AMD Ryzen 9 7900X CPU-Z Benched, Falls Short of Core i7-12700K in ST, Probably Due to Temperature Throttling

[Valantar]

All those improvements are huge? I'm seeing about 9-19% as the most apparent IPC boost in stuff you might actually use someday. Add 10% clockspeed and you're not in 'huge' territory at all.

1% is worst case, but 39% makes about as little sense and the average certainly won't be 20% either.

A 13% geomean uplift is ... fine? Decent? Something like that. It's well above what Intel tended to deliver in the 2nd-to-10th gen span, but that's about the best you can say about it. It's not unimpressive, but it isn't impressive either. These chips need the clock speed gains to show a major improvement - and 5nm does seem to deliver at that - but this doesn't make me all that hopeful for the future. Hopefully AMD has been working for a few years already on widening their cores (without this guzzling power). That's where Apple gets their massive IPC, and partially where ADL saw its biggest gains from (but also most likely its massive boost power needs).

 

[Tomorrow]

All those improvements are huge? I'm seeing about 9-19% as the most apparent IPC boost in stuff you might actually use someday. Add 10% clockspeed and you're not in 'huge' territory at all.

1% is worst case, but 39% makes about as little sense and the average certainly won't be 20% either.

Oh how times have changed. We used to have <5% gains with Intel. Now people complain when we get "only" 13%

 

[Bwaze]

And don't forget, that's "only 13%" in IPC alone, we get about 15% higher clocks too.

But people will of course compare it to newest Intel generation, which promises 20% higher gaming and 40% higher productivity scores for their flagship.

And AMD also has a problem of 5800X3D, which they will barely match in gaming with this brand new platform, with much more expensive CPUs, motherboards, RAM...

 

[AusWolf]

Amidst all this IPC comparison with the previous generation, let's not forget that it's not the previous generation that we should compare the performance increase to, but to the stuff that we own.

Does it perform as you'd expect? Is its power consumption and heat dissipation acceptable for your needs? Do you find the price of upgrade worth it? - these are the questions one should be asking.

Edit: 13% is really not a lot, but if you're coming from a 2700X, or let alone 1700X, the improvement will be significant.

 

[QuietBob]

The default CPU-Z benchmark (Version 17) measures FPU performance using SSE2 only. The ST score can be used to roughly estimate a CPU's performance in games, since most of them depend on high IPC and are limited by the main thread. As more and more titles employ the AVX set, it would be useful to compare the results using the new CPU-Z benchmark (Version 19), which can utilize up to AVX512.

As concerns the AMD slide: all the games presented here strongly depend on ST performance, hence the projected gains. I'm surprised GTA shows only marginal improvement, since it's a heavily ST bound title.

 

[Bwaze]

The ST score can be used to roughly estimate a CPU's performance in games, since most of them depend on high IPC and are limited by the main thread.

When 5800X3D trails in CPU-Z single core behind 5800X you know that this benchmark is a bad estimate on how games will run.

 

[rv8000]

This is, not the greatest article. Where’s all the supporting information for these claims? A few short weeks and the actual information will be available. Sad to see this kind of reporting here

 

[Makaveli]

Good it seems most people in the thread understand CPU Z scores don't mean much. Looking forward to the TPU review.

 

[AnarchoPrimitiv]

Seems like a mild case, maybe thanks to two CCXs
Future AM5 users - do prepare for a completely different heat management landscape.

Also do prepare to rethink if AM4 cooler Z height compatibility was worth the thermal sacrifice of such monstrously thick IHS

It's a catch 22, because we know how people hold AMD to standards they don't hold Nvidia or Intel too...(just remember how when they were condemned for raising prices on Zen3 despite having the best CPUs available....people expect AMD to compete with and beat Intel, and somehow still beat Intel on price despite literally Intel literally having an R&D budget 7.5x larger than AMD's)...if coolers weren't compatible with Zen4, they'd be screaming, Temps are high now, they're screaming....people want it both ways

 

[Vayra86]

Oh how times have changed. We used to have <5% gains with Intel. Now people complain when we get "only" 13%

Not complaining, just adding perspective. <5% is nothing. 13% is a bit more than nothing.

 

[phanbuey]

"In the multi-threaded test, however, the 7900X, with its 11822 points, is in the league of the next-generation Core i7-13700K (8P+8E) processor, which was recently spotted scoring 11877 points with a 6.20 GHz overclock."

So that leak has the P-Cores to 6.2Ghz. A stock 12900K gets 11700pts, so a 13700K at stock should beat that easily, and other leaks have also shown a score closer to 13000.

Intel Core i7-13700K has been overclocked to 6 GHz, in the CPU-Z ST test | dtechted

This last sentence just doesn't seem likely:
The 7900X will hence be pushed as a superior alternative to the i7-13700K for productivity and creator tasks, whereas its single-threaded score ensures that it falls behind the i7-13700K in gaming by a fair bit."


1. As @Valantar pointed out, CPU-Z is absolutely not indicative of gaming performance, so there's not a way to gauge that from this bench. The 5800x3d gets 639 ST score in CPU-Z and absolutely dominates in gaming, for example.
2. The 13700K has already posted higher MT scores in other leaks -it's EXTREMELY unlikely it will only be 3% faster in MT than a 12900k at stock.

If anything it will likely be the opposite - the 7900x and 13700k will trade blows depending on game, with a solid chance that 7900x is faster than the 13700k, and the 13700k will beat it in MT on non-avx 512 loads.

 

[HD64G]

They assume games are using single threaded tasks, which was true for older games but not for the new AAA ones like cyberpunk which makes full use of multi threaded CPUs

The slide has game IPC performance increases in it and it varies from 12-24%...

 

[Space Lynx]

View attachment 262213

So you can see that CPU-Z literally is the least representative of any benchmark for Zen4 IPC improvements. Don't forget this picture I posted above is IPC, once you add another 10 percent frequency all those improvements are huge.

CPU-Z is the least important metric. Least representative of the average improvement.

thank you for posting this, having the extra perspective was very helpful.

I don't like e-core shit, and I don't like the massive and consistent security flaws that Intel had year over year over year.... so I really want to stay with AMD.

 

[QuietBob]

When 5800X3D trails in CPU-Z single core behind 5800X you know that this benchmark is a bad estimate on how games will run.

Obviously, frame rates depend on multiple factors, not just the FPU. As evidenced by the 5800X3D, many titles can take advantage of greater bandwidth afforded by a larger cache. Still, the 3D variant remains a unique SKU, and games in general scale with ST performance.

 

[Valantar]

Obviously, frame rates depend on multiple factors, not just the FPU. As evidenced by the 5800X3D, many titles can take advantage of greater bandwidth afforded by a larger cache. Still, the 3D variant remains a unique SKU, and games in general scale with ST performance.

... but "ST performance" isn't a thing, it's a wildly complex combination of different factors in and of itself. The 5800X3D has better ST performance in many workloads precisely due to its cache - but it also has worse ST performance in others that can't utilize the cache, are more cache latency sensitive, or want higher clocks instead. Saying "games in general scale with ST performance" is perfectly in line with the 5800X3D being the fastest gaming CPU overall. And that's precisely the point.

As for it being a unique SKU - it is so far, but all signs point towards there being more X3D SKUs in the 7000 series.

 

[Dave65]

Waits for @GN and Hardware Unboxed reviews not this frivolous crap.

 

[QuietBob]

... but "ST performance" isn't a thing, it's a wildly complex combination of different factors in and of itself. The 5800X3D has better ST performance in many workloads precisely due to its cache - but it also has worse ST performance in others that can't utilize the cache, are more cache latency sensitive, or want higher clocks instead.

I agree. Rather than referring to abstract "absolute" ST performance of a CPU, we should be talking about its ST performance in a specific app. My point remains, though. Processors that score higher in CPU-Z ST benchmark tend to do better in games, with the notable exception of the 5800X3D. This is why I'm especially curious to see the new generation of V-cache on Zen 4.

 

[Speedyblupi]

CPU-Z also showed 5800X3D slower than normal 5800X in single and in multi core. It's very poor at representing anything meaningful.

To be fair, the 5800X3D isn't faster than the 5800X in most tasks except for gaming (though its lead in gaming performance is massive).

 

[Valantar]

To be fair, the 5800X3D isn't faster than the 5800X in most tasks except for gaming (though its lead in gaming performance is massive).

Gaming, and, ironically enough a lot of server/HPC workloads. Really odd how those seem to correlate - though it's by no means a 1:1 match. Few consumer apps care about the extra cache at all though.

 

[Jism]

probably a smaller die was a mistake

What mistake?

You got 8 cores crammed in a tiny little CCD with a bunch of cache that needs to be always powered on.

Once you start crossing the 5Ghz mark temps just fly upwards and it wont be better the higher or faster you go.

Both camps are pretty much at their limit to what they can do here really. Going smaller would only yield a more difficult source of heat to cool.

But really if i look at those graphs... 50 points behind on ST but a massive improvement on MT. I know my pick.

 

[AnotherReader]

A 13% geomean uplift is ... fine? Decent? Something like that. It's well above what Intel tended to deliver in the 2nd-to-10th gen span, but that's about the best you can say about it. It's not unimpressive, but it isn't impressive either. These chips need the clock speed gains to show a major improvement - and 5nm does seem to deliver at that - but this doesn't make me all that hopeful for the future. Hopefully AMD has been working for a few years already on widening their cores (without this guzzling power). That's where Apple gets their massive IPC, and partially where ADL saw its biggest gains from (but also most likely its massive boost power needs).

I remain cautiously optimistic about AMD finding average IPC increases for Zen 5 and other chips in the pipeline. AMD, unlike Intel and Apple, isn't focused on single threaded performance über alles. Their focus in on server performance, and since they use a single design for servers and desktops, they have opted to focus on a small core that can offer acceptable single threaded performance while delivering leading multi threaded performance for servers. They appear to have matched Alder Lake's single threaded performance with a much smaller core in terms of the out-of-order window by focusing on front-end and MLP (load/store) improvements. They can still use all the tricks that Intel has used for Golden Cove, but that would require a larger core.

This table comparing the out of order engines of Zen 3 and Alder Lake is reproduced from Chips and Cheese's excellent overview of Alder Lake/Golden Cove. I've changed the order of various entries, and eliminated some columns for brevity. The most important column is the last one; most structures in the out of order engine of Golden Cove are much larger than their counterparts in Zen 3, and even Sunny Cove (Ice Lake), has larger structures than Zen 3. Despite all that, Zen 3 beats all Intel architectures before Golden Cove.

Structure	Instruction affected if it..	Golden Cove Capacity	Zen 3 Capacity	Golden Cove vs Zen 3
Reorder Buffer (ROB)	Is waiting to retire (all)	512	256	2x
Load Queue	Reads from memory	192	116	1.5x
Store Queue	Writes to memory	114	64	1.78x
Branch Order Buffer	Affects control flow	128	48 Taken 117 Not Taken	Complicated, approximately 1.33x
Integer Register File	Writes to an integer register	280 (~248+32)	192 (173 measured+32?)	1.45x
Flags Register File	Sets flags (often tied to integer registers on x86)	248	121	2.04x
Floating Point/Vector Register File	Writes to a fp/vector register	332 (300+32)	160 (139 measured+32?)	2.07x
Total Scheduler Capacity	Is waiting on an execution unit	205	160	1.28x
Fill Buffer	Misses L1D	16	24	0.66x
Superqueue	Misses L2	48	64?	0.75x?

 

[DeathtoGnomes]

FPS is not an accurate measurement of performance, not alone. ST games are on the decline so less and less stock should be put into them. Game devs are finally getting the clue.

 

[Valantar]

I remain cautiously optimistic about AMD finding average IPC increases for Zen 5 and other chips in the pipeline. AMD, unlike Intel and Apple, isn't focused on single threaded performance über alles. Their focus in on server performance, and since they use a single design for servers and desktops, they have opted to focus on a small core that can offer acceptable single threaded performance while delivering leading multi threaded performance for servers. They appear to have matched Alder Lake's single threaded performance with a much smaller core in terms of the out-of-order window by focusing on front-end and MLP (load/store) improvements. They can still use all the tricks that Intel has used for Golden Cove, but that would require a larger core.

This table comparing the out of order engines of Zen 3 and Alder Lake is reproduced from Chips and Cheese's excellent overview of Alder Lake/Golden Cove. I've changed the order of various entries, and eliminated some columns for brevity. The most important column is the last one; most structures in the out of order engine of Golden Cove are much larger than their counterparts in Zen 3, and even Sunny Cove (Ice Lake), has larger structures than Zen 3. Despite all that, Zen 3 beats all Intel architectures before Golden Cove.

Structure	Instruction affected if it..	Golden Cove Capacity	Zen 3 Capacity	Golden Cove vs Zen 3
Reorder Buffer (ROB)	Is waiting to retire (all)	512	256	2x
Load Queue	Reads from memory	192	116	1.5x
Store Queue	Writes to memory	114	64	1.78x
Branch Order Buffer	Affects control flow	128	48 Taken 117 Not Taken	Complicated, approximately 1.33x
Integer Register File	Writes to an integer register	280 (~248+32)	192 (173 measured+32?)	1.45x
Flags Register File	Sets flags (often tied to integer registers on x86)	248	121	2.04x
Floating Point/Vector Register File	Writes to a fp/vector register	332 (300+32)	160 (139 measured+32?)	2.07x
Total Scheduler Capacity	Is waiting on an execution unit	205	160	1.28x
Fill Buffer	Misses L1D	16	24	0.66x
Superqueue	Misses L2	48	64?	0.75x?

Yeah, GC was a massive widening effort for Intel in general it seems - and they clearly needed it. IIRC Zen3 (and Zen2? Can't remember) was quite a bit wider than the Skylake-derived Intel cores in many ways - but crucially matched them in the decode stage (where ADL is wider again). But IMO AMD needs to follow suit, as they've already got core density down pat - and will increase it even further with Zen4c. From there, the only way forward is increasing IPC even for servers, as you can't push clocks higher, and increasing core counts is already getting a lot harder (IF power on EPYC Milan is ~100W!). So, barring a move to direct silicon interconnects - which are no doubt coming, but unlikely to hit 8CCD server CPUs for a while yet - they need more IPC to keep ahead (while not increasing power by much), especially in light of competition from ARM-based server chips. With Zen4 delivering 96c and Zen4c delivering 128c CPUs, Zen5 needs to focus on IPC improvements unless they want to be overtaken in everything outside of HPC and massively scalable server use cases. Though no doubt AMD's chip designers have been well aware of this for years already.

 

[Bomby569]

What mistake?

You got 8 cores crammed in a tiny little CCD with a bunch of cache that needs to be always powered on.

Once you start crossing the 5Ghz mark temps just fly upwards and it wont be better the higher or faster you go.

Both camps are pretty much at their limit to what they can do here really. Going smaller would only yield a more difficult source of heat to cool.

But really if i look at those graphs... 50 points behind on ST but a massive improvement on MT. I know my pick.

a bigger die would make for gaps for cooling like Intel does.

 

[Valantar]

a bigger die would make for gaps for cooling like Intel does.

Outside of disabled AVX units, Intel does not have meaningfully more dark silicon than AMD does. Their cores simply require more transistors and thus more space.