Framechasers cringelord heard you . It is hilarious since it's coming from a guy who is selling gaming tweaking/OC guides for the most expensive CPUs. Also he complains about something he can only detect ("AMDip") lolView attachment 111202
What a ride it has been. We went from the haters celebrating Zen5% and how lame that is because we needed more gaming performance than that generationally. To Arrow coming out and being a disaster. So the cope kicked in hard with the "People do more than game on their PC you know!"
Then the 9800X3D launching, and suddenly none of us need that much CPU for gaming.
"It doesn't matter at 4K!" "It doesn't matter if you don't have a 4090!" I am getting second hand embarrassment from reading it everywhere.
Yeah, definitely. Shows how limited zen 5 is by the memory subsystem. Strix halo will be a bit more unchained in that regard and I'm expecting strong productivity performance without the need for 3D vcache.have to say, the eye popping Gaming performance seems to have shadowed the impressive general application perf IMO
I haven't had a chance to pour over data from the 7800X3D in huge detail, , but just from hindsight and what I have seen the standout for me is how Zen 5 has responded to cache outside of gaming workloads. The effective 'ipc' increase seems notably more significant.. around 8% in CnC's Spec INT Testing, and in Computerbase's Tests for CB24 , we can see it's around 6% faster than the TDP unlocked 9700X in Multi core as they both sit at roughly 5.2Ghz under All core load in this case (conveniently - check their OG 9700X review)
In ST CB it sits around 4% higher iso clock estimated (The 9700X is 5.5% higher clocked, but only 2% higher performance) , Ditto WebXPrt
AMD's 9800X3D: 2nd Generation V-Cache
Following the first generation of V-Cache found in the Zen 3 and Zen 4 X3D SKUs, AMD is now following up with the second generation of V-Cache which is a major change for AMD in terms of packaging.chipsandcheese.com
AMD Ryzen 7 9800X3D vs. 7800X3D, 5800X3D & Core Ultra im Test: Anwendungs-Benchmarks
AMD Ryzen 7 9800X3D im Test: Anwendungs-Benchmarks / Multi-Core-Leistung / Single-Core-Leistungwww.computerbase.de
View attachment 111246
View attachment 111247
The end result is in MT in particular, some quite phenomenal overall performance really for an 8 core CPU @ 5.2Ghz .. It's like Zen 5 comes alive.
It would be great to see what it does with the wick wound down a bit too.
MSI 9950X3D show it will be worse in games than 9800X3D. Scheduling issues will undoubtedly cause anomalies that will bring its average vs 9800X3D down a bit.if leaks 9950X3D is true, and will be faster in games than 9800X3D, that also gonna be best universal CPU for gaming and work
Excellent post!What really needs to stop is the 10 versions of the same test.
Imagine a productivity suite that was just 10 different 3d renderers, most of which nobody uses. That is what CPU gaming benchmarks are pretty much, it just covers the AAA titles with very little covering the rest of the gaming market. GN atleast test Stellaris and some of the german sites test ANNO so there are a few places that at least throw in those sorts of titles but if you wanted a proper suite you would do something like the below
1 e-sports shooter. (CS2, Cod, Valorant)
1 e-sports MOBA. (LoL or Dota)
1 RTS (starcraft 2)
1 ARPG (path of exile, Diablo 4)
1 turn time test (civ 6, civ 7 next year)
1 sim rate test (HoI4, Cities skylines, Factorio, Satisfactory)
1 survival game (PubG, Rust)
1 Flight sim (MSFS, Xplane)
1 racing sim (ACC, iRacing)
1 Bethesda game (Modded Skyrim given that is more popular than Starfield or Starfield given that is the newest iteration of the engine)
1 CDPR game (CP2077 given that is their latest and is popular)
1 Larian game (BG3)
2 AAA games that are currently popular
That covers a 15 game suite and would give you a pretty broad look at a lot of different titles, some of which are AAA and some of which go beyond just looking at that.
Then imagine if there were 5 different outlets that each tested a different game in each category or where they do test the same game they each tested a unique scene. That would give you far more scope to make a purchasing decision.
It has gobs of memory bandwidth as well as a SLC.Yeah, definitely. Shows how limited zen 5 is by the memory subsystem. Strix halo will be a bit more unchained in that regard and I'm expecting strong productivity performance without the need for 3D vcache.
And ignoring the possibility of the CPU being just wasted piece of silicon after it's been zapped enough times by the ST boostsAs with most things in life, there is seldom a simplistic "best", and even against 7800X3D there are still some outliners where RPL is faster (ignoring power).
All measurements in programs were performed four times, while in computer games eight times (two separate rounds covering four passes, preceded by a platform reset). The result that differed the most from the rest was rejected, and then an arithmetic mean was taken from the rest. In the game performance graphs, the individual bars mean: the longest is the average number of frames generated per second (AVG FPS), the middle is the average value for the lowest frames per second (MIN FPS), the shortest are frames per second (1% LOW) that are below the average MIN FPS for 99% of the time. The above description may seem enigmatic to readers unfamiliar with the terminology, so 1% LOW should be read as a symptom of performance drops, stutters, drops, etc., if it significantly deviates from the MIN FPS indications.
...
DDR5 memories worked at 7000 MHz and latencies of 32-48-48-96, always working in asynchronous mode or Gear2 (1:2), which is mandatory for Intel LGA 1700 platforms and optional for the AMD AM5 platform. In the case of the AMD AM5 platform, the DDR5 7000 MHz CL 32-48-48-96 (1:2) settings are equivalent to the DDR5 6000 MHz CL 30-36-36-76 (1:1), as confirmed by THIS article ( LINK ) and the sample charts below. Therefore, I decided to use identical DDR5 RAM modules for both platforms. The problematic 1:1:1 configuration is no longer required to achieve optimal performance on the AMD AM5 platform, because the correlation between the Infinity Fabric bus, the memory controller and the RAM clock has changed compared to its predecessor. Therefore, the asynchronous mode in the AMD AM5 implementation is much more useful than before. The Infinity Fabric bus on the AMD AM5 platform worked at an automatic frequency (2000-2100 MHz),
...
Why are some games ray tracing enabled, when it is widely known that it additionally burdens the graphics card? This is partly true, but the above statement is imprecise, because ray tracing also affects the CPU utilization. The relationship between CPU/GPU depends on the implementation of ray tracing in specific titles, because in some cases the GPU utilization can even drop, while the CPU load increases. For example - Spider Man Remastered, Hogwarts Legacy or Dying Light 2 practically do not change the relationship between components, although enabling ray tracing reduces the overall performance, which allows you to check the configurations in the most demanding scenario.
...
In computer game tests, I use my own measurement locations, which I capture during long hours of sessions, but all of them are completely normal scenarios occurring during gameplay. I consistently avoid built-in benchmarks, because most of them do not reflect real conditions and distort the results, often being an element of manipulation by hardware manufacturers and drivers. Therefore, the best choice are test locations in popular locations or sequences played during the campaign, which cannot be skipped and/or affect the overall experience of the game. To increase the reliability of the measurements, I present my measurement locations below:
1:2 memory with 7000 mt/s -> straight to the trash bin Eastern Euro reviewers are like uber-pro Intel and pro-nVidia, as usualtuned settings
Insane if true. 9950X is going to be an absolute unit of a CPU, especially with new RTX 5090 coming soon. Sad times for intel though, they are being relegated to bulldozer levels of performance.Not sure if what he says is from rumor or talks with AMD (it which case he probably isn't supposed to be saying it) but Paul says 9900X3D and 9950X3D are to have vcache on both CCDs. Starts at correct time stamp, even with an image of cache on both.
I managed to get one here in Finland. I have two builds to do, one is for gaming and other one is for a photographer. This is nice, faster than 99050X.Amazon is now back to $479 even, and they are in stock ! newegg is sold out still.
Edit: But I just noticed " FREE delivery January 22 - 24 to amazon prime members"
They mesured CL30 6000 MT/S as well. But yeah win10 is strange and their overclocking part was trash IMO (compared what scatterbencher could achieve)1:2 memory with 7000 mt/s -> straight to the trash bin Eastern Euro reviewers are like uber-pro Intel and pro-nVidia, as usual
Only possible IF Zen 6 has more memory bandwidth available.What is most interesting to me (a non gamer) about the 9800X3D reviews is that it looks like the core architecture still has some room to grow. A few tweaks here and there to alleviate some bandwidth bottlenecks and it isn't hard at all to imagine a Zen 6 with 20-30% IPC uplift.
This thing really breaths well ..... and is very power efficient when it does so.
I am still thinking that the 285K CB 2024 (and Zen 5 with different memory speeds) is showing that bandwidth is more important in this CB than in 2023 and that 285K can move more than Zen 5 .... while Zen 5 clobbers 285K in latency to pretty much any memory.
Are these accurate conclusions?
No?And that will possible ONLY with new socket.
I really hope they do it. IMO it's the only way that makes sense.Insane if true. 9950X is going to be an absolute unit of a CPU, especially with new RTX 5090 coming soon. Sad times for intel though, they are being relegated to bulldozer levels of performance.
Only possible IF Zen 6 has more memory bandwidth available.
Latency is not the limiting factor, but bandwidth starts to be.
And that will possible ONLY with new socket.
and yet they test Intel 13th and 14th with unlimited settings, we all know what that means: inflated results from a review pov1:2 memory with 7000 mt/s -> straight to the trash bin Eastern Euro reviewers are like uber-pro Intel and pro-nVidia, as usual
Framechasers cringelord heard you . It is hilarious since it's coming from a guy who is selling gaming tweaking/OC guides for the most expensive CPUs. Also he complains about something he can only detect ("AMDip") lol
View attachment 111252
I really think hes just speculating / pulling that out of his arse. Personally I think there is .001% chance that they do that...UNLESS theyve figured out a way to UNIFY both CCDs across a single huge L3 and choose to do it this gen. I think the chances are even less of that though.Not sure if what he says is from rumor or talks with AMD (it which case he probably isn't supposed to be saying it) but Paul says 9900X3D and 9950X3D are to have vcache on both CCDs. Starts at correct time stamp, even with an image of cache on both.
Why not? They would "waste" less X3D chiplets per CPU, but they'd literally leave performance on the table. Keep the scheduling issues (that are a dealebreaker for some). And keep the 7900X3D as a zombie unsellable CPU ...I really think hes just speculating / pulling that out of his arse. Personally I think there is .001% chance that they do that...UNLESS theyve figured out a way to UNIFY both CCDs across a single huge L3 and choose to do it this gen. I think the chances are even less of that though.