AMD vs Intel Single threading?

[~>Dg<~]

amd current line up is basically the same as a intel 8700 non k x how many cores it has. so a 9700 is faster. its that simple. also a amd 4600af or 2600 is not the same performance as a 5820k or 6800 i7. this is already proven shown. especially once overclocked both of them overclocked are about the same as a stock 8700k or amd 3600. which none of the amd cpus mentioned are.

new intel chips coming are bout 5-10 percent quicker than the 9700k and 9900 so they already ahead of anything amd out now. so add a extra 10 percent on top. so the reality is AMD 4000 chips with 15-20 percent on top are the real cpus worth waiting for. hopefully so. nothing wrong with AMD 3000 cpus brilliant bang for buck for your money. if the 4000s are 15 -20 percent quicker we have some very good times ahead.

 

[Robert896r1]

You're trolling again ^^^^ i din't Insult @Robert896r1



Some people are running them higher all core than they boost single core, it depends.

As for IPC. "Ipc is not the same as single core performance" is a meaningless retort, If you have a given IPC on one core that IPC is the same in each core, if you're measuring core IPC its best to do it on a single core, for Example a 16 thread Zen 2 scales higher in performance than a 16 thread Coffeelake vs what they are on a single core, this is down to better SMT scaling on Zen 2.

I'm hard pressed to find evidence for your statement about this SMT scaling advantage. To demonstrate, I just ran R20 SC and MC on my 9900k at my daily profile:



When I compare that to the R20 thread here according to your claim, the Ryzen Cpu's should be flexing their SMT scaling but that's not the case. I'm ahead in Single score and I'm ahead in multiple core. At 16 threads there's no "SMT scaling" magic or anything like that that puts me at a disadvantage or lets the Ryzen CPU's catch up.

When I look at TNA's score of 544 it's done at 4.675 and let's say that's even with me on SC. https://forums.overclockers.co.uk/posts/33152922
For MC, nghtmare's score of 5365 is at 4.45ghz. https://forums.overclockers.co.uk/posts/32985645 with R20 scaling, he'd need TNA's frequency to match me in Multi so again, I don't see any SMT advantage. If there was a scaling advantage it would show up here.

If you have synthetics that demonstrate this SMT advantage, happy to run them. I can tell you now that something like TimeSpy which also scales well with threads is a massive advantage for me also over any Ryzen 16 thread CPU.

We're on a site called overclockers so what someone does with limited knowledge or experience isn't my concern. I'm on AIO that costs the same as a D15 so we can take exotic cooling etc out of the discussion also. I'm hoping you stay on the mainstream platform for discussion and don't start jumping around randomly to other platforms.

 

[Journey]

new intel chips coming are bout 5-10 percent quicker than the 9700k and 9900

There are no Intel chips coming with faster single thread performance this year.

 

[Perfect_Chaos]

amd current line up is basically the same as a intel 8700 non k x how many cores it has. so a 9700 is faster. its that simple. also a amd 4600af or 2600 is not the same performance as a 5820k or 6800 i7. this is already proven shown. especially once overclocked both of them overclocked are about the same as a stock 8700k or amd 3600. which none of the amd cpus mentioned are.

new intel chips coming are bout 5-10 percent quicker than the 9700k and 9900 so they already ahead of anything amd out now. so add a extra 10 percent on top. so the reality is AMD 4000 chips with 15-20 percent on top are the real cpus worth waiting for. hopefully so. nothing wrong with AMD 3000 cpus brilliant bang for buck for your money. if the 4000s are 15 -20 percent quicker we have some very good times ahead.

It's latency holding them back in games, at least that's what some like HU or Techspot are saying.

There are no Intel chips coming with faster single thread performance this year.

yer i thought these new chips didn't have any IPC increase, just a couple more cores on top and HT on the rest.

 

[humbug]

I'm hard pressed to find evidence for your statement about this SMT scaling advantage. To demonstrate, I just ran R20 SC and MC on my 9900k at my daily profile:



When I compare that to the R20 thread here according to your claim, the Ryzen Cpu's should be flexing their SMT scaling but that's not the case. I'm ahead in Single score and I'm ahead in multiple core. At 16 threads there's no "SMT scaling" magic or anything like that that puts me at a disadvantage or lets the Ryzen CPU's catch up.

When I look at TNA's score of 544 it's done at 4.675 and let's say that's even with me on SC. https://forums.overclockers.co.uk/posts/33152922
For MC, nghtmare's score of 5365 is at 4.45ghz. https://forums.overclockers.co.uk/posts/32985645 with R20 scaling, he'd need TNA's frequency to match me in Multi so again, I don't see any SMT advantage. If there was a scaling advantage it would show up here.

If you have synthetics that demonstrate this SMT advantage, happy to run them. I can tell you now that something like TimeSpy which also scales well with threads is a massive advantage for me also over any Ryzen 16 thread CPU.

We're on a site called overclockers so what someone does with limited knowledge or experience isn't my concern. I'm on AIO that costs the same as a D15 so we can take exotic cooling etc out of the discussion also. I'm hoping you stay on the mainstream platform for discussion and don't start jumping around randomly to other platforms.

Granted on SMT.

However, there is a lot of subject changing going on in this conversation......

Do you or do you not agree with:

Clocking the pins off your CPU's to keep ahead of the curve is not the way to go, as AMD learned.

Efficient high IPC CPU's is better than high clocked CPU's with very high power consumption needing high end cooling.

?

Fundamentally that's my argument, I'm not talking about you, you are obviously quite happy to pay What? £500 for the CPU, £250 for the Motherboard? £140 for the RAM? £100 for the cooler? so £1000? to get 110% of the performance that i'm getting for £350 and i don't sit sweating buckets in the summer because my CPU is kicking out about 250 Watts + of heat.

As a business model can Intel just keep on cranking the clocks up? because the next time AMD step it up even that may no longer be enough, and then what do Intel have, other than about one in a Thousand people who just love the Intel brand so much or hate the AMD brand so much they will buy Intel no matter what.

 

[MartinPrince]

i din't Insult @Robert896r1

You must have a tenuous grasp on reality if you can't see that saying to Robert896r1 "You're beginning to sound like a Bulldozer apologists." is not intended as an insult. Just by adding '' to it does not magically hide that fact.

My 3600 does a better job in Lightroom than his 9700K....!

To have this notion in the first place is not a problem. We all get things wrong, misunderstand or believe things we read or hear. The problems start when someone is presented with irrefutable evidence to the contrary and still tries to maintain and support those daft beliefs.

Charts are hard data, they trump anecdotal claims.

This is just such an unbelievably daft response coming from you and really shows you up.

If I run a test and then tell people about the result then that = Anecdotal.
If several people run the same test and share their results then that = Empirical.

Intel [email protected] 15secs Robert896r1
Intel [email protected] 15secs MartinPrince
ThreadRipper 3960X 18secs amigafan2003
Ryzen 3900X 18secs MartinPrince
Threadripper 2920X 22secs sandys
Ryzen 3600 24secs humbug
Intel [email protected] 24secs MartinPrince
Ryzen 2700@4Ghz 25secs CAT-THE-FIFTH(mate)
Ryzen 2600 27secs CAT-THE-FIFTH
Intel 4790s 38secs MartinPrince
Intel [email protected] 44secs MartinPrince

To the contrary of your statement your CPU did the test in 24secs compared to my 15secs. That's a 37% difference.

What I think most people consider to be a normal, mature and reasoned response is for someone to try and work out where/why they got something so badly wrong.

Where as you go back to your non sequitur responses of 'IPC' this, 'Ryzen all core clock' that and 'Intel needs to run at 5Ghz to catch up...' the other.

You are the first to start calling people 'Trolls' and 'Apologists' without realising that this is the typical case of the pot calling the kettle black and you are in fact the fanatical 'Troll' and the 'Apologist'.

 

[humbug]

You must have a tenuous grasp on reality if you can't see that saying to Robert896r1 "You're beginning to sound like a Bulldozer apologists." is not intended as an insult. Just by adding '' to it does not magically hide that fact.

To have this notion in the first place is not a problem. We all get things wrong, misunderstand or believe things we read or hear. The problems start when someone is presented with irrefutable evidence to the contrary and still tries to maintain and support those daft beliefs.

This is just such an unbelievably daft statement coming from you and really shows you up.

If I run a test and then tell people about the result then that = Anecdotal.
If several people run the same test and share their results then that = Empirical.

Intel [email protected] 15secs Robert896r1
Intel [email protected] 15secs MartinPrince
ThreadRipper 3960X 18secs amigafan2003
Ryzen 3900X 18secs MartinPrince
Threadripper 2920X 22secs sandys
Ryzen 3600 24secs humbug
Intel [email protected] 24secs MartinPrince
Ryzen 2700@4Ghz 25secs CAT-THE-FIFTH(mate)
Ryzen 2600 27secs CAT-THE-FIFTH
Intel 4790s 38secs MartinPrince
Intel [email protected] 44secs MartinPrince

To the contrary of you statements your CPU did the test in 24secs compared to my 15secs. That's a 37% difference.

What I think most people consider to be a normal, mature and reasoned response is for someone to try and work out where/why they got something so badly wrong.

Where as you go back to your non sequitur responses of 'IPC' this, 'Ryzen all core clock' that and 'Intel needs to run at 5Ghz to catch up...' the other.

You are the first to start calling people 'Trolls' and 'Apologists' without realising that this is the typical case of the pot calling the kettle black and you are in fact the fanatical 'Troll' and the 'Apologist'.

I remember this, this is a different benchmark variation to the slides that i posted. I'm talking about one thing and you're calling me out as if its the same thing. This was your own custom file.

You really do think i have "Learning Difficulties" Just stop...

 

[humbug]

PS: if @Robert896r1 feels he's insulted by my likening one of his comment to that of a Bulldozer apologist i'm pretty sure he would say so, he has his own Keyboard.
No need for you to be offended on his behalf for some stolen victimhood currency...

 

[Robert896r1]

"Efficient high IPC CPU's is better than high clocked CPU's with very high power consumption needing high end cooling."

Yes as long as the high IPC can generate enough frequency and sustain said frequency to remain competitive.

OC'ing is never power efficient regardless of node or chip. Take a few mins when you can and watch this as it'll save me typing! https://youtu.be/3WyfJyl9yfY?t=688

Fundamentally that's my argument, I'm not talking about you, you are obviously quite happy to pay What? £500 for the CPU, £250 for the Motherboard? £140 for the RAM? £100 for the cooler? so £1000? to get 110% of the performance that i'm getting for £350 and i don't sit sweating buckets in the summer because my CPU is kicking out about 250 Watts + of heat.

You buying a 400hp BMW for 60k does not mean a Ferrari at 240k should have 1600hp or it's a failed product. If that was the case, there would be no high end products for consumers. In hindsight the 9900k was a great move as it’s rare to have a cpu remain at the top for almost 2 years. As you know, flagship cpu's cost more due to lower yields which leads to a lower supply thus again negating some linear relation between price and performance to lower sku's.

The second argument is 250w. If we take a normal use case for a mainstream cpu such as gaming, rendering, productivity apps, you're hard pressed to get north of 200w. 50w of additional cooling is a lot given the thermal density of the chip. We're also talking an OC'd chip now which needs to respected based on the info in the video above. Now, If all you do is P95 AVX2 Small or major scientific calculations that hammer avx2 workloads continuously, then you have made a poor purchase, didn't research enough and should be looking at HEDT or server platforms as they are specialized for these workloads. Would it be great if the chip ran cooler and didn't need as much power? Ofcourse it would! but it is what it is. Thankfully it can produce results unlike the HEDT lineup which gets thoroughly trounced at this stage.

As a business model can Intel just keep on cranking the clocks up? because the next time AMD step it up even that may no longer be enough, and then what do Intel have, other than about one in a Thousand people who just love the Intel brand so much or hate the AMD brand so much they will buy Intel no matter what.

Intel has and had no intent to ride out 14nm and their node scaling failures are well documented. However, they've done well to mature a platform given the limitations. They're in a weird spot as it's hard to recommend their current lineup and Comet Lake is a dead end as it'll surely be short lived along with Z490. AMD on the other hand still needs bit of maturation with their bios/AGESA and their silicon seems very sensitive to voltage and prone to degrading if pushed as an OC enthusiast might.

I'd personally say we're in a weird spot in the buying cycle.

 

[Robert896r1]

On a diff note, if I got offended by strangers I'd get off the internet. These are just conversations. Take em or leave em.

 

[humbug]

Yes as long as the high IPC can generate enough frequency and sustain said frequency to remain competitive.

Define competitive? or better yet define how Zen 2 is not competitive.

OC'ing is never power efficient regardless of node or chip. Take a few mins when you can and watch this as it'll save me typing! https://youtu.be/3WyfJyl9yfY?t=688

I'm subbed to his channel, i have seen that. i agree but again i'm not sure what you're driving at with that.

You buying a 400hp BMW for 60k does not mean a Ferrari at 240k should have 1600hp or it's a failed product. If that was the case, there would be no high end products for consumers. In hindsight the 9900k was a great move as it’s rare to have a cpu remain at the top for almost 2 years. As you know, flagship cpu's cost more due to lower yields which leads to a lower supply thus again negating some linear relation between price and performance to lower sku's.

Driving a Ferrari is a very different experience to driving a BMW, it not just about performance, its about the way the car handles, about how it feels, about its cosmetics, Car Analogies for CPU's don't work. A CPU is a component with in a system, it serves a function, it is not in its self an experience, think: put the Ferrari engine in the BMW.

But to drill down into your point, the 9900K is a bit faster in most games than a 3900X, but that really is it, its the only claim it can make, that makes it niche, much like the Ferrari, but that is where the similarities end, it doesn't offer an experience in its self and with that is less Prancing Horse and more One Trick Pony.

The second argument is 250w. If we take a normal use case for a mainstream cpu such as gaming, rendering, productivity apps, you're hard pressed to get north of 200w. 50w of additional cooling is a lot given the thermal density of the chip. We're also talking an OC'd chip now which needs to respected based on the info in the video above. Now, If all you do is P95 AVX2 Small or major scientific calculations that hammer avx2 workloads continuously, then you have made a poor purchase, didn't research enough and should be looking at HEDT or server platforms as they are specialized for these workloads. Would it be great if the chip ran cooler and didn't need as much power? Ofcourse it would! but it is what it is. Thankfully it can produce results unlike the HEDT lineup which gets thoroughly trounced at this stage.

It uses an awful lot of power for the performance that you're getting, is my point.

Intel has and had no intent to ride out 14nm and their node scaling failures are well documented. However, they've done well to mature a platform given the limitations. They're in a weird spot as it's hard to recommend their current lineup and Comet Lake is a dead end as it'll surely be short lived along with Z490. AMD on the other hand still needs bit of maturation with their bios/AGESA and their silicon seems very sensitive to voltage and prone to degrading if pushed as an OC enthusiast might.

AMD have not sat on the same architecture for 10+ years, its started with Zen, then Zen+ was a small tweak on Zen and Zen 2 is again quite different, it seem to me AMD isn't keeping an architecture around long enough for it to mature as much as Intel have, i would argue this is Intel's problem, not AMD's.

 

[Phil2008]

You buying a 400hp BMW for 60k does not mean a Ferrari at 240k should have 1600hp or it's a failed product. If that was the case, there would be no high end products for consumers. In hindsight the 9900k was a great move as it’s rare to have a cpu remain at the top for almost 2 years. As you know, flagship cpu's cost more due to lower yields which leads to a lower supply thus again negating some linear relation between price and performance to lower sku's.

The second argument is 250w. If we take a normal use case for a mainstream cpu such as gaming, rendering, productivity apps, you're hard pressed to get north of 200w. 50w of additional cooling is a lot given the thermal density of the chip. We're also talking an OC'd chip now which needs to respected based on the info in the video above. Now, If all you do is P95 AVX2 Small or major scientific calculations that hammer avx2 workloads continuously, then you have made a poor purchase, didn't research enough and should be looking at HEDT or server platforms as they are specialized for these workloads. Would it be great if the chip ran cooler and didn't need as much power? Ofcourse it would! but it is what it is. Thankfully it can produce results unlike the HEDT lineup which gets thoroughly trounced at this stage.

Thats the thing apps or games will only use so many cores,, so for example a 20core cpu is a tad pointless unless your running quite a few apps at once. Totally pointless if your running 1 app or game and even more so if the clock speed isn't very high.

 

[Grim5]

Thats the thing apps or games will only use so many cores,, so for example a 20core cpu is a tad pointless unless your running quite a few apps at once. Totally pointless if your running 1 app or game and even more so if the clock speed isn't very high.

His car analogy works well then. If you were to buy a 1000hp ferrari, roads only allow for a small fraction of that.

 

[Skyrover]

Which is not that great, its only better, a little, due to the clock speed difference.

The ST IPC on Zen 2 is about is 13% higher than Coffeelake.

Score 535: Intel Core i7 8700K at 5.2Ghz, Chaos666
Score 532: AMD Ryzen R5 3600 at 4.575Ghz, RavenXXX2

Intel needs to run at 5Ghz+ to remain relevant and the thing with that is smaller lithography nodes just don't clock as high, its why they are still with 14nm on desktop, a 5Ghz 14nm Coffeelake CPU its still faster than their next generation architecture is at 4Ghz.

AMD are going a different rout, they are on smaller nodes and with that cannot clock them as high, so they get there with higher per clock performance, and if the rumours are true Zen 3 will have 30% high IPC than Coffeealke, no 5.3Ghz 300 watt 8 core Coffeelake is going to keep up with that.

True enough, but obviously Intel's current IPC and clockspeed advantage sill give it an edge in single core performance. Older benchmarks like this are a good indicator of this.

Things may change going forward as AMD's IPC may increase enough to offset Intels frequency advantage.

 

[Robert896r1]

Thats the thing apps or games will only use so many cores,, so for example a 20core cpu is a tad pointless unless your running quite a few apps at once. Totally pointless if your running 1 app or game and even more so if the clock speed isn't very high.

Pretty much always comes down being honest with your use cases and isolating your research to that criteria.

 

[Sargatanas2511]

Thats the thing apps or games will only use so many cores,, so for example a 20core cpu is a tad pointless unless your running quite a few apps at once. Totally pointless if your running 1 app or game and even more so if the clock speed isn't very high.

I would agree with you. Hopefully with the next gen of consoles coming on what is essentially Zen architecture games will start to more efficiently use the cores and we will see it translate on PC. This should have happened a long time ago though to be honest as even the PS3 had 7 'cores' to use.

 

[Grim5]

I would agree with you. Hopefully with the next gen of consoles coming on what is essentially Zen architecture games will start to more efficiently use the cores and we will see it translate on PC. This should have happened a long time ago though to be honest as even the PS3 had 7 'cores' to use.

back then developers had no idea what to do with 7 cpu cores, multithreaded developement was still seen as witchcraft.

as such very few games from that generation ever pushed the ps3 cpu much it was just chilling in hole the gpu did all the lifting

 

[Grim5]

Can someone explain this to me.

when I had the 8700k I noticed that Intel CPUs will draw almost max power draw and get super hot even if just one core is maxed out and the rest are idling. Because of this older games that were bound to one or two threads made the cpu run a good 30c hotter than well threaded games.

but this is not the case on my ryzen 3000 cpu, max load on a single core draws less than half the power that it would with multi thread loads and as such the cpu stays very cool, typically on 10c higher than idle while one core is maxed out

So why does heavy single thread loads make Intel cpu **** the bed and Ryzen just laughs at single threaded loads?

 

[sideways14a]

There are lots of reasons why zen2 is producing less heat, amd are on a better node and have a more power efficient design.
Plus intel are cranking up the power to run the outdated cores faster to hold up the performance.

 

[Sargatanas2511]

back then developers had no idea what to do with 7 cpu cores, multithreaded developement was still seen as witchcraft.

as such very few games from that generation ever pushed the ps3 cpu much it was just chilling in hole the gpu did all the lifting

Cell wasn't exactly the easiest to code for either from what I know so that wouldn't help. The Jaguar cores of the current gen though should have been ripe for multithreading and they seem to be sort of taken advantage of but I am not convinced it's fully utilised.