after all they came there for help. On the flip side of the coin, people who are getting max boost clocks aren't making posts for help because they don't need help in the first place. You cannot base the current statues of a product based on support threads alone. If you did nearly every product ever launched would be a faulty piece of junk when in fact the number of people posting help threads for many products is a small amount. It should also be noted that as sales increase, number of RMA cases will increase proportionately. A better selling product is bound to have more people that need help.

I've seen a few posts about not reaching max boost clocks but nothing I would call widespread. Given the number of sales they are making based off the few retailers that actually release their numbers, I'd say this is well within expected.
 
Manually setting core affinity should effectively do the same thing, right? It would be great if there was a tool that could save per application profiles.
Hi

There are two applications which allow you to save affinity...

Prio is great it installs then is a new option in windows task manager.
In the details section right-click the process you want to set affinity on and option to save affinity is available too for when relaunched.

Tho I find process lasso much better. Sometimes Prio doesn't do its job! Process lasso is very advanced task manager which does certainly save affinity and works as intended.
 

Puiu

Posts: 5,062   +3,923
TechSpot Elite
No, it's just strangely difficult to find anything comparing IPC of the 6700K, 7700K, or 8700K. In fact TechSpot doesn't even have IPC numbers for it's 6700K and 7700K reviews. I did find that GamersNexus link though that does an good job of illustrating the point.
yep, zero gains over many "generations". we are lucky that they were forced to add more cores or we would be sitting pretty with just 4 cores in 2019.