AMD Ryzen 7 9800X3D Linux Performance: Zen 5 With 3D V-Cache
(phoronix.com)149 points by mfiguiere 7 days ago
149 points by mfiguiere 7 days ago
I'm curious to see if AMD will release a 9950X3D this time around. I can foresee that kind of CPU dominating everything else across most workloads given how good this 8-core is holding up against CPUs with double or more cores.
I have a 5950x that is now a few years old and I planned to upgrade to a 9950x.
I have never had one of the 3D V-Cache processors and am curious how it would improve the benchmarks for my multi-threaded data management system that does many operations against a set of 4K blocks of data.
I heard rumors that a 9950x3D version will be available in January. I am trying to figure out if I should wait.
While true, also keep in mind that the iPad Pro (M4) which has no active cooling, and uses only 1/4th the power ... is still faster (single & multicore) than this 9800X3D - and it's also been on the market for 1/2 year now already.
In multi-threaded workloads, the M4 gets 13,380, the 9800X3D gets ~19,000 (varies by build), and the 9950X gets 22,000-24,000 depending on build.
The M4 Max you can pre-order gets around 26,000 multicore but is significantly more expensive than the 9950X ($569) or 9800X3D ($479). The M4 Max is a $1,200 premium over the M4 on the 14 inch MacBook Pro and a $1,100 premium over the M4 Pro on the 16 inch.
The M4 Max is only available in the MacBook Pro at present. The Mac Mini and iMac will only get the base M4. The Mac Studio is still based on the M2.
This is just a summary of performance and cost. Portability, efficiency, and compatibility factors will weigh everyone's choices.
UPDATE TO MY COMMENT: The new small Mac Mini does have an option for the M4 Pro but not the M4 Max. For the curious, the M4 Pro supposedly gets around 22,000 in Geekbench. It's an $800 premium over the base M4 Mac Mini while adding 8GB of RAM and 256GB of storage.
Single core yes, but multi core no.
The Geekbench scores cannot compare laptop CPUs with desktop CPUs, because the tasks that are executed are too short and they do not demonstrate the steady-state throughput of the CPUs. The desktop CPUs are much faster for multithreaded tasks in comparison with laptop/tablet CPUs than it appears in the GB results.
The Apple CPUs have a much better instructions-per-clock-cycle ratio than any other CPUs, and now in M4 they also have a relatively high clock frequency, of at least 4.5 GHz. This allows them to win most single-threaded benchmarks.
However the performance in multi-threaded benchmarks has a very weak dependence on the CPU microarchitecture and it is determined mostly by the manufacturing process used for the CPU.
If we were able to compare Intel, AMD and Apple CPUs with the same number of cores and made with the same TSMC process, their multithreaded performance would be very close at a given power consumption.
The reason is that executing a given benchmark requires a number of logic transitions that is about the same for different microarchitectures, unless some of the design teams have been incompetent. An Apple CPU does more logic transitions per clock cycle, so in single thread it finishes the task faster.
However in multithreaded execution, where the power consumption of the CPU reaches the power limit, the number of logic transitions per second in the same manufacturing process is determined by the power consumption. Therefore the benchmark will be completed in approximately the same number of seconds when the power limits are the same, regardless of the differences in the single-threaded performance.
At equal power, an M4 will have a slightly better MT performance than an Intel or AMD CPU, due to the better manufacturing process, but the difference is too small to make it competitive with a desktop CPU.
> The Geekbench scores cannot compare laptop CPUs with desktop CPUs, because the tasks that are executed are too short and they do not demonstrate the steady-state throughput of the CPUs. The desktop CPUs are much faster for multithreaded tasks in comparison with laptop/tablet CPUs than it appears in the GB results.
Bullshit. What you're talking about is the steady-state of the heatsink, not the steady state of the chip. Intel learned the hard way that a fast CPU core in a phone really does become a fast CPU core in a laptop or desktop when given a better cooling solution.
> However in multithreaded execution, where the power consumption of the CPU reaches the power limit, the number of logic transitions per second in the same manufacturing process is determined by the power consumption. Therefore the benchmark will be completed in approximately the same number of seconds when the power limits are the same, regardless of the differences in the single-threaded performance.
No, microarchitecture really does matter. And so does the macro architecture of AMD's desktop chips that burn a huge amount of power on an inefficient chip to chip interface.
For an apples to apples comparison, you'll need to compare Zen 5 with M3, or whatever Zen 6 is going to be with M4.
Apple is paying for exclusive access to TSMC's next node. That improves their final products, but doesn't make their architecture inherently better.
Do you though? M4 is what is on the market now and this chip is just coming out. Maybe they are on different processes, but you still have to compare things at a given point in time.
Why would a consumer care about what node something is on? You should only care about a set of processors that is available in the market at the same time. The M4 is available now and Zen 6 is not. Once zen 6 is here we probably have an M5.
> for single threaded 9800x3d scored about 3200, whilst m4 had 4400... The m4 is so far above the rest it's ridiculous.
Except the fact that your computer runs more than one thread. Pity that this "single core" performance cannot be utilized at its maximum potential.
And the OS is terrible, so it's practically useless for me.
Hehe ... yeah, single threaded, in some benchmarks. Very impressive chip, the M4. Multi-threaded loads that take more than 30 seconds, no way, come on. But to see the X3D chips really shine above their competitors, you need to slot in a high-end graphics card, and load up a ... uh well you can't compare to Apple Silicon at that point ...
https://tpucdn.com/review/amd-ryzen-7-9800x3d/images/efficie...
Raw gaming performance increase is good but its gaming efficiency seems to have taken a dip compared to 7800X3D.
So AMD chose to decrease efficiency to get more performance this generation.
Source: https://www.techpowerup.com/review/amd-ryzen-7-9800x3d/23.ht...
The efficiency is only worse because the CPU can use the power without burning itself up unlike the last generation's X3D. And efficiency is always better at lower clocks. You can get this generation's efficiency uplift by limiting its power to the levels where last generation's CPU started throttling to keep its 89C Tjmax, but that will inevitably also limit the frequency that's the main performance uplift for the CPU
For comparison on how limited last gen's X3D was wrt power, tom's hardware has it on 71W with all core AVX, while my 7600X with 2 fewer cores consumes up to 130W
If I can summarize what you wrote: Same IPC gain as normal Zen5 but more power can be drawn to increase performance due to moving the cache chiplet to the bottom.
The previous 3D cache solutions were not just limited thermally, but also the cache chiplet could not tolerate the high voltages that AMD's CPU cores use at high frequencies. Even with excellent cooling, you weren't going to get a 7800X3D or 5800X3D to match frequencies with the non-3D parts. (This might have been less of a problem if AMD could put the extra cache on a different power rail, but that's hard to retrofit into an existing CPU socket.) This new cache chiplet still has a lower voltage limit than the CPU cores, but it's not as big a disparity.
9800X3D is supposed to have Eco mode with a lower TDP cap, similarly to other AMD processors. I don't see it included in the initial reviews, but it would be curious to see the followup data. If the history is anything to go by, it would significantly decrease the power consumption with only a marginal performance impact.
I have the 7950x, and if I set it to 65W eco mode, I still have basically the same geekbench score
65W: https://browser.geekbench.com/v6/cpu/6126001
105W: https://browser.geekbench.com/v6/cpu/5821065
I actually haven't tested it with 170W (which is the default for the 7950x) for whatever reason, but the average 7950x score on geekbench is basically the same as my geekbench scores with lower than normal TDP.
https://browser.geekbench.com/processors/amd-ryzen-9-7950x
I wouldn't be surprised if the same is possible with the newer CPUs.
Nice added bonus is that my PC fans barely spin (not at audible speeds)
Bad arch decision are punishing. AMD was absolutely dwarfed in the early core iX days and never really came back until Ryzen. The whole bulldozer linage was DoA to the point Opteron just never factored in.
Hopefully Intel pull something out again but they look asleep a the wheel.
For a long time, x86 chips are happy to give you a little more performance for a lot more watts at the top end of the performance chart.
Watts/fps @ max fps makes for an interesting graph, but not a very clear comparison. It would be better to compare watts used when locked at a given fps, or fps available when locked at a given wattage. Or watthours to do a video encode (with max wattage, and at various watt limits).
Nice to actually have a decent release this generation of CPUs.
The rest of Zen5 was maybe a 5% bump on average, and Intel's new series actually regressed in performance compared to 14th gen.
Seems like the Zen5X3D's will be the only good parts this time around.
Hardware Unboxed has the interesting theory that the I/O die, which is unchanged between Zen4 and Zen5, is a significant bottleneck especially for the latter. The 3D v-cache would then ease the pressure there, and so see the cpu get an extra boost beyond that expected from increased cache.
To cut Intel some slack, this latest version overhauls their old architecture, and they were fairly upfront about the lack of development in performance in this generation.
The idea is the new platform will allow for better development in future, while improving efficiency fairly significantly.
From a consumer standpoint - this doesn't matter. You can't buy that future product that may exist. You can only choose whether to buy the current product or not. And right now, that product is bad.
I certainly hope the next generation is a massive bump for Intel, but we'll see if that's the case.
I think the new T-equivalent CPU could be very interesting if Intel releases one. Those variants are optimized for 35W TDP, and they can be used for building high-performance fanless systems that can sustain their performance for quite some time. The lower power requirements for Arrow Lake might be a really good match there.
it's scrapped, the new design: https://www.pcworld.com/article/2507953/lunar-lakes-integrat...
So why buy this generation and not wait unless your computer broke and you NEED Intel?
> To cut Intel some slack, this latest version overhauls their old architecture...
... and their 13th/14th generation processors had serious problems with overvoltage-induced failures - they clearly needed to step back and focus on reliability over performance.
9800X3D looks like an all-around winner, so if you don't mind spending $500 on just the CPU, I don't see why anyone would get anything else.
All-around winner in what? For $500 you can get a lot more cores.
All-around winning, $500, 8 cores makes no sense.
This thing has a premium gaming price tag because there is nothing close to it other than their own 7800X3D.
In theory, yes. But in the real world the bottleneck of the same 128 bit wide memory, interface that's been popular way back since the time of dual core chips.
Less cache misses (on popular workloads) helps decrease power and increase performance enough that few things benefit from 12-16 cores.
Thus the M3 max (with a 512 bit wide memory system) has a class leading single core and multi-core scores.
I'm not so sure about memory actually being the bottleneck for these 8 core parts. If memory bandwidth is the bottleneck this should show up in benchmarks with higher dram clocks. I can't find any good application benchmarks, but computerbase.de did it for gaming with 7800MHz vs 6000MHz and didn't find much of a difference [1]
The apple chips are APUs and need a lot of their memory bandwidth for the gpu. Are there any good resources on how much of this bandwidth is actually used in common cpu workloads? Can the CPU even max out half of the 512bit bus?
[1] https://www.computerbase.de/artikel/prozessoren/amd-ryzen-7-...
What would you suggest instead?
It is pretty competitive on the Multi-Core rating: https://browser.geekbench.com/v6/cpu/8633320 compared to other CPUs: https://browser.geekbench.com/processor-benchmarks
As a C++ programmer, I just bought a 9900X for my first PC build. Sure, it won't game as well, but I like fast compile times, and the 9900X is on sale for $380 right now. That's $100 cheaper than the 9800X3D launch price.
Yeah, these Zen 5 are killer for that kind of workload. I also replaced my workstation with a 9900-series CPU since my Intel 14900K fried itself, and I am very pleased with every aspect, except idle power consumption which is a minor drawback.
It looks like the X3D is no better than the 9900X for non-game single-threaded workloads like browsers, and it's much worse than the 12 or 16 core parts in terms of overall throughput, so for a non-gamer the plain X seems much better than the X3D.
What's your idle power consumption for AMD vs Intel if you don't mind me asking? I'm getting avg 125W for my 13900k build, measured at the wall and it mildly bugs me when I think of it, I thought it'd be closer to 80. And power is very expensive where I live now.
I'm about to build a new system and am planning on using the 9900X. It's primarily for coding, Adobe CC, and Ableton, with maybe a rare gaming session here and there. It seems that the 9900X is the best bang for the buck right now. It games just fine, BTW.
Intel can still be kind of faster for "productivity" stuff? At least if you are willing to pay for the >8000 MHz CUDIMMs (which i don't think AMD even supports at full speed?) which can result in pretty impressive performance. Of course the value/price is probably not great...
The last Intel machine I will ever build was my 13900K, primarily because I liked the fact that I could use cheaper DDR4 memory.
Next rig and everything for the forseeable future will be AMD. I've been a fanboy since the Athlon XP days - took a detour for a bit - but can't wait to get back.
Even if Intel wasn't chugging so badly right now, their recent handling of the overvoltage and oxidation fiasco where they only thought about covering their asses instead of working the problem would leave me with a pretty disgusting taste in my mouth if I bought anything Intel for the foreseeable future. Customer relations should mean something, just look at Noctua.
> I've been a fanboy since the Athlon XP days - took a detour for a bit - but can't wait to get back.
Same. But already built a 3700X and then a 7700X.
I've got this feeling the wife she's gonna upgrade her 3700X to a 7700X soon, meaning I'll get build a 9000 series AMD!
The results for decompression, but no compression, are all surprisingly bad compared to other benchmarks, how comes? For example 7-zip decompression performs worse than my 7700X (84 K mips vs 93 K mips on my 7700X). Other decompression benchs are equally depressing. But compression performs as expected (as much as 30% faster than my 7700X).
What can explain those disappointing results but only on decompression?
Sharing links from websites with intrusive video advertisements should be prohibited. The websites should be banned, and those who share links to them should receive a paddling.
This is precisely what Scroll (1) used to do. It seems it didn't end up well, unfortunately.
I am surprised at how much this thing is just straight up crushing it with just 8 cores.
I think it topping the machine learning benchmarks has to do with having only 8 cores to share the 96MB of L3 cache, which ends up having a ratio of 1core having 1MBL2 + 12MB L3 which is huge, that means EACH THREAD has more cache than i.e the entire nvidia 3090 (6mb l2 total), and this ends up taking FULL advantage of the extra silicon of various avx extensions.