Today we're taking our first look at AMD's new Zen 5 processors, starting with the Ryzen 7 9700X, their latest 8-core / 16-thread CPU designed to replace the 7700X.
Today we're taking our first look at AMD's new Zen 5 processors, starting with the Ryzen 7 9700X, their latest 8-core / 16-thread CPU designed to replace the 7700X.
One very easy explanation: Zen5 has increased L1 data cache to 48kB. Since Intel CPUs used 32kB L1 data cache from Core 2 (2006) to Sunny Cove (2019), many games were for obvious reasons optimized for 32kB L1 data cache. Tbh Intel's Cove series CPUs are still much worse in games than they "should be".Wow... That doesn't make any sense. Perhaps today's games simply aren't optimized for Zen5, but in any case this performance doesn't match what you'd expect from a design with so many changes.
Looking forward to seeing Chips and Cheese come up with some explanation for this, obviously there's some bottleneck in the architecture, perhaps the insistence on not increasing the L3 cache (to justify the X3D line?).
Zen5 is made for modern software. Most software tested are old ones.Yo what… idk something must be wrong here or AMD lied hard. Wow. Turning into Intel I guess lol
Interesting - I'm seeing reviews all across the board. HWCooling, for eg, shows the same results you do though they go more in-depth with testing non-gaming applications and show that there's something like a 20/30% increase in efficiency (ie slightly higher performance + slightly lower wattage). That leads overhead for PBO to ramp up in applications though this doesn't seem to translate to that much improvement in gaming performance.
Meanwhile, sites like TechPowerUp are showing massive increases in performance with the 9600X on PBO at/slightly under the 7900X with games at 4K when paired with a 4090. This makes me wonder if the variance between individual chips themselves is extremely wide, if there's something about the motherboard (BIOS/VRM/?) that may be limiting the processor, or something else.
Also for folks who saw this review and immediately want to jump on Intel - LOL, look up 'Intel degradation' and tell me if a slightly underwhelming processor model is the worst thing you can end up with.
Averaged over these 49 tests, the Ryzen 7 9700X achieves a 6% performance uplift over the Ryzen 7 7700X, which is less than expected. One of the reasons is that heavy multithreaded loads push the CPU into its power limit, so it will throttle. When we removed the power limit and enabled PBO auto overclocking we gained an additional 6% on average, up to 15% in single tests like rendering. A further 6% gain from enabling PBO is much more than what we typically see on modern CPUs, which are usually configured for peak performance out of the box. While some may argue that the TDP limit is set too low and limits performance, I view it as a strategy to enhance thermals and energy efficiency without significantly compromising performance.
Quick guess:How are you measuring power draw? Compared to your previous reviews, it seems like you've added ~80w on the 12700k MT power draw, 60 on the 12900k, 40-50 on the 14600k. How?
The chips were already running unlimited power in the previous review.Quick guess:
New BIOS or other motherboard?! Maybe "Intel Extreme Profile"