Intel Core i9-10900 10-core CPU Pictured

[ARF]

Without HT would drastically drop the temperatures and power consumption, and allow much higher clocks. The 9700k's lack of HT allows users to run it at 5.2-5.3ghz at relative ease (compared to 9900k which CAN hit those clocks but it's much more difficult). Most people don't need 16 threads and SMT/HT even hurts in some situations. The 9700k lacks a bit of cache, but with 2 more cores, another 4-8mb of cache and a new process that allows it to hit 5.4-5.5 ghz with decent cooling without causing a small brown out would be really ideal for most gamers.

Turning HT off on the current gen of chips yields pretty dramatic temperature differences and allows for a 200-300mhz higher OC as a result.

View attachment 144846

It's pretty common to see a 9700K at 5.2 beating a 9900K at 5.0 in games/non-heavily threaded loads while consuming less power.

This game doesn't scale beyond 4 or 6 cores at all. At least, it scales with frequency

 

[phanbuey]

This game doesn't scale beyond 4 or 6 cores at all. At least, it scales with frequency

that's true for literally 90% of games this year. even outer worlds or red dead don't scale... and even the game that do scale like the latest battlefield, the 9700k still games slightly faster than a 3900x
think... 8t vs 24t same fps..... on a threaded game...

i mean if you want to sit around and play cinebench on your gaming rig then yes, threads are greatr

 

[Vayra86]

Without HT would drastically drop the temperatures and power consumption, and allow much higher clocks. The 9700k's lack of HT allows users to run it at 5.2-5.3ghz at relative ease (compared to 9900k which CAN hit those clocks but it's much more difficult). Most people don't need 16 threads and SMT/HT even hurts in some situations. The 9700k lacks a bit of cache, but with 2 more cores, another 4-8mb of cache and a new process that allows it to hit 5.4-5.5 ghz with decent cooling without causing a small brown out would be really ideal for most gamers.

Turning HT off on the current gen of chips yields pretty dramatic temperature differences and allows for a 200-300mhz higher OC as a result.

View attachment 144846

It's pretty common to see a 9700K at 5.2 beating a 9900K at 5.0 in games/non-heavily threaded loads while consuming less power.

'Beating' right. But what are we really looking at here. For this game. A full 100mhz yields you exactly less than ONE minimum FPS, and a maximum of TWO average FPS.

And that is with Hitman not being limited in any other way. Ie GPU, or the engine/game code itself, or RAM. Game is also offline, so no latency hit, no tickrate limitations or impact, network load, etc.

The gist of it all is that AMD is now 'close enough' making Intel's overpriced top end completely ridiculous. And if you did buy this for 'high thread counts' then yes, Ryzen all the way because it simply has more of them. The wiggle room that is left for an Intel rig has decreased substantially and the potential gain over a lower priced AMD alternative with often double the thread count, is negligible for gaming. Yes, even for high refresh, these days.

Only if you feel like spending >2500 on your gaming PC, which is a total waste of time, is Intel really going to offer any sort of noticeable impact. Below that, you stand to gain more with a good GPU choice and a CPU that will last longer than any Intel 6- or 8 core /thread CPU. Those 9700's.... obsolete faster than any HT / SMT CPU even if they clock higher. There is just no question about it. So yay, you gain 4 FPS today...

The 3570k proved that non HT CPUs just run into trouble faster when thread count is saturated, while the HT quads had at least a few more years in them. Why repeat this for such a meagre gain? So its really a double edged blade here. Short term gain is at odds with future proofing, really.

Why would it be different? These CPUs are sold to the same clients (maybe putting gamers aside).
It's the same boost-idle-boost-idle cycle.

Actually it's the other way round (Intel vs AMD in expectations). AMD looks great in Cinebench or batch encoding. People buy them, run a few benchmarks, post results on forums - great. And one day they notice that their office laptop boots quicker, opens websites faster and actually is perfectly fine for everything they need. So why did they buy this huge desktop? And how to use 12 cores?

LOL on crunching workloads. How many people here actually do some heavy computing on their uber fast PCs? And I mean concious useful activity, not running benchmarks and distributed computing projects.

Also, you would have to manually limit the CPU to force it to run at those 2.8GHz (which will happen in SFF OEM machines). Leave it alone, provide decent airflow - it'll boost all day long if needed.

You're forgetting this is a top end i9 part we are speaking of.

Average Joe does not consider this, making your entire argument irrelevant. You are definitely looking at performance oriented use cases here with much lower idle times. As you say, most people will not ever see the extra performance - nor will they run Cinebench, or buy a 400>+ dollar CPU, fast RAM and pricy board. Those that do, often are using the extra performance, they do have sustained loads.

 

[cucker tarlson]

that's true for literally 90% of games this year. even outer worlds or red dead don't scale... and even the game that do scale like the latest battlefield, the 9700k still games slightly faster than a 3900x
think... 8t vs 24t same fps..... on a threaded game...

i mean if you want to sit around and play cinebench on your gaming rig then yes, threads are greatr

I'll tell you more - if you look at cpu core load analysis,9700k can handle distributing gaming loads much more evenly than smt cpus.

you will likely see higher peak usage on one of 3900x 24 threads than you'll see on any of 9700k's eight.

 

[ARF]

I'll tell you more - if you look at cpu core load analysis,9700k can handle distributing gaming loads much more evenly than smt cpus.

you will likely see higher peak usage on one of 3900x 24 threads than you'll see on any of 9700k's eight.

What does "more evenly" even mean? You mean more micro-stutter and/or that the games are very buggy and behave in ridiculous way.

 

[cucker tarlson]

What does "more evenly" even mean? You mean more micro-stutter and/or that the games are very buggy and behave in ridiculous way.

grab a dictionary pal.

Metro Exodus: Sam's Story тест GPU/CPU | Action / FPS / TPS | Тест GPU

Главным героем Metro Exodus: Sam's Story является солдат Сэм - американский солдат, который отлично себя чувствовал как

gamegpu.com

peak usage
75% on 9700k
80% on 3900x
81% on 3700x
83% on 9900k

same story in most of other modern games

70% on 9700k vs 83% on 3900x

Wolfenstein: Youngblood тест RTX | Action / FPS / TPS | Тест GPU

Wolfenstein: Youngblood — События игры происходят через 19 лет после событий Wolfenstein II: The New Colossus, в альтернативном 1980 году

gamegpu.com

44% on 9700k vs 64% on 3900x

Red Dead Redemption 2 тест GPU/CPU 2019 | Action / FPS / TPS | Тест GPU

Америка, 1899 год. Эпоха Дикого Запада подходит к концу. Служители закона методично охотятся на остатки банд.

gamegpu.com

84% on 9700k vs 92% on 3900x

Star Wars Jedi: Fallen Order тест GPU/CPU 2019 | Action / FPS / TPS | Тест GPU

В «Звёздные Войны Джедаи: Павший Орден», боевике с видом от третьего лица от Respawn Entertainment, вас ждут приключени

gamegpu.com

 

[ARF]

grab a dictionary pal.

Metro Exodus: Sam's Story тест GPU/CPU | Action / FPS / TPS | Тест GPU

Главным героем Metro Exodus: Sam's Story является солдат Сэм - американский солдат, который отлично себя чувствовал как

gamegpu.com

peak usage
75% on 9700k
80% on 3900x
81% on 3700x
83% on 9900k

same story in most of other modern games

70% on 9700k vs 83% on 3900x

Wolfenstein: Youngblood тест RTX | Action / FPS / TPS | Тест GPU

Wolfenstein: Youngblood — События игры происходят через 19 лет после событий Wolfenstein II: The New Colossus, в альтернативном 1980 году

gamegpu.com

44% on 9700k vs 64% on 3900x

Red Dead Redemption 2 тест GPU/CPU 2019 | Action / FPS / TPS | Тест GPU

Америка, 1899 год. Эпоха Дикого Запада подходит к концу. Служители закона методично охотятся на остатки банд.

gamegpu.com

84% on 9700k vs 92% on 3900x

Star Wars Jedi: Fallen Order тест GPU/CPU 2019 | Action / FPS / TPS | Тест GPU

В «Звёздные Войны Джедаи: Павший Орден», боевике с видом от третьего лица от Respawn Entertainment, вас ждут приключени

gamegpu.com

This is not peak usage! This is just one thread loaded to that figure. You also have threads on the Ryzen 9 3900X with 3%, 6% and 19% load.

 

[cucker tarlson]

This is not peak usage! This is just one thread loaded to that figure.

 

[ARF]

SWJFO:



RDR2:



WY:



MESS:

 

[ppn]

Threads boost performance by 25% but make 1x 4.8Ghz core perform like 2x 3.0Ghz and this is bottlenecking the main or AI thread or introducing stutters and latencies.

 

[EarthDog]

Threads boost performance by 25% but make 1x 4.8Ghz core perform like 2x 3.0Ghz and this is bottlenecking the main or AI thread or introducing stutters and latencies.

If we compare non gaming enchmarks, HT/SMT typically adds about 50% more performance compared to actual cores and threads. So if you have a quad with HT vs an octo without, we'll see the octo win by around 33%. Now, obviously this scales differently depending on the load, but at least with hwbot benchmarks (some of which are IRL loads) that has been the trend.

As far as hitching in gaming... some games improve when disabling HT (assuming it has enough real cores to handle it). Some games it doesnt matter. So to me, those situations are game dependent and the game's/engine issue, not the cpu.

How does it bottleneck a core when it is getting more work done? Your analogy there doesnt make sense to me. If ht goes away, it is still slower... its bottlenecked compared to two real cores, but we arent dealing with that as a comparison.

 

[JackCarver]

The Intel CPUs distribute gaming load better than the Ryzen CPUs, as one can see on those graphs Cucker Tarlson posted. Also the 9900K distributes load better over real Cores and HT Cores.

Wolfenstein Youngblood:
3900X: 83% load on core 2 and 19% load on core 5, lol

Core 2 operates nearly on full load whereas core 5 idles. The Intel distribution is way better here

 

[TheoneandonlyMrK]

The Intel CPUs distribute gaming load better than the Ryzen CPUs, as one can see on those graphs Cucker Tarlson posted. Also the 9900K distributes load better over real Cores and HT Cores.

Wolfenstein Youngblood:
3900X: 83% load on core 2 and 19% load on core 5, lol

Core 2 operates nearly on full load whereas core 5 idles. The Intel distribution is way better here

The Os controls work thread distribution not the CPU ,and each Os release the usage of cores improves.
Note this answer is regardless of brand.

Your inflating your own opinion, I game more on Intel than AMD , I see no in-game difference between the two besides minor differences in max FPS, but nothing worth caring about.

 

[ARF]

The Intel CPUs distribute gaming load better than the Ryzen CPUs, as one can see on those graphs Cucker Tarlson posted. Also the 9900K distributes load better over real Cores and HT Cores.

Wolfenstein Youngblood:
3900X: 83% load on core 2 and 19% load on core 5, lol

Core 2 operates nearly on full load whereas core 5 idles. The Intel distribution is way better here

We shouldn't say that the load distribution is better. AMD has assigned priority to its cores depending on their quality and ability to reach the maximum boost frequency.
If everything normal, the BIOS and chipset driver should inform the OS, so that the OS assigns the load mostly to the best cores.

The load can be intensive or extensive. Extensive is when you measure how much time the thread is loaded, and intensive is when you measure the real load factor at a single point in time.

 

[cucker tarlson]

The Intel CPUs distribute gaming load better than the Ryzen CPUs, as one can see on those graphs Cucker Tarlson posted. Also the 9900K distributes load better over real Cores and HT Cores.

Wolfenstein Youngblood:
3900X: 83% load on core 2 and 19% load on core 5, lol

Core 2 operates nearly on full load whereas core 5 idles. The Intel distribution is way better here

The lower peak load the better.

 

[ARF]

The lower peak load the better.

I don't believe it. Lower thread peak load would mean lower frequency.

And we also don't know that if the OS rapidly switches between the available threads, there won't be conflicts between the apps competing for those threads.

It's all relative and unless you have 100% of the information and the whole picture, you can't claim anything.

 

[neatfeatguy]

Reminds me when people bought i7-2600Ks and turned the HT off..

Isn't the whole idea of buying a hyperthreaded CPU to have more threads?

lol... but this is a bit different. You are paying less for less.

You have to remember that HT years ago actually hindered the performance of some games, so folks did indeed disable HT for that very specific reason.

Things have improved greatly over the years that either Intel's/MS/game developer (or some combination of them all) for writing better software so issues like HT causing issues doesn't seem to come up anymore, at least not that I'm aware of.

 

[JackCarver]

The Os controls work thread distribution not the CPU ,and each Os release the usage of cores improves.
Note this answer is regardless of brand.

Ok but the core load Distribution is definitely curious, especially in Wolfenstein Youngblood. The 3900X has three cores with load 61% to 83% and all others are below 41%. If the game scales for example over 6 cores shouldn't be there at least 6 cores with allmost same load?
The 9900K with HT has a way more evenly load Distribution in comparison.

 

[ARF]

Ok but the core load Distribution is definitely curious, especially in Wolfenstein Youngblood. The 3900X has three cores with load 61% to 83% and all others are below 41%. If the game scales for example over 6 cores shouldn't be there at least 6 cores with allmost same load?

Well, this is a very good direction of thinking. I also think that if the game needs 6 cores, it should assign the best cores and work only with them, thus offloading the rest completely.

But AMD doesn't think so. They have decided that the threads should be switched off rapidly on and off in order to distribute the heat, thus the clock optimally.

 

[cucker tarlson]

I don't believe it. Lower thread peak load would mean lower frequency.

Why?
Is that how ryzen boost works?
Seems kinda pointless having 12 cores and pushing one to 90% while five sit at 5%.
Not really the point of buying a high core count CPU

 

[JackCarver]

Seems kinda pointless having 12 cores and pushing one to 90% while five sit at 5%.

That's definitely my thinking too.

 

[TheoneandonlyMrK]

Ok but the core load Distribution is definitely curious, especially in Wolfenstein Youngblood. The 3900X has three cores with load 61% to 83% and all others are below 41%. If the game scales for example over 6 cores shouldn't be there at least 6 cores with allmost same load?
The 9900K with HT has a way more evenly load Distribution in comparison.

And.

So that's an issue, you can't have. Three heavier loaded threads with superfluous activity distributed across the rest?.

What's curious to me is people's theory's of how processors work these days and why intel are better.

That boost algorithm which both use ,designed per SKU also has some say.

Some of the most vocal against multi core and. Ryzen in general are he'll bent on keeping their four core hyperthreaded chips relevant via downplaying multi core ryzens.

Let's see where that gets them.

It's the Aliens, simple.

 

[ARF]

Why?
Is that how ryzen boost works?
Seems kinda pointless having 12 cores and pushing one to 90% while five sit at 5%.
Not really the point of buying a high core count CPU

The maximum performance always comes at 100% load.
In Ryzen, there are CCX, these a quad-core complexes which share cash. When you rapidly switch load on and off to them, you add quite a bit of latency, thus your overall performance decreases.

It's in your best interest to use the cores in the same CCX and not switch to the the other CCX for thread off/onloading.

I mean - yeah, it's better to use only as many threads as your app needs, and leave the rest alone, for other apps.

Of course, when your CPU is at 100% full load, then this rapid switching becomes 100% pointless.


The same could be said for videocards. There are different driver versions which might load the shaders up to 95% or 99%, of course with 99% load you will get a few more frames/second.

 

[cucker tarlson]

The maximum performance always comes at 100% load.
In Ryzen, there are CCX, these a quad-core complexes which share cash. When you rapidly switch load on and off to them, you add quite a bit of latency, thus your overall performance decreases.

It's in your best interest to use the cores in the same CCX and not switch to the the other CCX for thread off/onloading.

I mean - yeah, it's better to use only as many threads as your app needs, and leave the rest alone, for other apps.

Of course, when your CPU is at 100% full load, then this rapid switching becomes 100% pointless.


The same could be said for videocards. There are different driver versions which might load the shaders up to 95% or 99%, of course with 99% load you will get a few more frames/second.

4 core ccx hold high core count cpus at times from better utilization.
All on all it seems to benefit ryzens performance,this and cache size.
100% load is never a good thing for games though.

 

[heflys20]

It's the Aliens, simple.

Reminds me of why I stopped participating in gaming/pc forums for the most part. I got dragged into that "brand-centric" lifestyle. Made me stop caring, particularly with how fast-paced this industry is. Performance vs $ is the name these days. Someone told me yesterday that my thought process was equivalent to driving a Civic vs his Bugatti, computer wise. He's right, I couldn't imagine the stress of owning a Bugatti or the money required.