AMD Ryzen 5 9600X Review, Extremely BAD Value!

[Kenrou]

Actually slower than the 7600x in some games/apps, WTH?

00:00 - Welcome to Hardware Unboxed
00:31 - Ryzen 5 9600X
01:20 - Test System Specs
01:28 - Zen 5 is not as efficient as you’ve been told!
03:14 - Cinebench 2024
03:52 - Cinebench 2024 Power
04:07 - 7-Zip File Manager
04:27 - Blender Open Data
04:43 - Corona 10 Benchmark
05:03 - Adobe Photoshop 2024
05:21 - Adobe Premiere Pro 2024
05:34 - Baldur's Gate 3
05:48 - The Last of Us Part 1
06:02 - Cyberpunk 2077: Phantom Liberty
06:23 - Hogwarts Legacy
06:34 - Assetto Corsa Competizione
06:53 - Spider-Man Remastered
07:03 - Homeworld 3
07:11 - A Plague Tale: Requiem
07:25 - Counter-Strike 2
07:48 - Starfield
07:58 - Horizon Forbidden West
08:09 - Hitman 3
08:17 - Watch Dogs: Legion
08:26 - 13 Game Average
08:53 - Power Gaming
09:27 - Cost per Frame CPU
09:57 - Cost per Frame Upgrade
10:16 - Final Thoughts

TLDR:

 

[mackerel]

Don't start me on another rant about cost per frame being a bad metric.

I don't think I mentioned it on this site yet. PCWorld posted a long video on why they're not publishing a review yet. They did the testing, but found their results about 10% short of AMD's double check values. It's a long video and doesn't sound like they fully understand what's going on even after contacting AMD about it. They said GN found their 9600X 5% short, which they said could have been in tolerance territory, but there might be something going on with BIOSes or elsewhere causing lower than expected perf. I've not compared test results between different reviewers that might give a better picture if these are outliers or are they generally consistent. This is one to watch.

BTW it looks like I'm saved from temptation. Doesn't look like anywhere I looked at in UK will have stock for another couple weeks. Either it sold out in the 40 minutes between release and me looking, or it is a paper launch for this region.

 

[Woomack]

I'm only curious if memory controllers are really better and if motherboards can handle 8200+ stable. Then I would rerun everything for RAM reviews on AMD as Intel is acting weird at some settings, and I tested way too many 13/14 gen CPUs already. I haven't seen any leaks about Ryzen 9k and its high RAM frequency yet, but maybe some will appear after the higher models premiere. Most reviewers have no idea how to overclock RAM, so they run basic tests at XMP/EXPO and skip everything else.
I was thinking about the 9950X, but if RAM won't be much better, then I may wait for the X3D version.

 

[]-[itman]

I'm only curious if memory controllers are really better and if motherboards can handle 8200+ stable. Then I would rerun everything for RAM reviews on AMD as Intel is acting weird at some settings, and I tested way too many 13/14 gen CPUs already. I haven't seen any leaks about Ryzen 9k and its high RAM frequency yet, but maybe some will appear after the higher models premiere. Most reviewers have no idea how to overclock RAM, so they run basic tests at XMP/EXPO and skip everything else.
I was thinking about the 9950X, but if RAM won't be much better, then I may wait for the X3D version.

It's the same memory controller as in Zen 4, that wasn't changed. The new boards coming out later this year might provide better signaling between the RAM and CPU, though, which may enable higher stable memory speeds. I don't really think it makes much difference, though, as the IF will be a bottleneck either way.

 

[mackerel]

	Cinebench 2024		Cinebench R23
	ST	MT	ST	MT
Anandtech	131	1111	2162	19538
Ancient Gameplays	134	1136
Bitwit			2214	20090
der8auer				19504
Eurogamer	130	1172
Gamer Meld	133	1122	2194	20018
Gear Seekers	135	1163
GN
Hardware Canucks	136	1219	2219	21069
HUB	136	1201
JayzTwoCents	132	1140	2197	20352
Kitguru		1177		20135
Level1Techs	134	1166
LTT	135	1187
Machines & More	135	1178	2208	19834
OC3D	135	1198	2224	20797
Paul's Hardware	137	1131	2222	19277
PC Centric	134	1184
PC Garage			2209	20878
PCWorld	132	1100
TechPowerUp	135	1208
Toms Hardware	132	1155	2219	20608
Average	133.9	1163.8	2206.8	20175.0

I went and skimmed a whole bunch of 9700X reviews (including some channels I never heard of before) to see how much results varied. Some tested more than one configuration, and in that case I picked whatever was closest to stock. No OC. Different sites have different policies on ram which could have some impact here. I ignored channels publishing results before yesterday's embargo.

While I looked at many tests, CB 2024 and R23 were the most common between them, and here are those results. Note R23 isn't very memory sensitive if at all, but 2024 definitely is. So expect more variation in 2024 MT depending on if the site used JEDEC or some blazing fast ram.

It was an eye opener even if sites used same software, there are still so many variables. For example, I thought Blender would be a good one. Nope. Different versions over time. Different sub tests. Results could be expressed as throughput or time, individually or aggregated. Some had their own custom renders. So at the end of the day, Cinebench was the most stable test there.

 

[Woomack]

It's the same memory controller as in Zen 4, that wasn't changed. The new boards coming out later this year might provide better signaling between the RAM and CPU, though, which may enable higher stable memory speeds. I don't really think it makes much difference, though, as the IF will be a bottleneck either way.

It's the same, but AMD specifies it higher somehow, so I wonder if anything has changed. They already made the 1:2 IF ratio work in the past and stabilized higher memory clocks in general, so who knows ...

 

[]-[itman]

It's the same, but AMD specifies it higher somehow, so I wonder if anything has changed. They already made the 1:2 IF ratio work in the past and stabilized higher memory clocks in general, so who knows ...

I think from AMD’s perspective, they just validated a higher officially supported speed of DDR5-5600.

 

[Woomack]

The new optimal memory clock at the recommended 1:1 ratio is supposed to be 6400 for the Ryzen 9000 when it was 6000 for the Ryzen 7000 series. Many Ryzen 7000 CPUs don't work stable at 6400 1:1. This is why I wonder if they improved something around the same IC or if it's only marketing talking. If there are improvements, then I doubt if it's anything significant, but nearly every Ryzen 7000 runs with memory at 8000 1:2, and it looks like the limit is more because of motherboards than CPUs. It's hard to believe every CPU hits a wall at the same memory clock, regardless of the motherboard and its brand. I tested multiple CPUs, and on 5 different motherboards, all were 100% stable with RAM at 8000 and couldn't even post at 8200. From other users' reports (those who actually know how to OC RAM and not some randoms from weird forums), it's about the same with one exception, which is Gigabyte B650E Tachyon that I had no chance to check. This is the only motherboard that runs with Ryzen 7000 higher than 8000, but it's also impossible to buy.

 

[SF101]

Mine (7600x) wont even post at 6400.
Wish they would just put the 9000X3D chips out already and stop sandbagging.
Who knows maybe ill gamble on Arrow lake I've done crazier buys before.

 

[Sologuy]

My previous 7600X which was lapped but not overkill with that mod could do 6400 1:1 stable in MTP with my Hynix M die kit. Current testing with 9700X (same RAM kit) is not doing well with 6400 1:1 in MTP.
But I reckon the agesa needs updating as its only in the first iteration of that code for Zen 5 compared to when I did the testing on the 7600X in the later parts of last year. I'm just bringing this point up as its claimed widely that Zen 4 & 5 share the same IMC. Also, not sure if windows is having issues with scheduling & that could be causing stability issues cause' I've heard a bit about that recently with Zen 5; or on the other hand MTP needs updating as it hasn't been updated since 22'. In any case I'll try again with 9700X at 6400 1:1 with prime 95 as that's updated fairly regularly. Can only do this when I find the time, post those results in another thread.

 

[Woomack]

We already have at least 2 AGESAs for 9K CPUs. BIOS releases with new CPU support are for two months, so they had the time for any tuning (it's hard to say if they did or want to keep it for new motherboards). I only wonder if there are any real changes, optimized code or anything like that, as AMD mentioned (I can't find it now) that the RAM can work higher on the 9K series. It was mentioned that 6400 1:1 will be the new sweet spot, while I still see DDR5-6000 as suggested by AMD in the reviewers pack (probably every CPU review sponsored by AMD uses a 6000 kit).
The new Ryzen DDR5 series, like the recently released G.Skill Royal Neo, is up to 6400 CL30. If IMC were any better, they would release 6600-6800 kits.

If I won't be forced to get a new CPU for motherboard reviews, I will keep the 7950X until the 9950X3D release.