Sunday, July 31st 2022
Intel Core i5-13600K and Core i7-13700K QS CPUs Benchmarked
Is there anything better than yet another benchmark leak of upcoming products? This time around we don't have to make do with Geekbench or some other useless benchmark, as a bilibili user in the PRC has posted a video where he has put the upcoming Intel Core i5-13600K and Core i7-13700K CPUs through 10 different games, plus 3DMark Fire Strike and Time Spy. This has been done at 1080p, 1440p and 2160p at that, using a GeForce RTX 3090 Ti graphics card. Both CPUs are QS or Qualification Samples, which means they're going to be close to identical to retail chips, unless there are some last minute issues that are discovered. The CPUs were tested using an ASRock Z690 Steel Legends WiFi 6E motherboard, well, two actually, as both a DDR4 and a DDR5 version were used. The DDR4 RAM was running at 3600 MHz with slow-ish timings of 18-22-22 in gear 1, whereas the DDR5 memory was running at 5200 MHz, most likely at 40-40-40 timings, although the modules were rated for 6400 MHz, in both cases we're looking at 32 GB.
Courtesy of @harukaze5719, we have some much easier to read graphs than those provided by the person that tested the two CPUs, but we've included the full graphs below as well. Each CPU was compared to its current SKU equivalent from Intel and in many of the games tested, the gain was a mere percent or less to three or four percent. However, in some games—at specific resolutions—especially when paired with DDR5 memory, the performance gain was as much as 15-20 percent. A few of the games tested, such as FarCry 6 at 4K, the game ends up being GPU limited, so a faster CPU doesn't help here as you'll see in the graphs below. There are some odd results as well, where the DDR5 equipped systems saw a regression in performance, so it's hard to draw any final conclusions from this test. That said, both CPUs should offer a decent performance gain, as long as the game in question isn't GPU limited, of around five percent at 1440p when paired with DDR5 memory.
Sources:
bilibili video, bilibili graphs, @harukaze5719 graphs
Courtesy of @harukaze5719, we have some much easier to read graphs than those provided by the person that tested the two CPUs, but we've included the full graphs below as well. Each CPU was compared to its current SKU equivalent from Intel and in many of the games tested, the gain was a mere percent or less to three or four percent. However, in some games—at specific resolutions—especially when paired with DDR5 memory, the performance gain was as much as 15-20 percent. A few of the games tested, such as FarCry 6 at 4K, the game ends up being GPU limited, so a faster CPU doesn't help here as you'll see in the graphs below. There are some odd results as well, where the DDR5 equipped systems saw a regression in performance, so it's hard to draw any final conclusions from this test. That said, both CPUs should offer a decent performance gain, as long as the game in question isn't GPU limited, of around five percent at 1440p when paired with DDR5 memory.
84 Comments on Intel Core i5-13600K and Core i7-13700K QS CPUs Benchmarked
Suppose that a processor uses a constant 100 watts to complete a workload in one minute, but another one uses 75 watts to complete a workload in one minute and 20 seconds. The 100W processor is then more efficient than the 75W one.
A faster core that chugs power but completes a task quicker is more desirable over one that is leaner on power but takes a longer span of time to complete the same task in terms of total power consumption.
What probably will happen is both will have their strengths and weaknesses, but given Intel's aggressive clocks AMD will be more efficient.
I still remember they adjusted the score composition to beat Ryzen, but instead they made the i3 beats everything including Intel's own 7980xe.
This time it should be some trickery about the AVX512
They do not expect Zen4 with AVX512 support would 'break' their biased benchmark. In this case, power consumption of the the 13600k and 13700k were also leaked before
13600k ~ 12700k
13700k ~ 12900k
Almost no improvement in efficiency
Just more watts more performance.
Modern CPUs can max out their frequency in 1 ms, so bursting above the TDP for a small amount of time in order to rapidly complete a task actually *saves* power. Of course the usefulness of such feature on desktops is limited, and only serve as TDP rebalancing to enable higher frequencies in situations that do not leverage all processor cores.
DDR5 is basically doubling it again, so you need twice as fast on the memory to match up. (Over simplified, but close enough)
This is why DDR4 3600 is more or less winning in most of the performance charts for both platforms, as they can all run it 1:1 with enough room to keep latencies down - I'm not sure what the ideal value is for DDR5 yet, if people have figured that out
To me it is more like an 'Excuse made to cover a problem' .
Everyones selling current stock at discounts (look at the 5900x and how low its gone), which implies the new products might be cheap enough they'd have trouble selling the old stock at current prices
There is no way in hell amd will be more efficient. Only the 7950x stands a chance at that, the rest of the lineup won't do much when testing at same wattage
But overall, that really look like the Athlon X2 / Pentium 4 Northwood era all that. But this time, instead of a GHz rush by intel, it's a ultra fat core that is power hungry. We all know they ended up in a wall, but i suspect they are just trying to buy time while they get on a new leaner architecture like they did with the Core 2 architecture. And also, we now all know that the E-core are not for low power usage or background task but more for improving the multicore performance without using too much silicon space. Not a bad strategy to buy time while they work on chiplets.
On AMD, i think they still have huge place to improve their cores, The fact that a R7 5800 get so much more performance with 3D-Vcache probably mean the core are starving. I do not think they need that much rework on the execution side. They seems to need better prefetching, more/faster L2 cache and better branch prediction. All that without using too much power.
I think they will be competitive on single thread but the multiplication of e-core will give them trouble on productivity task that benefits from many cores. They should probably gave them option to have 3 chiplets, or 1x8 Zen4 chiplets + 1 x16Zen4c chiplets or something like that.
But we will see, i still think AMD can surprise with Zen4 but the fight will be hard.
In the end, the one that benefits the most is the end user. Intel is trying to push really hard the cadence after having milking the market for few generation and that prevent AMD of milking the market right now.
That will mean competitive pricing, that will mean good reason to upgrade, that will mean the baseline configuration for new games will have more and more resources. Hope they use it well.
With higher frequencies, new architecture, I'm pretty sure they will be both close with AMD more efficient.
So you can't even say let's just test at x watt.