Has AMD passed Intel or just pulled even or still slightly behind

Joined
May 30, 2016
Messages
62
This is not meant to start a flame war by fanboys.

I want an honest answer as I have read so many different things conflicting each other

It seems intel still has a slight edge in gaming by like 3-5% on a clock for clock basis as some said games are still better optimized for intel Core ringbus architecture even though AMD is even or slightly ahead clock for clock in single threaded and multithreaded raw compute power at same clock speed.

Although some say intel still has a slightly better IPC at same clock speed, but the gap is almost nill when it comes to 8 core CPUs.

Although the confusing thing is, it seems AMD 16 core 3950X is break even or even slightly ahead of intel's Core i9 10980XE 18 core CPU Despite haivng 2 less cores: https://www.forbes.com/sites/antony...-better-than-amds-ryzen-9-3950x/#4eea8f17405f

And that is with 2 less cores!! ANd evne some bewnchmarks show it winning against intel's 28 core $3000 CPU which would have to mean single threaded performance is much better for AMD.

https://www.pcgamesn.com/amd/ryzen-9-3950x-beats-intel-28-core-cpu



How could that be if intel is even or still slightly ahead on Coffee Lake 8 core CPUs compared to Ryzen 8 core CPUs clock for clock?

Is it that intels higher core counts are mesh topology which is inferior to AMD's current Zen 2 architecture and even intel's own ring bus topology for current geb apps?

Obviously AMD wins on being able to stuff more cores into CPU easier and at least at almost break even performance IPC because of lower power consumption and easier to cool it with 7nm fabric vs intel's 14nm fabric

Although intel Coffee Lake CPUs overlcok better while being stable as long as the cooling is up to par to deal with the more heat.

What are everyone's thoughts on AMD seeming to win on 16 or more core CPUs vs intel's even higher core counts, but lose or be only even with intel on 8 core Zen 2 CPUs vs Intel' Coffee Lake 8 core CPUs?
 
This is not meant to start a flame war by fanboys.

I want an honest answer as I have read so many different things conflicting each other

It seems intel still has a slight edge in gaming by like 3-5% on a clock for clock basis as some said games are still better optimized for intel Core ringbus architecture even though AMD is even or slightly ahead clock for clock in single threaded and multithreaded raw compute power at same clock speed.

Although some say intel still has a slightly better IPC at same clock speed, but the gap is almost nill when it comes to 8 core CPUs.

Although the confusing thing is, it seems AMD 16 core 3950X is break even or even slightly ahead of intel's Core i9 10980XE 18 core CPU Despite haivng 2 less cores: https://www.forbes.com/sites/antony...-better-than-amds-ryzen-9-3950x/#4eea8f17405f

And that is with 2 less cores!! ANd evne some bewnchmarks show it winning against intel's 28 core $3000 CPU which would have to mean single threaded performance is much better for AMD.

https://www.pcgamesn.com/amd/ryzen-9-3950x-beats-intel-28-core-cpu



How could that be if intel is even or still slightly ahead on Coffee Lake 8 core CPUs compared to Ryzen 8 core CPUs clock for clock?

Is it that intels higher core counts are mesh topology which is inferior to AMD's current Zen 2 architecture and even intel's own ring bus topology for current geb apps?

Obviously AMD wins on being able to stuff more cores into CPU easier and at least at almost break even performance IPC because of lower power consumption and easier to cool it with 7nm fabric vs intel's 14nm fabric

Although intel Coffee Lake CPUs overlcok better while being stable as long as the cooling is up to par to deal with the more heat.

What are everyone's thoughts on AMD seeming to win on 16 or more core CPUs vs intel's even higher core counts, but lose or be only even with intel on 8 core Zen 2 CPUs vs Intel' Coffee Lake 8 core CPUs?


Simple answer is Intel needs the higher clocks to beat the AMD Ryzen chips. However when Intel adds more cores they cant clock up as high and they start to fall behind the AMD chips. AMD in fact has better clock speeds when the core count gets above 16. Any IPC advantage Intel has is highly dependent on being at least 5 GHz or better.
 
I'd say a lot has to do with the new infinity fabric changes. Also the huge 36MB cache on single ccd chips and 72mb cache on 2 ccd chips has a huge play in performance. Also by having the uncore on a separate chiplet that can be clock matched to your ram frequency has a huge play in making the actual cores more efficient at being cores. Where the cores can be clocked higher by not being slaved to the uncore directly


Inf fab vs mesh has lots of articles out but most of them are 2017 to 18 time frame and are obsolete. Inf fab had many changes to it with zen 2. For instance infinity fab on zen 1/1.5 threadrippers had separate Numa nodes and only 2 of 4 ccds could have direct access to memory. In zen 2 i.e. 3000 series there is only one single numa node across the whole chip so cores get direct mem access.

One can go on and on. I'd say the chips are neck and neck. The issue is looking at the micro vs macro. If all care about is gaming the Intel is a wee bit faster due to frequency advantage. Productivity amd offers more effective SMT and far more l3 cache so they tend to win. But the macro scale would say the chips are neck to neck but AMD wins with a better price per core per performance point of view. That is changing though as Intel is finally dropping prices due to true direct competition.

In short AMD has a superior chiplet approach while intel is still held back by rather obsolete monolithic designs.

Intel will indeed catch up and surpass AMD. Competition is awesome! Finally there is.
 
Last edited:
Who has the advantage assuming the clock speed is the same across all cores and core count is equal?

Zen 2 8 cores or Coffee Lake 8 cores say at 4.5GHz each with turbo boost turned off meaning both say the Ryzen 8 3800X and Core i9 9900K all cores run 4.5GHz all the time under any load and at idle? For both single and multi threaded performance?

Of course this also assumes video card and RAM amount and type and speed is also the same and disk storage is also the same.
 
AMD have higher IPC but Intel have higher clockspeed.
It depends on your use case (number of cores needed vs boost from clock speed) which will serve you better.
 
Who has the advantage assuming the clock speed is the same across all cores and core count is equal?

Zen 2 8 cores or Coffee Lake 8 cores say at 4.5GHz each with turbo boost turned off meaning both say the Ryzen 8 3800X and Core i9 9900K all cores run 4.5GHz all the time under any load and at idle? For both single and multi threaded performance?

Of course this also assumes video card and RAM amount and type and speed is also the same and disk storage is also the same.
The problem is, it depends on the work load. Intel still has higher IPC with games. If both AMD and Intel are clocked exactly the same as you say, Intel will still slightly out pace AMD in general. If you are talking about something like rendering. AMD tends to do better and will finish quicker than Intel given the same clock speed. It's not as simple as ones faster or slower anymore, it's dependent on the work load (for both single and/or multi threaded). I would say (at least in my opinion) that AMD is superior to Intel at this point for most instances/use cases. If you are a twitch gamer, then Intel is probably your market. If you use your PC for more than just gaming it's a toss up. If you do any sort of video editing or 3d rendering, more than likely you're going to be best with an AMD build. I would much prefer a 3600 for $200 that keeps up with the 9600k $250 in most games and outperforms handily at heavily threaded work loads. Take into account the $50 savings, plus it comes with a usable CPU cooler in the box, it's a no brainer for me to recommend. Of course, motherboard costs can vary a lot, but a B450 isn't very expensive and works wonderfully well with the 3600.
Anyways, that turned into more than I anticipated, but really it depends on the work load, but for all intents and purposes they are pretty equal at this point with one better at some and the other better at some.
 
Here is my 3960x underwater hitting GPU compression speeds via h265.

Just testing a 1080p to 1080p 264 to265 encode. This cpu is retarded powerful.

upload_2019-11-30_14-30-40.png


I need MOAR RAM!

upload_2019-11-30_14-31-58.png
 
The 3950x is AMD being very loyal to AMD 4 and Gen 1 buyers but TR40 is the next step for Intel to climb as the 3960/70 are working parts and not the future as that's already here .
 
The problem is, it depends on the work load. Intel still has higher IPC with games. If both AMD and Intel are clocked exactly the same as you say, Intel will still slightly out pace AMD in general. If you are talking about something like rendering. AMD tends to do better and will finish quicker than Intel given the same clock speed. It's not as simple as ones faster or slower anymore, it's dependent on the work load (for both single and/or multi threaded). I would say (at least in my opinion) that AMD is superior to Intel at this point for most instances/use cases. If you are a twitch gamer, then Intel is probably your market. If you use your PC for more than just gaming it's a toss up. If you do any sort of video editing or 3d rendering, more than likely you're going to be best with an AMD build. I would much prefer a 3600 for $200 that keeps up with the 9600k $250 in most games and outperforms handily at heavily threaded work loads. Take into account the $50 savings, plus it comes with a usable CPU cooler in the box, it's a no brainer for me to recommend. Of course, motherboard costs can vary a lot, but a B450 isn't very expensive and works wonderfully well with the 3600.
Anyways, that turned into more than I anticipated, but really it depends on the work load, but for all intents and purposes they are pretty equal at this point with one better at some and the other better at some.

Stop mixing IPC and other metrics.

It's all about latency to memory and core to core.

Intel has better Latency with their Ringbus cpu's, Mesh has Worse.
Meaning HCC chips like 7900x, 10900x, 10980xe etc have worse latency than their Coffee lake and even zen2.

Amd beats intel in IPC across the board with zen2 and IPC is not related to clock speeds
their HCC (Skl-x, cfl-x) chips run AVX512 and can win in ipc... not very common for consumers, not very useful and if it is for you, you know it so don't ask.


Take all 3 archs at 4300 mhz allcore:

Latency matters, coffee lake is king, then Zen2, Then Skylake-X.
If skylake-X beats zen2 in games it's probably cause the engine does something stupid and it's optimization.
If coffee lake beats zen2 then it's latency which makes it win
If zen2 beats coffee lake then the game doesn't to a lot of cross CCX talk or it all fits into their CCX L3 cache so it never goes far to get data hence never really get any latency penalty.
 
This is not meant to start a flame war by fanboys.

I want an honest answer as I have read so many different things conflicting each other

It seems intel still has a slight edge in gaming by like 3-5% on a clock for clock basis as some said games are still better optimized for intel Core ringbus architecture even though AMD is even or slightly ahead clock for clock in single threaded and multithreaded raw compute power at same clock speed.

Although some say intel still has a slightly better IPC at same clock speed, but the gap is almost nill when it comes to 8 core CPUs.

Although the confusing thing is, it seems AMD 16 core 3950X is break even or even slightly ahead of intel's Core i9 10980XE 18 core CPU Despite haivng 2 less cores: https://www.forbes.com/sites/antony...-better-than-amds-ryzen-9-3950x/#4eea8f17405f

And that is with 2 less cores!! ANd evne some bewnchmarks show it winning against intel's 28 core $3000 CPU which would have to mean single threaded performance is much better for AMD.

https://www.pcgamesn.com/amd/ryzen-9-3950x-beats-intel-28-core-cpu



How could that be if intel is even or still slightly ahead on Coffee Lake 8 core CPUs compared to Ryzen 8 core CPUs clock for clock?

Is it that intels higher core counts are mesh topology which is inferior to AMD's current Zen 2 architecture and even intel's own ring bus topology for current geb apps?

Obviously AMD wins on being able to stuff more cores into CPU easier and at least at almost break even performance IPC because of lower power consumption and easier to cool it with 7nm fabric vs intel's 14nm fabric

Although intel Coffee Lake CPUs overlcok better while being stable as long as the cooling is up to par to deal with the more heat.

What are everyone's thoughts on AMD seeming to win on 16 or more core CPUs vs intel's even higher core counts, but lose or be only even with intel on 8 core Zen 2 CPUs vs Intel' Coffee Lake 8 core CPUs?

Others have touched on this, but I'll give it a shot anyway. Intel does have a slight edge in gaming. In some cases, its a bit more than slight. The aggregate advantage generally agreed upon is about 5-6%. The range however is more like 0-15%. AMD does have a very slight IPC advantage over Intel which is offset by Intel's higher clocks. AMD doesn't have much of a lead here, so clocking an Intel CPU anywhere from 400MHz to 600MHz higher negates this IPC advantage. Therefore, Intel still leads in terms of single-threaded performance in many cases. This is true of any of their mainstream desktop processors, which are all generally capable of clocking to 5.0GHz if its a K part.

Mesh topology doesn't come into play on the desktop parts. On the HEDT side, it is hurting Intel but that's not really the issue. At 14nm, Intel's CPU's simply run too hot and consume way too much power to be competitive with AMD in multi-threaded workloads where AVX-512 doesn't come into play. Intel does have Xeons with core counts as high as 56 which are the undisputed king of Deep Learning tests as far as CPU's go. The problem is that these higher core count Xeons are physically massive and are much larger than their desktop counterparts. They use LGA 3647 which is about 4 times the size of LGA 1151. These CPU's have a TDP of 400 watts at 2.60GHz and only boost to around 3.8GHz. This is fine in multi-threaded applications, but makes for a poor desktop CPU. Clock speed still matters.

AMD's flagship mainstream part, the 3950X still loses to Intel's 10980XE quite a lot of the time and this lead for Intel increases when you really start pushing its clocks. However, this is no victory as the Intel has to have two more CPU cores and pulls about 200 watts more to do it. And again, you need to overclock the Intel in order to get a definitive win across the board. Even then, there are still a handful of things that favor AMD. Cinebench R20 is a good example of that. It's also necessary to point out that with the 10980XE, Intel's clock speed advantage virtually disappears. The 3950X clocks to around 4.1GHz or so all core and 4.7GHz single core. Comparatively, Intel clocks at 4.6GHz with turbo boost 2.0 (single core) and 4.8GHz with Turbo Boost Max Technology 3.0 (4-core). All core clocks are around 4.0GHz. Single core clocks aren't enough to push Intel ahead and all core clocks are too low to compete either, thus two more cores have to make up the slack.

You can get Turbo Boost Max 3.0 to clock to 4.8GHz some of the time, but again, this isn't a massive lead over AMD. It's probably enough in some cases, but not definitively. If you overclock these CPU's, you can get around 4.5GHz to 4.8GHz all core, but the CPU runs extremely hot. It can reach over 100c with a custom loop if pushed really hard. You can't simply slap an AIO on this thing or a big ass air cooler and call it a day.

Lastly, you get into the cost issue. The only CPU's that compete with AMD are the 9900K (and by extension its variants) and the 10980XE. Intel's 9900K, KF, and KS are king in the gaming arena, and still beat out AMD's 8c/16t CPUs in some cases. But price wise, they are almost as expensive as a 3900X which dominates them at every turn outside of gaming and is generally "close enough" on that front. Intel's 10980XE competes well against the 3900X and 3950X CPU's, but there are a laundry list of caveats I've touched on already. However, the biggest one is price. The 3950X costs $749.99 if you can find one. The even harder to find 10980XE costs $979 in 1,000 unit quantities or around $1,050 street price on the low end.

We haven't even touched on platform prices yet. On the AMD side, you have a lot more choices as you can go with earlier and lower end chipset based boards. I don't necessarily recommend this route, but its certainly doable and it can make sense for some people. The important factor here is choice. That means a high end X470 or mid-range X570 and so on. X570 motherboards can be ridiculously expensive, and given the age of X299 motherboards, you can certainly pick up some for similar to X570 pricing. That said, the average X299 motherboard will probably bemore expensive than the average X570 motherboard. You do get more PCIe lanes, more memory bandwidth and less bullshit than you do dealing with AMD's offerings. Z390 compared to X470 or X570 often favors X470 and in the X570 realm, its the same thing as X299 vs. X570. The Z390 can be lower priced than the more expensive X570's but the average price is probably similar.

There are still a few more reasons why Intel, even where it wins is a little behind the curve. On the Z390 front, this isn't really an issue. In fact, you can throw about any old RAM into one and it will be fine. RAM differences don't impact Intel as much as they do AMD, so that's a win for our 9900K, KF and KS CPU's. However, on the HEDT side, things are much, much worse. You need four modules to feed these CPU's enough memory bandwidth. This further increases the cost of the Cascade Lake-X family over X570 / Ryzen 3000. The last, ugly and often forgotten truth is this. The Intel system will need a much more powerful power supply. In my test configuration, with the GPU sitting idle, the 10980XE pulled 617 watts overclocked. Add in a beefy GPU like an RTX 2080 Ti, water cooling hardware, lots of fans and your over 85% capacity on a 1,000 watt power supply. In other words, 1,000 watt PSU is the minimum standard for a Intel 10980XE system built for gaming. Even then, I would want a bigger PSU than that. This adds to the cost.

I hope I've painted a clear picture of the situation.
 
They're neck and neck in gaming at stock, amd crushes Intel in almost everything else, Intel slightly ahead in gaming overclocked.
 
Its good to see Amd come into market and bring prices down a bit. I'm looking forward to moving to a 3900x in the near future to give amd a try. I would say they are right on the tail of intel
 
Its good to see Amd come into market and bring prices down a bit. I'm looking forward to moving to a 3900x in the near future to give amd a try. I would say they are right on the tail of intel

AMD can be given credit for a lot of things but the only prices they've brought down are Intel's. AMD's top end mainstream part is $750. That's $250 more than Intel had priced theirs at. AMD's two Threadripper CPU's are at $1,399 and $1,999 for the 3960X and 3970X respectively. This is roughly what the old 9980XE was priced at. I've said it many times, when AMD can get away with it, the company charges as much or more for its products than Intel does.
 
AMD can be given credit for a lot of things but the only prices they've brought down are Intel's. AMD's top end mainstream part is $750. That's $250 more than Intel had priced theirs at. AMD's two Threadripper CPU's are at $1,399 and $1,999 for the 3960X and 3970X respectively. This is roughly what the old 9980XE was priced at. I've said it many times, when AMD can get away with it, the company charges as much or more for its products than Intel does.


My mainstream 9900k was almost $500.00 which I'm happy with the cpu, but for $250 more you are getting double cores and threads, thats not a bad deal at all for the 3950x. Intel was forced to drop the prices on the Hedt cpus due to Amd taking market share at least imo. Im still learning more on this side of things since I do more gaming and streaming then actual work on my pc.
 
My mainstream 9900k was almost $500.00 which I'm happy with the cpu, but for $250 more you are getting double cores and threads, thats not a bad deal at all for the 3950x. Intel was forced to drop the prices on the Hedt cpus due to Amd taking market share at least imo. Im still learning more on this side of things since I do more gaming and streaming then actual work on my pc.

AMD is giving you a lot for the money, but again, they are actually increasing the price points set by Intel in recent years. Mainstream is now up $250 and HEDT is up to about $100 more give or take depending on the retailer. The entry level into AMD's new HEDT line is also $800 more than Intel's. Of course you still have plenty of 2nd generation Threadrippers out there at reasonable prices but Intel is fairly competitive with those.
 
Last edited:
AMD is giving you a lot for the money, but again, they are actually increasing the price points set by Intel in recent years. Mainstream is now up $250 and HEDT is up to about $100 more give or take depending on the retailer. The entry level into AMD's new HEDT line is also $800 than Intel's. Of course you still have plenty of 2nd generation Threadrippers out there at reasonable prices but Intel is fairly competitive with those.

I do not like high prices, but with AMD I can see why they have a price premium. They offer 2 things that Intel cannot. PCI-E 4.0 and more cores (at a reasonable price). So, I am all for AMD charging more since no other company can offer what AMD does.

Now, once Intel catches up to AMD, then I would expect AMD To lower prices.
 
AMD can be given credit for a lot of things but the only prices they've brought down are Intel's. AMD's top end mainstream part is $750. That's $250 more than Intel had priced theirs at. AMD's two Threadripper CPU's are at $1,399 and $1,999 for the 3960X and 3970X respectively. This is roughly what the old 9980XE was priced at. I've said it many times, when AMD can get away with it, the company charges as much or more for its products than Intel does.

Saying they've raised prices ignores the performance increase they've delivered for that price. The 3950x doubles the performance of the previous top mainstream part, the 3970x gives a 40% performance jump for ~10% more money vs the previous top hedt part. Intel has added almost no performance so they had no choice, and even then their top hedt part barely beats amds top mainstream part while costing way more.

Plus, the only selling point for Intel, AI training, is vastly faster on gpus so isn't even much of a selling point imo.
 
AMD is giving you a lot for the money, but again, they are actually increasing the price points set by Intel in recent years. Mainstream is now up $250 and HEDT is up to about $100 more give or take depending on the retailer. The entry level into AMD's new HEDT line is also $800 than Intel's. Of course you still have plenty of 2nd generation Threadrippers out there at reasonable prices but Intel is fairly competitive with those.


Good point Dan, Im sure I could be fine with the 3800x but the 3900x would be nice to get on the amd side. I'm late the the whole pc building world and what not but its a good time non the less
 
Saying they've raised prices ignores the performance increase they've delivered for that price. The 3950x doubles the performance of the previous top mainstream part, the 3970x gives a 40% performance jump for ~10% more money vs the previous top hedt part. Intel has added almost no performance so they had no choice, and even then their top hedt part barely beats amds top mainstream part while costing way more.

Plus, the only selling point for Intel, AI training, is vastly faster on gpus so isn't even much of a selling point imo.

Thats kinda where I'm like well maybe its time to rethink how i give intel my money and look at the value instead of i gotta have the best of the best
 
Threadripper 3960x/3970x/3990x do not compete with Intel HEDT. Intel HEDT has been killed off by Ryzen 3900x/3950x. TR3 competes with Intel Workstation class chips and they are delivering them at half the cost. Compare the 3960x vs the W-3175x. Even when overclocked and requiring nearly sub ambient cooling, the 3175x still gets beat down. And it gets worse when you move up Intel's product stack.

Levelontechs guy has touched on this a few times. TR3 is cheap when compared to a direct competitor.
 
Saying they've raised prices ignores the performance increase they've delivered for that price. The 3950x doubles the performance of the previous top mainstream part, the 3970x gives a 40% performance jump for ~10% more money vs the previous top hedt part. Intel has added almost no performance so they had no choice, and even then their top hedt part barely beats amds top mainstream part while costing way more.

Plus, the only selling point for Intel, AI training, is vastly faster on gpus so isn't even much of a selling point imo.

I'm not ignoring the performance increases. They are newer products, so it stands to reason there should be a performance increase. Especially in light of the fact that the Zen+ parts had a number of deficiencies compared to the competition which have now largely if not entirely been eliminated. I figured that was a given.
 
Stop mixing IPC and other metrics.

It's all about latency to memory and core to core.

Intel has better Latency with their Ringbus cpu's, Mesh has Worse.
Meaning HCC chips like 7900x, 10900x, 10980xe etc have worse latency than their Coffee lake and even zen2.

Amd beats intel in IPC across the board with zen2 and IPC is not related to clock speeds
their HCC (Skl-x, cfl-x) chips run AVX512 and can win in ipc... not very common for consumers, not very useful and if it is for you, you know it so don't ask.


Take all 3 archs at 4300 mhz allcore:

Latency matters, coffee lake is king, then Zen2, Then Skylake-X.
If skylake-X beats zen2 in games it's probably cause the engine does something stupid and it's optimization.
If coffee lake beats zen2 then it's latency which makes it win
If zen2 beats coffee lake then the game doesn't to a lot of cross CCX talk or it all fits into their CCX L3 cache so it never goes far to get data hence never really get any latency penalty.
Umm... Stop confusing reality? Intel clocked the same as AMD in games is faster. This has been proven time and again. AMD is generally (clock for clock) faster in the majority of everything else (with some exceptions). I hate to be the bearer of this news, but if they are at the same clock and one is running faster than it's performing more instructions per clock than the other. Whether that's because core latency, memory controller, specific op codes being performed, it's still indicitave of "actual" IPC, not just theoretical. IPC is not directly related to clock speeds, but it is indirectly as there are other latencies which will do affect real world throughput. Again, theory vs reality. The fact remains clock for clock some work loads to to Intel (gaming) and other go to AMD (most others). Stop confusing your dream of reality for reality itself. You can't isolate IPC from the rest of the system, nor would this metric be useful because it wouldn't help understand how it performs in real world processes.
 
There are some predictions out there that AMD Epyc could drive Intel marketshare in the server arena below 90% in 2020. And that's when things start to get truly interesting.
 
Zen 2's IPC is superior, except in extremely latency-sensitive tasks (gaming being one), where it is still slightly behind. Slightly.

Intel has superior clocks. So at same core/thread count, Intel can sometimes pull wins even in tasks where AMD's IPC is superior.

However, AMD offers more cores/threads at almost any given price point, negating that Intel advantage in most non-AVX512 multithreaded workloads, or with mainstream vs mainstream parts. Gaming still suffers, but by only 5-6% on average, sku vs. sku. AMD sometimes outright wins in some games, but Intel still wins more. We're really at "close enough" territory here, generally.

The long and short of it is, if you don't use AVX512, AMD wins outright in almost all categories in HEDT.

In gaming, AMD is better in the value/budget space because of superior price/performance ratio and free cooler.

Intel competes well with only two CPUs, the 9900k for outright gaming builds (and the 9700k kinda-sorta, but I wouldn't buy a non-HT CPU these days), and the 10980XE for certain *specific* workloads. But either a 3950X or a Threadripper will be a better value than the 10980XE in most cases.
 
Zen 2's IPC is superior, except in extremely latency-sensitive tasks (gaming being one), where it is still slightly behind. Slightly.

Intel has superior clocks. So at same core/thread count, Intel can sometimes pull wins even in tasks where AMD's IPC is superior.

However, AMD offers more cores/threads at almost any given price point, negating that Intel advantage in most non-AVX512 multithreaded workloads, or with mainstream vs mainstream parts. Gaming still suffers, but by only 5-6% on average, sku vs. sku. AMD sometimes outright wins in some games, but Intel still wins more. We're really at "close enough" territory here, generally.

The long and short of it is, if you don't use AVX512, AMD wins outright in almost all categories in HEDT.

In gaming, AMD is better in the value/budget space because of superior price/performance ratio and free cooler.

Intel competes well with only two CPUs, the 9900k for outright gaming builds (and the 9700k kinda-sorta, but I wouldn't buy a non-HT CPU these days), and the 10980XE for certain *specific* workloads. But either a 3950X or a Threadripper will be a better value than the 10980XE in most cases.

That sums it up pretty good.

If Intel was offering similar threads-cores/dollar, it would be very competetive across the desktop parts. But as it is, they are at a significant thread deficit at every price point.

So it's rare for Intel to win anything outside gaming in a price comparison, and really the gaming margins are almost irrelevant.
 
  • Like
Reactions: N4CR
like this
In truth, very irrelevant. Even with AMD's increased pricing, they're still better than Intel right now. But, with that said, I don't own a Ryzen, so can't speak to the reliability / stability of the chipset, etc... (that stuff is important too).

Some "news" has been a mixed bag. Claims of unreachable clocks, memory compatibility issues, etc... I know people don't post unless to complain, but a large number of success stories would help. Especially from folk folks with systems under relatively good loads that are on 24 x 7 and are more than 6 months old (with full specs as well, as not all parts are created equal).
 
Zen 2's IPC is superior, except in extremely latency-sensitive tasks (gaming being one), where it is still slightly behind. Slightly.

Intel has superior clocks. So at same core/thread count, Intel can sometimes pull wins even in tasks where AMD's IPC is superior.

However, AMD offers more cores/threads at almost any given price point, negating that Intel advantage in most non-AVX512 multithreaded workloads, or with mainstream vs mainstream parts. Gaming still suffers, but by only 5-6% on average, sku vs. sku. AMD sometimes outright wins in some games, but Intel still wins more. We're really at "close enough" territory here, generally.

The long and short of it is, if you don't use AVX512, AMD wins outright in almost all categories in HEDT.

In gaming, AMD is better in the value/budget space because of superior price/performance ratio and free cooler.

Intel competes well with only two CPUs, the 9900k for outright gaming builds (and the 9700k kinda-sorta, but I wouldn't buy a non-HT CPU these days), and the 10980XE for certain *specific* workloads. But either a 3950X or a Threadripper will be a better value than the 10980XE in most cases.

That sums it up pretty good.

If Intel was offering similar threads-cores/dollar, it would be very competetive across the desktop parts. But as it is, they are at a significant thread deficit at every price point.

So it's rare for Intel to win anything outside gaming in a price comparison, and really the gaming margins are almost irrelevant.

It does. The only addendum to that would be regarding the 10980XE itself in the mid-range HEDT market space. The older 2nd generation Threadripper CPU's like the 2920X and 2950X are going to lose to the 10980XE in most cases. The 2970WX and 2990WX are going to be better in some workloads, but they carry all the baggage of Zen 1 and Zen+ with them and thus, falter badly in gaming and despite the core advantage, lose to the 10980XE where AVX-512 can be leveraged. If your working and doing nothing else, the 2970WX and 2990WX can represent a better value for the right price. However, they come with caveats just as the 10980XE does. That midrange space is a bit strange but that's the only place where Intel has any value.

I'm actually quite fond of it, and in the mid-range, the 10980XE would be my personal leading choice over the 2970WX and 2990WX as those two suck at gaming. When I tested the 2990WX, I was sorely disappointed by its gaming performance. The entire line sucks at it, but the 2990WX is worse given its massive CCX complexes and NUMA issues.
 
It does. The only addendum to that would be regarding the 10980XE itself in the mid-range HEDT market space. The older 2nd generation Threadripper CPU's like the 2920X and 2950X are going to lose to the 10980XE in most cases. The 2970WX and 2990WX are going to be better in some workloads, but they carry all the baggage of Zen 1 and Zen+ with them and thus, falter badly in gaming and despite the core advantage, lose to the 10980XE where AVX-512 can be leveraged. If your working and doing nothing else, the 2970WX and 2990WX can represent a better value for the right price. However, they come with caveats just as the 10980XE does. That midrange space is a bit strange but that's the only place where Intel has any value.

I'm actually quite fond of it, and in the mid-range, the 10980XE would be my personal leading choice over the 2970WX and 2990WX as those two suck at gaming. When I tested the 2990WX, I was sorely disappointed by its gaming performance. The entire line sucks at it, but the 2990WX is worse given its massive CCX complexes and NUMA issues.

The 3960X and 3970X rectify that, generally, it seems. Of course, good luck getting one!
 
This is not meant to start a flame war by fanboys.

I want an honest answer as I have read so many different things conflicting each other

It seems intel still has a slight edge in gaming by like 3-5% on a clock for clock basis as some said games are still better optimized for intel Core ringbus architecture even though AMD is even or slightly ahead clock for clock in single threaded and multithreaded raw compute power at same clock speed.

Although some say intel still has a slightly better IPC at same clock speed, but the gap is almost nill when it comes to 8 core CPUs.

Although the confusing thing is, it seems AMD 16 core 3950X is break even or even slightly ahead of intel's Core i9 10980XE 18 core CPU Despite haivng 2 less cores: https://www.forbes.com/sites/antony...-better-than-amds-ryzen-9-3950x/#4eea8f17405f

And that is with 2 less cores!! ANd evne some bewnchmarks show it winning against intel's 28 core $3000 CPU which would have to mean single threaded performance is much better for AMD.

https://www.pcgamesn.com/amd/ryzen-9-3950x-beats-intel-28-core-cpu



How could that be if intel is even or still slightly ahead on Coffee Lake 8 core CPUs compared to Ryzen 8 core CPUs clock for clock?

Is it that intels higher core counts are mesh topology which is inferior to AMD's current Zen 2 architecture and even intel's own ring bus topology for current geb apps?

Obviously AMD wins on being able to stuff more cores into CPU easier and at least at almost break even performance IPC because of lower power consumption and easier to cool it with 7nm fabric vs intel's 14nm fabric

Although intel Coffee Lake CPUs overlcok better while being stable as long as the cooling is up to par to deal with the more heat.

What are everyone's thoughts on AMD seeming to win on 16 or more core CPUs vs intel's even higher core counts, but lose or be only even with intel on 8 core Zen 2 CPUs vs Intel' Coffee Lake 8 core CPUs?

So it boils down to heat and clocks, and also the fact 3950x is not a full fledged HEDT processor unlike 10980xe which is HEDT (and hence why 3950x is not designated a Threadripper)

Intel coffee lake R clocks are not a concern as even all 8 cores at 5ghz heat is manageable

But for HEDT its not the case.

10900x/10920x max all core boost is 4.3ghz
10940x 14c max all core boost is 4.1ghz
10980xe 18c max all core boost is 3.8ghz
And all this already at 165w tdp

If Intel (or the end user) could keep the max all core boost for 10980xe similar to 10900x/10920x there would be less issue. But because heat and power issues linked to old 14nm process clock has to drop after 12c for stock clocks.

Thinking about it very simply, there is 500mhz lost per core on the 10980xe vs 10920x due to heat lowering clocks. Multiply that out and you have an effective 9ghz, or the equivalent of two cores, lost to heat. So the 18c CPU essentially loses 2c worth of work to that lowered clock speed. The 10900x-10940x end up being faster than the 10980xe with stock turbo tables on lighter loads/games due to heat also. If you look at it that way it's more like this

With clock boost drops accounted for and apples to apples comparison:
10900x = 10 core effective
10920x = 12 core effective
10940x = 13.5 core effective
10980xe = 16 core effective

And that's how the 10980xe ends up trading blows with 3950x. Note also 3950x has simpler design than 10980xe due to only 24 vs 48 CPU lanes, only dual vs quad channel memory, and no avx512 - losing those features on 3950x allows more of CPU go towards making more basic single and multicore tasks faster. There is less feature discrepancy between coffee lake r and mainstream Ryzen than Cascade lake X and mainstream Ryzen, hence another reason there is room for AMD to perform higher in basic single/multi tasks in the latter case... And also why the 3950x outperforms the 3960x/3970x in some lower core scenarios

Note if you overclock the 10980xe on water to overcome the stock clock drops it will have beastly performance. Just make sure u bring 1200w psu along :D
 
Last edited:
As you pointed out, Intel's Cascade Lake-X line seems to overclock very well if the 10980XE is any indicator. Mine will do 4.7GHz on all cores. Yes, its' extremely hot but the performance is definitely there and that's enough to generally beat a 3950X. Although, I will put this theory to the test as I have a 3950X inbound for review. Sorry I'm going to be later than I wanted with that, but AMD didn't sample me this time. (Blame them.)

I wouldn't bother with Intel's HEDT without water cooling. I'm not kidding, when you start to push them they get ridiculously hot and the power consumption is outrageous.
 
As you pointed out, Intel's Cascade Lake-X line seems to overclock very well if the 10980XE is any indicator. Mine will do 4.7GHz on all cores. Yes, its' extremely hot but the performance is definitely there and that's enough to generally beat a 3950X. Although, I will put this theory to the test as I have a 3950X inbound for review. Sorry I'm going to be later than I wanted with that, but AMD didn't sample me this time. (Blame them.)

I wouldn't bother with Intel's HEDT without water cooling. I'm not kidding, when you start to push them they get ridiculously hot and the power consumption is outrageous.

When you do test the 3950x. Could you test the overclocking results with SMT off? I am curious if you can get better all core clock with SMT off.
 
When you do test the 3950x. Could you test the overclocking results with SMT off? I am curious if you can get better all core clock with SMT off.

I'm not sure that I like that idea. Would anyone really turn SMT off on these things? I can try it and see what happens, but I can't imagine anyone would run one of these CPU's like that. I know I wouldn't.
 
  • Like
Reactions: Mega6
like this
I'm not sure that I like that idea. Would anyone really turn SMT off on these things? I can try it and see what happens, but I can't imagine anyone would run one of these CPU's like that. I know I wouldn't.

Well I do have a 3950x on the way, will be here Wednesday.Well the reason I was curious is (doesn't mean you have too, was just curious), I would like to see if you can hit 4.6-4.7ghz with just 16 threads, just to see how close it would get to a 9900k in gaming related data.

Either way I do plan to fuck around with my 3950x just to see for shits and giggles.
 
Well I do have a 3950x on the way, will be here Wednesday.Well the reason I was curious is (doesn't mean you have too, was just curious), I would like to see if you can hit 4.6-4.7ghz with just 16 threads, just to see how close it would get to a 9900k in gaming related data.

Either way I do plan to fuck around with my 3950x just to see for shits and giggles.

The 3900X topped out at 4.3GHz. that seems to be the general wall for a CPU like that one. Sometimes you see 4.4GHz, so even with SMT off I don't see 4.6GHz as remotely realistic. I don't even know if turning SMT off even helps these chips.
 
Back
Top