News AMD Ryzen 9 7950X and Ryzen 5 7600X Review: A Return to Gaming Dominance

Page 4 - Seeking answers? Join the Tom's Hardware community: where nearly two million members share solutions and discuss the latest tech.

johnnyboy5520

Distinguished
Apr 9, 2013
9
3
18,515
Which won't matter if the performance increase is only because of the cache/that's how high the cache can go.
That's a really big 'If'. A youtuber using lower latency RAM saw noticeable gains with the 5800X3D so there's performance still on the table even with that chip.
 

KyaraM

Admirable
Well, if I'm wrong, they're not testing Zen4 with the recommended memory and EXPO settings (someone correct me on that). And, Zen4 with 3d v-cache will retake the crown anyway. Until their next architecture, Intel has little to no chance to dominate in gaming.
From the article:
For our overclocked configurations, we enabled the DDR5-6000 EXPO profile for the memory kit.
They did testing with DDR5-6000.
 

tomachas

Distinguished
Oct 12, 2014
94
12
18,535
One more viewpoint.

View: https://www.youtube.com/watch?v=JM-twyjfYIw




Not exactly a glowing review.
It is kind of tempting to get 7600X until you do the math then DDR5 taxes you, then you need proper cooling to keep high ghz count if you don't I am not sure why then buy this CPU. Extra few hundred dollars is a lot and in my case better spend them elsewhere. At this point 12600K makes perfect sense.
 
As expected, the 7600X is kind of screwed until B650 comes out. The future doesn't look too bleak for AMD all things considered taking into account two important things:
1.- VCache will launch early next year and we, more or less, know the price points they'll use. They better justify their price hikes though.
2.- Platform cost will, most likely, go down from here. The same thing happened with X570 and Z690 as well. Nothing to be super alarmed as this is just early adopter tax and it happens all the time.

What matters here, the performance, is there. That's good. Also, as a side note, most numbers from AMD slides were verified and this just goes to show that AMD, at least for the time being, can be trusted somewhat. The benefit of the doubt in their claims was deserved.

I still don't like the "let's go BRRRR to 95°C", but we'll have to see how bad it really is down the line. My HTPC with the 5800X3D does get toasty, but not 95°C toasty... It hovers around 85°C, which I think it's "fine", but not ideal... Maybe I need to adjust my expectations on temps?

Also, I did watch der8auer's delid and that honking slab of metal does indeed hamper temps, which is a darn shame; I was expecting it, but looks like AMD could have done way better there. Looks like der8auer will start selling his delidding kits again, LOL. I wonder if AMD will have the same bending problem with the socket? Doesn't seem like it or at least no one has mentioned it yet.

EDIT: Good timing on this one.
View: https://www.youtube.com/watch?v=k-kdEVfF2T8

EDIT2: As it wasn't linked, der8auer's delid video, which is hella interesting!
View: https://www.youtube.com/watch?v=y_jaS_FZcjI


Regards.
 
Last edited:
  • Like
Reactions: Roland Of Gilead
The operating temps (as "intended" by AMD) are indeed a really turn down. Either they are safe or not, time will tell, I do not think many people will like to have such heater on their systems, with the more expensive cooling solution need it to keep it under 95°C.

What I would love to see next is what are the ways (if theres any) to make this cpus run a bit cooler (other than the ECO Mode or direct contact like Der8uer showed) even if it means losing a little bit of performance.

But yeah I would ratter upgrade to a 5600/5600X or 5700X no my current mobo, or get an Intel 12400/12600/12700, than to this Zen4 for now.

About the platform cost, those mobo price do look expensive compared to what we are used to see. The main problem here is that AMD said they will keep the AM5 socket till 2025, but far as I know they never said you will be abel to use the last AM5 zen "x" chip on a X670E/X670/B650E/B650 mobo.

If I knew I would be able to install the newest Zen x CPU in 2025 in a motherboard I bought in 2022, then that would make a small difference in the high cost platform perception.

For now, with the information I got so far, those 95°C and the high power consumption (compared to Zen 3) are really, really bad deals for me.
 
2.- Platform cost will, most likely, go down from here. The same thing happened with X570 and Z690 as well. Nothing to be super alarmed as this is just early adopter tax and it happens all the time.

What matters here, the performance, is there. That's good. Also, as a side note, most numbers from AMD slides were verified and this just goes to show that AMD, at least for the time being, can be trusted somewhat. The benefit of the doubt in their claims was deserved.

Well said!
 

KyaraM

Admirable
Thoroughly reading TPU's test of the CPU kinda shows a reverse of this test in gaming here, which is very interesting. Here, even the 12700K beats any of the new Ryzen 7000 chips by 2% in average gaming performance. The 3080 used holds back the tested CPUs at times even in 720p, but most of the time, Alder Lake seeme to actually lead. This means that it is really game dependent who is better, just like with the 5800X3D. GN also showed Alder Lake leading in specific gaming benchmarks. Power consumption in gaming is at Alder Lake level in the test, again at lower performance as a whole. All in all, that turns results from "expected" to "disappointing" for me, at least for gaming. The chips are very strong in applications for the most part, but the real competition is yet to come. Raptor Lake isn't even released yet so it's still early to judge; however, people might maybe shouldn't celebrate too early. Full test by TPU here:

Looking forward to their 50 games test suite, that will certaily be enlightening.
 
  • Like
Reactions: alceryes
Also the chonky IHS on these add 20c+ to the temps; der8auer video.
One of the main reasons for the IHS chonkyness is to somewhat alleviate offcenter hotspots. Some Ryzen 5000 coolers actually had an issue with this. The missing surface area due to those SMDs definitely hurt thermals but the flange pieces (legs) of the IHS also provide more surface area to help with cooling.
As far as the 95ºC goes, that isn't due to the IHS. AMD has completey changed how their chips work from a thermal perspective. Even if AMD had their old IHS these chips would still march right up to 95ºC before starting to throttle. This is now by design.
 
Thoroughly reading TPU's test of the CPU kinda shows a reverse of this test in gaming here, which is very interesting. Here, even the 12700K beats any of the new Ryzen 7000 chips by 2% in average gaming performance. The 3080 used holds back the tested CPUs at times even in 720p, but most of the time, Alder Lake seeme to actually lead. This means that it is really game dependent who is better, just like with the 5800X3D. GN also showed Alder Lake leading in specific gaming benchmarks. Power consumption in gaming is at Alder Lake level in the test, again at lower performance as a whole. All in all, that turns results from "expected" to "disappointing" for me, at least for gaming. The chips are very strong in applications for the most part, but the real competition is yet to come. Raptor Lake isn't even released yet so it's still early to judge; however, people might maybe shouldn't celebrate too early. Full test by TPU here:

Looking forward to their 50 games test suite, that will certaily be enlightening.
A lot of these scores are within the 3% margin of error. I wouldn't call it a win (for either AMD or Intel) for anything that lands within 3%. If we take away all those margin of error scores both AMDs new flagship chip and Intel's last gen flagship chip seem to be close to functionally equivalent (in gaming, at least). Steve talks about this in his test suites over at Gamers Nexus. Only CSGO, Rainbow 6, and maybe Far Cry 6 offered some meaningful differences but it's very difficult to create a CPU bottleneck with these powerful chips.

This may be where the super high frequencies of Intel's new Rocket Lake chips show a difference. 'Course, then we have AMD's X3D chips, etc., etc. ;)
 
One of the main reasons for the IHS chonkyness is to somewhat alleviate offcenter hotspots. Some Ryzen 5000 coolers actually had an issue with this. The missing surface area due to those SMDs definitely hurt thermals but the flange pieces (legs) of the IHS also provide more surface area to help with cooling.
As far as the 95ºC goes, that isn't due to the IHS. AMD has completey changed how their chips work from a thermal perspective. Even if AMD had their old IHS these chips would still march right up to 95ºC before starting to throttle. This is now by design.
Sorry, but the chonky IHS is to keep backwards compatibility, which don't get me wrong, is a great idea. The reality/consequence of that decision was to have that chonky slab of metal on top and it turns out it is hampering temps to the tune of at least 10°C and up to 20°c (taking into account der8auer used his latest and greatest liquid metal thingy). That's a massive increase in operating temps for the CPUs.

Now, to your point, AMD does indeed said "no problem with these", but der8auer demonstrated there is performance left on the table because of the chonkier IHS. How much? Well, it's actually a non-trivial amount TBH. The 7600X was able to boost even higher to 5.4Ghz all-core and stay under 80°C, so that is actually a good 200Mhz by just improving the IHS/temps alone. That's just bonkers in a good way, but bad for the IHS showing.

Regards.
 
Sorry, but the chonky IHS is to keep backwards compatibility, which don't get me wrong, is a great idea. The reality/consequence of that decision was to have that chonky slab of metal on top and it turns out it is hampering temps to the tune of at least 10°C and up to 20°c (taking into account der8auer used his latest and greatest liquid metal thingy). That's a massive increase in operating temps for the CPUs.

Now, to your point, AMD does indeed said "no problem with these", but der8auer demonstrated there is performance left on the table because of the chonkier IHS. How much? Well, it's actually a non-trivial amount TBH. The 7600X was able to boost even higher to 5.4Ghz all-core and stay under 80°C, so that is actually a good 200Mhz by just improving the IHS/temps alone. That's just bonkers in a good way, but bad for the IHS showing.

Regards.
Yup. I knew about the cooler compatibility. I'll see if I can find what I read about the offset hotspot issue that the thicker IHS helps with.

That is a HUGE difference with the delid. No offence to der8auer, but I'm going to take his results with a grain of salt. He's got financial incentive to showcase the difference. I need to see several other delid results giving the same amount of benefit.
 
  • Like
Reactions: -Fran-
Up to 8 cores, so looking at real against real cores, zen 4 is on par with alder lake, so as long as raptor has zero IPC increase and zero clock increase AMD will be fine.
The only mystery is how close the additional e-cores are going to make the final score for multithreaded workloads.


(PC world review)
18_Ryzen_9_7950X_Thread_Scores_R23_vs_12900K.png
 
Sorry, but the chonky IHS is to keep backwards compatibility, which don't get me wrong, is a great idea. The reality/consequence of that decision was to have that chonky slab of metal on top and it turns out it is hampering temps to the tune of at least 10°C and up to 20°c (taking into account der8auer used his latest and greatest liquid metal thingy). That's a massive increase in operating temps for the CPUs.

Now, to your point, AMD does indeed said "no problem with these", but der8auer demonstrated there is performance left on the table because of the chonkier IHS. How much? Well, it's actually a non-trivial amount TBH. The 7600X was able to boost even higher to 5.4Ghz all-core and stay under 80°C, so that is actually a good 200Mhz by just improving the IHS/temps alone. That's just bonkers in a good way, but bad for the IHS showing.

Regards.

Its interesting, whatever the reason was it was the wrong one for me.

If they used the chonky IHS to keep backwards compatibility with current AM4 coolers, then thats a backfire for a whole lot of people. Most people runing a Zen, Zen+, Zen 2 or Zen 3 cpus are most likely not using a high-end cooling system (meaining no 280mm radiators or higher, or the likes of the NH-D15).

If this was AMDs idea to not add another cost to the new platform then it was the wrong idea considering how this new CPU seems to work.

I hope in the next weeks they figure out a way to extract decent performance out of this chips without needing to go as high as 95°C under a 280mm radiator.
 
Up to 8 cores, so looking at real against real cores, zen 4 is on par with alder lake, so as long as raptor has zero IPC increase and zero clock increase AMD will be fine.
The only mystery is how close the additional e-cores are going to make the final score for multithreaded workloads.
From what the leaks and Intel are touting, Raptor will beat AMD in gaming, due to the crazy clocks the've got the P-cores up to. Until the next gen AMD X3D chips come out, that is.
It does look like, at the top end, ALL of these flagship chips are too powerful to show any meaningful FPS difference in games though.
 

hannibal

Distinguished
The point of these CPUs is that the more powerfull cooler system you have, the more power they will consume, because there is more headroom in temperature.

If you use cheap cooler, the CPUs use less power, because the heat limit becomes earlier and the CPU can not boost as high!

Lets hope that we get bios options to reduce the temperature limit if we want to do it so! I would consider running this at 80 degree of celcius to limit the power usage and clockspeeds.

But using normal cheap cooler also help!
 

KyaraM

Admirable
A lot of these scores are within the 3% margin of error. I wouldn't call it a win (for either AMD or Intel) for anything that lands within 3%. If we take away all those margin of error scores both AMDs new flagship chip and Intel's last gen flagship chip seem to be close to functionally equivalent (in gaming, at least). Steve talks about this in his test suites over at Gamers Nexus. Only CSGO, Rainbow 6, and maybe Far Cry 6 offered some meaningful differences but it's very difficult to create a CPU bottleneck with these powerful chips.

This may be where the super high frequencies of Intel's new Rocket Lake chips show a difference. 'Course, then we have AMD's X3D chips, etc., etc. ;)
This is correct, though the difference between the 12900K and the 7950X is actually over 4%; the 2% is for the 12700K. Differences are greater with the smaller chips, naturally. The thing is. It's not actually Alder Lake vs Raphael. It's Raptor Lake vs Raphael... and those did show improvements over Alder Lake in leaks etc. It would be very weird if it was different. So parity in gaming between Alder Lake and Raphael is actually bad for AMD. It gives Intel a real shot there. Also of note is the energy inefficiency in gaming when comparing especially the 12700K to the bigger Raphael chips. Only way to make them more efficient was to castrate the 7950X, which further reduced FPS. And you can do that with Alder Lake, too...
 
Lets hope that we get bios options to reduce the temperature limit if we want to do it so! I would consider running this at 80 degree of celcius to limit the power usage and clockspeeds.

But using normal cheap cooler also help!
At 80 degrees the CPU will never hit the thermal wall and so will always use as much power as it can find in the attempt to get to the thermal limit, or it will stop if it hits the power limit.
This turbo setting needs to have an off setting for it to stop trying that but then you are constrained to a locked all core clock which is going to cost you a lot of boosting in lesser threaded stuff.
 
Up to 8 cores, so looking at real against real cores, zen 4 is on par with alder lake, so as long as raptor has zero IPC increase and zero clock increase AMD will be fine.
The only mystery is how close the additional e-cores are going to make the final score for multithreaded workloads.


(PC world review)
18_Ryzen_9_7950X_Thread_Scores_R23_vs_12900K.png
That's actually a really good graph and, I can say right now (using it as a base), Raptor Lake will win in lower threaded, but will lose on the higher threaded part, unless they increase all-core clocks of E-cores significantly; this will mean increasing the overall package power and, from what I've read, they will only increase if by 10 watts or so? I also like how linear the scaling of the E-cores look, but we know after a certain amount they won't add as much due to clocks going lower (that's what basically happens with AMD in the graph) if Intel wants to keep power in check. I wonder how motherboards will handle that when Raptor Lake releases.

Regards.
 
  • Like
Reactions: alceryes
unless they increase all-core clocks of E-cores significantly; this will mean increasing the overall package power and, from what I've read, they will only increase if by 10 watts or so?
If they have zero efficiency improvement then 10W more will not do much, it will run slower but more cores might still be better than fewer but faster.

Also we all know that reviewers will run them without any power limit so the actual power draw might be very high but also give very high clocks.
 
This is correct, though the difference between the 12900K and the 7950X is actually over 4%; the 2% is for the 12700K. Differences are greater with the smaller chips, naturally. The thing is. It's not actually Alder Lake vs Raphael. It's Raptor Lake vs Raphael... and those did show improvements over Alder Lake in leaks etc. It would be very weird if it was different. So parity in gaming between Alder Lake and Raphael is actually bad for AMD. It gives Intel a real shot there. Also of note is the energy inefficiency in gaming when comparing especially the 12700K to the bigger Raphael chips. Only way to make them more efficient was to castrate the 7950X, which further reduced FPS. And you can do that with Alder Lake, too...
Yup. For gaming and low-ish core count workloads, Intel's Raptor will probably win.

As I said, I may actually wait for the X3D variants to come out. I mean, anything will show a nice boost over my current 9900k but I want gaming longevity. X3D chips are where that's at!
 
  • Like
Reactions: digitalgriffin
If they have zero efficiency improvement then 10W more will not do much, it will run slower but more cores might still be better than fewer but faster.

Also we all know that reviewers will run them without any power limit so the actual power draw might be very high but also give very high clocks.

AMD boost to 95C automatically. It's already at it's limits out of the box unless you go some exotic LN cooling.
 

Oldcompsci

Reputable
Apr 30, 2019
33
21
4,535
So my Ryzen 7 5700x (65W TDP) with 64GB ram and RTX 3050 card will cover my needs for quite a while then ... awesome since my money is continually depreciating in value right now, I'm going to have to string out any future upgrades.