13900K vs 7950X

[Laakso]

Hi!

 

I made a post a few years ago about what GPU to go with. I have been using my i7 6700K with my 3070 for about 2 years now, it has mostly been going great even though I know I am leaving a lot of performance on the table.

But when MW2 came out, I started experiencing FPS drops like never before. Which means that it's time for a well needed upgrade. But I'm not sure what I should choose.

I've been looking for the top of the top on both side, which are 13900K and 7950X, because I don't want to upgrade in a LOONG time after this.

 

A little background on what I do. I mostly play games, and mostly fps games, but I like content creation and streaming. So I want to be able to do it all without a problem.

At the moment I have 3 screens, 170hz 1440p, 144hz 1080p and 60hz 1080p. Though I'm planning on upgrading the two 1080p to 1440p aswell.

 

So which one would be the best for my upgrade? I understand that there may not be a clear winner here, mostly wanna hear what you think and if you have experience with any of them.

 

Thank you so much in advance, feel free to ask me anything if something is unclear.

 

Have a great day everyone!

Edited December 4, 2022 by Laakso
Grammar and spelling to make it easier to understand

[QuantumSingularity]

It's the usual topic, but on the flagship side. On one side you have the 13900K which in most cases is either on par with the fastest gaming CPU or really close to it, but is now on a dead platform. On the other you have the first gen of the new AM5 platform and with the added expenses of the early adoption. But since you HAVE to get a whole new system whichever way you go, i'd go AMD with 7950X This way at least when the next gen comes out or the new 3D V-cache CPUs come out, i'll be just a bios flash away from a new upgrade. 

[Maxxtraxx]

Both are great. Some points to consider:

generally the X670 are expensive the B650 are cheaper but both require DDR5 (you could potentially reuse your ddr4 from your old system with z690/h670/b660)
 

Both CPU's are very highly tuned at stock settings

 

the 13900k has Very high peak power draw (not likely an issue while gaming) and also runs hot as a result

the 7950x runs hot but has considerably lower peak power draw

 

both require high performance cpu cooling (recommended 240AIO or greater)

 

Both CPU's are turned up to 11... both run MUCH cooler and draw MUCH less power when turned down from 11 to 10 or 9(on a scale of 1-10) at the potential expense of peak clock frequencies(undervolting and negative PBO offsetting)

7950x seem to peak at about 6000-6400 MHz ddr5 speeds when on 2 dim memory configurations, when running 4 dimms and more capacity the speed drops to below 5000MHz

13900k has a much faster peak memory speeds but faster memory is also more expensive (not sure if more dimms or higher capacity effects the intel as severely)

 

PERSONLLY: coming from a history of cpu's as follows: 2500k --> 4790k --> 6700k --> 9900k --> 7950X
I'm not sold on the actual usefulness of intel's E-cores, I remember (cuz i'm old) when 4 cores was all you ever needed for gaming.... then 8 cores became future proof... will games adopt even more cores in the future? when a game can run on 10 or 12 or 16 threads will the E-cores be a liability?

I'm not interested in Windows 11 yet either... and Intel's thread scheduler to specifically assign tasks to the proper cores(P core vs E core) is not available on Windows 10, so unless you want to assign all cores priority and preference to programs yourself... you're out of luck or stuck with windows 11.

I (obviously) decided that AMD was the preferential choice for me... if more expensive choice, I find myself impressed by the longevity of the AM4 socket and the impressive performance scaling they achieved and this gives me hope that AM5 could be similar(zero hope of this from intel).

 

FOR YOU: both are closely matched and very high performance, get what you want, what fits in your budget, what fits your use case and what makes you smile... Sorry for the long opinion and long non answer.

[Chiyawa]

Well, it's quite complicated. Here's my take on this subject:

 

If you are planning to use your system for a long time and possibly upgrade it along the way, you may consider AMD system. AMD just adopt new platform, and with their records on keeping a platform alive for a long time (AM2, AM2+ which is backwards compatible with AM2 processors, AM3, AM3+, AM4 which is a very long platform as well), so there will be lots of upgrade options in the future.

 

If you're like 'I upgrade everything in 5 years or so', then, well, I mean you can go for AM4 platform which proves very stable and still on par with modern hardware, but if you max up the specs you won't have much upgrade option. If you're going to Intel 13th gen route, of course it's much more newer than AM4 so it's more compelling, but no matter what you choose, they are consider stagnate with very limited options for upgrades.

 

In summary:

AM5 with AMD 7950x: More options to upgrade in the future.

LGA1700 with 13th gen Intel: Upgrade options limited.

But no matter which you choose, as long as it will serve you well, go for it. There'll always new things coming in the future.

[JT89]

It's a hard decision, but if cost is any concern, I think the i7-13700k or kf is the way to go. The extra E cores on the 900 won't do anything for you. 

 

Intel 13th will allow for better DDR5 speeds, and since you don't plan on upgrading, I think that's an important consideration. I don't think either will lead you wrong, but intel has the fastest cores (even if you're only getting 8 of them), which will improve gaming, and the fastest DDR5 capability.

[Ebony Falcon]

35 minutes ago, Maxxtraxx said:

Both are great. Some points to consider:

generally the X670 are expensive the B650 are cheaper but both require DDR5 (you could potentially reuse your ddr4 from your old system with z690/h670/b660)
 

Both CPU's are very highly tuned at stock settings

 

the 13900k has Very high peak power draw (not likely an issue while gaming) and also runs hot as a result

the 7950x runs hot but has considerably lower peak power draw

 

both require high performance cpu cooling (recommended 240AIO or greater)

 

Both CPU's are turned up to 11... both run MUCH cooler and draw MUCH less power when turned down from 11 to 10 or 9(on a scale of 1-10) at the potential expense of peak clock frequencies(undervolting and negative PBO offsetting)

7950x seem to peak at about 6000-6400 MHz ddr5 speeds when on 2 dim memory configurations, when running 4 dimms and more capacity the speed drops to below 5000MHz

13900k has a much faster peak memory speeds but faster memory is also more expensive (not sure if more dimms or higher capacity effects the intel as severely)

 

PERSONLLY: coming from a history of cpu's as follows: 2500k --> 4790k --> 6700k --> 9900k --> 7950X
I'm not sold on the actual usefulness of intel's E-cores, I remember (cuz i'm old) when 4 cores was all you ever needed for gaming.... then 8 cores became future proof... will games adopt even more cores in the future? when a game can run on 10 or 12 or 16 threads will the E-cores be a liability?

I'm not interested in Windows 11 yet either... and Intel's thread scheduler to specifically assign tasks to the proper cores(P core vs E core) is not available on Windows 10, so unless you want to assign all cores priority and preference to programs yourself... you're out of luck or stuck with windows 11.

I (obviously) decided that AMD was the preferential choice for me... if more expensive choice, I find myself impressed by the longevity of the AM4 socket and the impressive performance scaling they achieved and this gives me hope that AM5 could be similar(zero hope of this from intel).

 

FOR YOU: both are closely matched and very high performance, get what you want, what fits in your budget, what fits your use case and what makes you smile... Sorry for the long opinion and long non answer.

I can’t see ecores becoming a liability 

consider the 9900k or even your 7950x

both have  hyper threading/smt

The hyper thread and smt thread are considerably slower than a e core 

if u had the choice between 1 p core with hyperthreading or one p core and 1 e core the e core will be much faster configuration 

We have been using faster and slower threads mixed for a hella long time before ecores showed up 

[Maxxtraxx]

22 hours ago, Ebony Falcon said:

I can’t see ecores becoming a liability 

consider the 9900k or even your 7950x

both have  hyper threading/smt

The hyper thread and smt thread are considerably slower than a e core 

if u had the choice between 1 p core with hyperthreading or one p core and 1 e core the e core will be much faster configuration 

We have been using faster and slower threads mixed for a hella long time before ecores showed up 

I'm not sure what SMT or HT being slower than an E-core means.

But the main interest is, would you rather have 16 P-cores or 8 P cores with backup from E-cores if you need more than the 8 Pcores can offer?

IMO, 16 P-cores would be the better choice. Intel added E-cores to match AMD's highly threaded performance on paper and possibly to help their power consumption in mobility. No one really wants the E-cores doing anything important. Obviously I have chosen the path i think is best via AMD and you've chosen the path you think is best via Intel so we're both biased.

 

ANANDTech: 

"In multi-threaded workloads, we saw an average uplift in performance of +22% when SMT was enabled. Most of our tests scored a +5% to a +35% gain in performance. A couple of workloads scored worse, mostly due to resource contention having so many threads in play – the limit here is memory bandwidth per thread. One workload scored +60%, a computational workload with little-to-no memory requirements; this workload scored even better in AVX2 mode, showing that there is still some bottleneck that gets alleviated with fewer instructions.

On gaming, overall there was no difference between SMT On and SMT Off, however some games may show differences in CPU limited scenarios. Deus Ex was down almost 10% when CPU limited, however Borderlands 3 was up almost 10%. As we moved to a more GPU limited scenario, those discrepancies were neutralized, with a few games still gaining single-digit percentage points improvement with SMT enabled.

For power and performance, we tested two examples where performance at two threads per core was either saw no improvement (Agisoft), or significant improvement (3DPMavx). In both cases, SMT Off mode (1 thread/core) ran at higher temperatures and higher frequencies. For the benchmark per performance was about equal, the power consumed was a couple of percentage points lower when running one thread per core. For the benchmark were running two threads per core has a big performance increase, the power in that mode was also lower, and there was a significant +91% performance per watt improvement by enabling SMT."

[Ebony Falcon]

7 minutes ago, Maxxtraxx said:

I'm not sure what SMT or HT being slower than an E-core means.

But the main interest is, would you rather have 16 P-cores or 8 P cores with backup from E-cores if you need more than the 8 Pcores can offer?

IMO, 16 P-cores would be the better choice. Intel added E-cores to match AMD's highly threaded performance on paper and possibly to help their power consumption in mobility. No one really wants the E-cores doing anything important. Obviously I have chosen the path i think is best via AMD and you've chosen the path you think is best via Intel so we're both biased.

 

ANANDTech: 

"In multi-threaded workloads, we saw an average uplift in performance of +22% when SMT was enabled. Most of our tests scored a +5% to a +35% gain in performance. A couple of workloads scored worse, mostly due to resource contention having so many threads in play – the limit here is memory bandwidth per thread. One workload scored +60%, a computational workload with little-to-no memory requirements; this workload scored even better in AVX2 mode, showing that there is still some bottleneck that gets alleviated with fewer instructions.

On gaming, overall there was no difference between SMT On and SMT Off, however some games may show differences in CPU limited scenarios. Deus Ex was down almost 10% when CPU limited, however Borderlands 3 was up almost 10%. As we moved to a more GPU limited scenario, those discrepancies were neutralized, with a few games still gaining single-digit percentage points improvement with SMT enabled.

For power and performance, we tested two examples where performance at two threads per core was either saw no improvement (Agisoft), or significant improvement (3DPMavx). In both cases, SMT Off mode (1 thread/core) ran at higher temperatures and higher frequencies. For the benchmark per performance was about equal, the power consumed was a couple of percentage points lower when running one thread per core. For the benchmark were running two threads per core has a big performance increase, the power in that mode was also lower, and there was a significant +91% performance per watt improvement by enabling SMT."

You would rather have 8 p cores with no hyperthreading and 8 e cores over 8 p cores with hyperthreading 

[Dedayog]

23 hours ago, JT89 said:

It's a hard decision, but if cost is any concern, I think the i7-13700k or kf is the way to go. The extra E cores on the 900 won't do anything for you. 

 

Intel 13th will allow for better DDR5 speeds, and since you don't plan on upgrading, I think that's an important consideration. I don't think either will lead you wrong, but intel has the fastest cores (even if you're only getting 8 of them), which will improve gaming, and the fastest DDR5 capability.

Even if cost isn't a concern, thought should be given to the use case and the proper CPU forit.

 

Neither the 13900k nor the 7950x are the best gaming cpus of their generation.

 

Especially if you're using a 3070 and 1440p.

 

[Mark Kaine]

2 hours ago, Dedayog said:

Neither the 13900k nor the 7950x are the best gaming cpus of their generation.

 

Especially if you're using a 3070 and 1440p.

 

And how come,  arent those the fastest cpus, on paper?

 

(you kinda need to explain why when you say something like this - i wouldn't know the answer, but to me it would seem the fastest cpus are best for gaming,  even if they're bad value perhaps) 

[Maxxtraxx]

On 12/5/2022 at 7:12 PM, Ebony Falcon said:

You would rather have 8 p cores with no hyperthreading and 8 e cores over 8 p cores with hyperthreading 

The Data as I interpret it does not seem to correlate. As we can see, 16 cores with hyperthreading does a nearly identical amount of total work as does 8 cores with HT and 16 E cores. I'm specifically considering the Cinebench R23 data which shows the 7950X and 13900k scoring very close to one another. This would seem to indicate that with an Identical number of logical cores the performance is very similar and thusly 2 Ecores are doing very similar work to 1 core with HT.

My point is that with that number being similar, would you rather have Pcores doing important work beyond the 16 thread mark or have E-cores doing it(because we KNOW that there is a definite performance difference between intel's Pcores and their Ecores).

For example in Cinebench R23, 13900k Pcores that is 8 cores with HT score 24k-25k(only Pcores) with a normal score for 13900k of 38k-40k that leaves all 16 e cores only scoring the remaining 15k, So of the total 40k R23 score the 8 Pcores with HT scores ~25k of the total and 16 Ecores scores ~15k of the total.

So yes, I would rather have 16 Pcores with HT than 8 Pcores with HT and 16 E cores.

Thats my logic and i'm sticking to it(for now).

Edited December 7, 2022 by Maxxtraxx

[George McGnarly]

I have the 7950x and love it. It can get a little spicy at times but I ignore it. I have it installed on an Asus x670e-e Wifi. Unfortunately, this mobo does not play well with Armourycrate at all. 

[Mark Kaine]

On 12/4/2022 at 4:49 PM, Laakso said:

the top of the top on both side, which are 13900K and 7950X,

ive seen several people claiming the 5800X3D is better for gaming ¯\_(ツ)_/¯ 

 

On 12/4/2022 at 4:49 PM, Laakso said:

I don't want to upgrade in a LOONG time after this.

then you should wait for 7700x3d or something cause those two "top" chips will be outdated very soon .

[Ebony Falcon]

2 hours ago, Maxxtraxx said:

The Data as I interpret it does not seem to correlate. As we can see, 16 cores with hyperthreading does a nearly identical amount of total work as does 8 cores with HT and 16 E cores. I'm specifically considering the Cinebench R23 data which shows the 7950X and 13900k scoring very close to one another. This would seem to indicate that with an Identical number of logical cores the performance is very similar and thusly 2 Ecores are doing very similar work to 1 core with HT.

My point is that with that number being similar, would you rather have Pcores doing important work beyond the 16 thread mark or have E-cores doing it(because we KNOW that there is a definite performance difference between intel's Pcores and their Ecores).

For example in Cinebench R23, 13900k Pcores that is 8 cores with HT score 24k-25k(only Pcores) with a normal score for 13900k of 38k-40k that leaves all 16 e cores only scoring the remaining 15k, So of the total 40k R23 score the 8 Pcores with HT scores ~25k of the total and 16 Ecores scores ~15k of the total.

So yes, I would rather have 16 Pcores with HT than 8 Pcores with HT and 16 E cores.

Thats my logic and i'm sticking to it(for now).

That’s not what I said Intel only 8p cores with ht is worse than  8 p cores no ht plus 8 ecores 

[JT89]

On 12/5/2022 at 8:25 PM, Dedayog said:

Even if cost isn't a concern, thought should be given to the use case and the proper CPU forit.

 

Neither the 13900k nor the 7950x are the best gaming cpus of their generation.

 

Especially if you're using a 3070 and 1440p.

 

Ok, I'll bite. Which ones are? I'd like to hear it.

[Mark Kaine]

6 hours ago, Maxxtraxx said:

Thats my logic and i'm sticking to it(for now)

i unno if you base your logic on cinebench, a completely arbitrary  "stress test" with no relation to real workloads, maybe you should rethink it.

 

1 hour ago, JT89 said:

Ok, I'll bite. Which ones are? I'd like to hear it.

i already asked, we will probably never know ... ¯\_(ツ)_/¯ 

[Maxxtraxx]

On 12/7/2022 at 5:03 PM, Ebony Falcon said:

That’s not what I said Intel only 8p cores with ht is worse than  8 p cores no ht plus 8 ecores 

Ok, you've clarified.

Have you any data to backup your assertion?

I've been trying to combine my reasoning with relevant data to determine the proper course.

[Maxxtraxx]

21 hours ago, Mark Kaine said:

i unno if you base your logic on cinebench, a completely arbitrary  "stress test" with no relation to real workloads, maybe you should rethink it.

 

i already asked, we will probably never know ... ¯\_(ツ)_/¯ 

It certainly depends upon the workload, but i would say that this Hardware Unboxed video shows some of the potential performance disparity between the P and E cores in a VERY significant way.

 

[Mark Kaine]

14 minutes ago, Maxxtraxx said:

shows some of the potential performance disparity between the P and E cores in a VERY significant way.

yes, i think its commonly accepted that e core are not very useful for gaming?  

 

but i think i must say i find this "chart" very unfortunate. 

 

why not compare 

6p cores 

against

6p cores + 4e cores

 

or something

*that* would definitely prove how much they add or if they add anything at all.  Objectively. 

 

 

But its HWU so it's not objective of course. always some narrative with the guy.

[Maxxtraxx]

21 hours ago, Mark Kaine said:

i unno if you base your logic on cinebench, a completely arbitrary  "stress test" with no relation to real workloads, maybe you should rethink it.

 

Interesting Anandtech article exploring Ecore efficency on the 12900k comparing 8 E-cores against a 7 year old 6700k (4c, 8t): HERE

The results seem to show many CLOSE performance comparisons between the 6700k and the Ecores. I'm not sure if I'm impressed that a 7 year old 4 core/8 thread CPU can keep up with 8 newer Ecores or that the Ecores seem to be a close match for Skylake (which was better than I expected).

[Mark Kaine]

1 minute ago, Maxxtraxx said:

Interesting Anandtech article exploring Ecore efficency on the 12900k comparing 8 E-cores against a 7 year old 6700k (4c, 8t): HERE

The results seem to show many CLOSE performance comparisons between the 6700k and the Ecores. I'm not sure if I'm impressed that a 7 year old 4 core/8 thread CPU can keep up with 8 newer Ecores or that the Ecores seem to be a close match for Skylake (which was better than I expected).

well , again, i don't disagree that e cores are mostly useless (for gaming at least) and I've also stated many times at other occasions that i think its incredible stupid and more or less a marketing gag...

 

So this is like saying water is wet. i will not argue against that either.

 

i was just saying cb is no real use case scenario. because it isnt. 

 

[Maxxtraxx]

21 minutes ago, Mark Kaine said:

well , again, i don't disagree that e cores are mostly useless (for gaming at least) and I've also stated many times at other occasions that i think its incredible stupid and more or less a marketing gag...

 

So this is like saying water is wet. i will not argue against that either.

 

i was just saying cb is no real use case scenario. because it isnt. 

 

You are correct, but, The point of a full throttle comparative test is not necessarily to be a good judge of daily usage, just as a 1/4mile drag race is not a good judge of your daily usage experience for your car, but it does offer a comparative insight putting apples against apples as part of getting the most complete picture of performance.

[Mark Kaine]

1 hour ago, Maxxtraxx said:

You are correct, but, The point of a full throttle comparative test is not necessarily to be a good judge of daily usage, just as a 1/4mile drag race is not a good judge of your daily usage experience for your car, but it does offer a comparative insight putting apples against apples as part of getting the most complete picture of performance.

cb is useful for

-checking if your temps are ok

-if the chip works more or less how it's supposed to...

 

the advantage of cb is that there is very little load variation,  which i think is probably exactly what it's designed for... but beyond that other than say, yes this chip can run this frequency at that temp it doesn't say much about performance because most workloads will just have way more variables and different requirements... 

 

so no its not useless but its limited in what it tells you about a real life working scenario (like games will act totally different,  for example) 

[Xkillerpn]

This is an easy one, buy a 13900k and yes you get the faster cpu(discarding temps , power). Now do you want a platform that will last more years and you can upgrade down the line then get the 7950x, its a bit slower yes, in 3 or 4 years you will be capable of upgrading to a faster cpu. I dont have a cristal ball but I think it will be faster then the 7950 and probbaly on par with whatever intel will have. This way you wont spend money on a board that are expensive to upgrade to another cpu. This is just my view on it.