AMD Ryzen 7 9800X3D review: devastating gaming performance

On November 7, 2024By Littum

AMD’s $480 Ryzen 7 9800X3D comes armed with eight cores and 16 threads paired with a new version of the company’s game-enhancing 3D V-Cache technology that delivers impressive performance and convincingly takes the throne as the fastest gaming CPU on the market. AMD claims the chip is 20% better than Intel’s latest flagship and 8% faster than the previous generation Ryzen 7 7800X3D. However, our benchmarks show it beating Intel’s current-gen flagship Core 9 285K by an almost unrealistic 35% on average in our test suite. It even beats Intel’s fastest competing gaming chip, the Core i9-14900K, by an equally incredible 30% on average in our test suite. Of course, the winnings will vary by title due to the vagaries of the 3D V-Cache technology. Still, we recorded incredible performance that firmly cements the Ryzen 9 9800X3D as the best CPU for gaming.

AMD achieved this feat by leveraging its proven 3D V-Cache technology, which uses a powerful, vertically stacked 64 MB L3 cache chip to boost gaming performance. To further improve performance, AMD moved the cache chip from the top of the processor to the bottom. This provides significantly greater thermal headroom, enabling higher clock speeds, while the additional cache unlocks the Zen 5 architecture’s highest gaming performance. As we’re used to with game-optimized X3D processors, not all games will benefit from the technology, and a few are even slightly slower (1 to 2%) than the previous generation model. Still, the chip also significantly improves 1% low frame rates, even in these titles. The higher 1% lows will also help when you’re GPU limited, like when you’re gaming at higher resolutions.

AMD has now also enabled full overclocking for the 9800X3D we tested, allowing you to squeeze out more performance in both gaming and productivity work. Speaking of which, the new design also helps remove some of the heavy penalties in productivity work that you’ll often pay for using a gaming-optimized X3D processor, allowing the 9800X3D to match and sometimes even exceed a comparable octa-core Ryzen 9000 CPU in multiple productivity workloads.

Swipe to scroll horizontally

AMD Zen 5 Ryzen 9000 ‘Granite Ridge’ specifications and prices
	Street/MSRP	Arch	Cores / Threads	Base / Boost Clock (GHz)	Cache (L2/L3)	TDP / PPT	Memory
Ryzen 9 9950X	$649	Zen 5	16/32	4.3 / 5.7	80 MB (16+64)	170W / 230W	DDR5-5600
Ryzen 9 9900X	$499	Zen 5	12/24	4.4 / 5.6	76 MB (12+64)	120W / 162W	DDR5-5600
Ryzen 7 9800X3D	$480	Zen 5 X3D	8/16	4.7 / 5.2	104 MB (8+96)	120W / 162W	DDR5-5600
Ryzen 7 9700X	$359	Zen 5	8/16	3.8 / 5.5	40 MB (8+32)	65W / 88W / 105W	DDR5-5600
Ryzen 5 9600X	$279	Zen 5	6/12	3.9 / 5.4	38 MB (6+32)	65W / 88W / 105W	DDR5-5600

The $480 Ryzen 9 9800X3D is AMD’s lone chip launch today, and the chip will be available on shelves tomorrow. The 9800X3D saves the day for gaming enthusiasts facing not just one, but two disappointing chip launches this year: AMD’s Zen 5-powered Ryzen 9000 series debuted with weak gains in gaming performance when it arrived earlier this year, but as you’ll see in our fresh tests we did for this review, newer firmware and a Windows tweak have wringed out more performance. Intel also released its long-awaited Arrow Lake Core Ultra series just two weeks ago, but these chips failed to match the gaming performance of Intel’s previous-gen chips. Unfortunately, there is no obvious solution in sight, setting the stage for a disappointing year ahead for gaming chips.

The Ryzen 7 9800X3D changes that completely, bringing a top-tier gaming chip to the market at a reasonable price considering its capabilities. The chip will slot into existing AM5 motherboards and won’t require the expensive complementary components, such as the expensive high-speed RAM kits, heavy-duty coolers and motherboards needed to extract the maximum performance from other top-tier chips. Meanwhile, Intel seems completely unable to counter AMD’s 3D V-Cache technology even after two and a half long years and now three chip generations after it debuted the now legendary Ryzen 7 5800X3D. Let’s look at the technology, then move on to the surprisingly good gaming benchmarks.

AMD Ryzen 7 9800X3D Prices and specifications

The Ryzen 7 9800X3D has the Zen 5 microarchitecture, which you can read more about here. The processor is compatible with AM5 motherboards, and AMD’s partners finally have X870E models available on the market.

The eight-core, 16-thread Ryzen 7 9800X3D debuts at $480, a $30 premium over the launch price of the Ryzen 7 7800X3D. The 9800X3D has a 120W TDP and has 500 MHz higher base clocks and 200 MHz higher boosts than the previous gen Zen 4 powered Ryzen 7 7800X3D.

The 9800X3D’s 5.2 GHz boost is impressive, but it lags the directly comparable Zen 5, the eight-core Ryzen 7 9700X, by 300 MHz. However, AMD has compensated with a 900 MHz higher base clock, allowing the 9800X3D to outperform the 9700X in some threaded workloads, as you’ll see in our benchmarks on the following pages.

As with the previous generations, the Ryzen 7 9800X3D does not come with a cooler. AMD recommends at least a 240-280mm liquid cooler (or equivalent). The processor has the same maximum temperature of 95 C (TjMax), but the new L3 cache chiplet design allows the chip to operate at higher clock speeds for longer (improved boost residency), which equates to stronger performance gains within the same TDP envelope.

Ryzen 7 9800X3D is fully overclockable – you can adjust the CPU cores, fabrics and memory to your liking. But while multiplier-based overclocking is available, as we’ve seen with other Zen 5 processors, most users with conventional cooling will be best served by the automatic overclocking Precision Boost Overdrive (PBO) feature. We have extensive testing with this feature enabled on the following pages.

AMD also bumped memory support up to DDR5-5600 from the DDR5-5200 found with the previous generation 7800X3D.

AMD Ryzen 7 9800X3D architecture

Picture 1 of 7

In the illustrations above, you can see AMD’s original approach with its 3D V-Cache technology. With previous designs utilizing the Zen 3 and Zen 4 architectures, AMD stacked an additional L3 SRAM chiplet directly in the center of the compute die (CCD) chiplet to isolate it from the heat-generating cores. We’ve covered the details of the first generation of this technology here.

AMD used hybrid bonding technology to connect the cache chiplet to the underlying computing die. This increased the cache capacity to 96MB to accelerate performance in latency-sensitive applications such as games.

AMD placed the chip and several pieces of structural silicon on top of the compute die, but as we demonstrated with thermal throttling tests of the Ryzen 9 7950X3D processor, the cache chip and structural silicon shims acted as a thermal blanket that inhibited effective thermal transfer, essentially trapping waste heat. As a result, AMD limited the effective voltage and frequency range of the 3D V-Cache equipped die to curb heat generation. Simply put, the chip got hot and thus had to run slower.

Picture 1 of 3

Ryzen 7 9800X3D — (Image credit: Tom’s Hardware)

AMD has now placed the cache chiplet under the die to alleviate the thermal challenges, placing the compute die closer to the thermal interface material, the IHS, and ultimately the CPU cooler. This improves the amount of computing performance AMD can extract from the 120W TDP envelope, and it also improves boost residency/duration. However, it comes with a new set of challenges – this design requires AMD to route the power and signal TSVs for the entire chip through the underlying L3 cache chiplet.

The new L3 cache chip is based on the same 7nm SRAM-optimized process node that AMD used in the previous two generations of 3D V-Cache technology in a technique that AMD now refers to as its ‘Second-Gen 3D V- Cache’ technology (this is an odd branding choice – this is actually the third improvement to the technology).

As before, AMD thins both the L3 cache chip and the 4nm (N4P) compute die to meet standard Z-height requirements for the processor package. AMD uses the same hybrid copper-to-copper bond with a 9-micron bump pitch to connect the two dies.

The chip now spans the entire bottom of the computer die, eliminating the need for the structural silicon shims. AMD doesn’t share the size of the L3 cache chip or the transistor density, but the compute die measures 70.6mm^2, so it’s a safe bet that’s the size of the cache chip as well. That’s much larger than the 36mm^2 L3 cache chip used for the Zen 4 X3D models – this is despite using the same processor node and the same 64MB L3 capacity. AMD probably has a lot of ’empty’ silicon on the cache chip that is simply there for structural support without the need to add separate shims.

Picture 1 of 2

AMD Ryzen 7 9800X3D — (Image credit: Tom’s Hardware)

Swipe to scroll horizontally

3D V-Cache technology
Row 0 – Cell 0	“New” 2nd-Gen 7nm 3D V-Cache Die	2nd-Gen 7nm 3D V-Cache Die	First-Gen 7nm 3D V-Cache Die	4nm Zen 5 Core Complex Die (CCD)	5nm Zen 4 Core Complex Die (CCD)	7nm Zen 3 Core Complex Die (CCD)
Size	36 mm^2	36 mm^2	41 mm^2	70.6 mm^2	66.3 mm^2	80.7 mm^2
Transistor count	?	~4.7 billion	4.7 billion	8.6 billion	6.57 billion	4.15 billion
MTr/mm^2 (transistor density)	?	~130.6 million	~114.6 million	121.81 MTr/mm^2	~99 million	~51.4 million

Of course, AMD has to feed more power and signal TSVs through the L3 cache chip, which now acts as a kind of interposer, up to the computer. AMD says that power TSVs are distributed throughout the SRAM array to feed the compute array located above. Extra power TSVs are distributed in areas unused for other functionality (presumably as more cache storage), which should at least partially account for the larger size.

Swipe to scroll horizontally

AIDA L3 Cache Latency Measurements — Ryzen 7 9800X3D
Memory Latency – Tom’s hardware	DDR5	CUDIMM DDR5	L3 latency
Ryzen 7 9800X3D	78.5 ns (DDR5-6000)	reach	13.2 ns
Ryzen 7 7800X3D	73.4 ns (DDR5-6000)	reach	12.8 ns
Ryzen 7 9700X	69.4 ns (DDR5-6000)	reach	11.3 ns
Core Ultra 9 285K	94.1 ns (DDR5-5600)	91.9 ns	16.6 / 15.8 ns
Core i9-14900K	79.1 ns (DDR5-5600)	REACH	21.8 ns

AMD says accessing the L3 cache chip carries the same four-clock penalty as with the previous generation. We measured a sub-ns increase in L3 latency compared to the previous 3D V-Cache equipped Ryzen 7 7800X3D, which falls within the expected variance. AMD says the bandwidth between the computer and the L3 chip is the same as the previous generation at 2.5 TB/s, but the bandwidth varies with clock speeds.

Access to both main memory and I/O must now travel through the L3 cache chip. As you can see in the measurements above, we see a significant increase in DDR5 latency compared to the standard Ryzen 7 9700X and Ryzen 7 7800X3D. However, AMD says there is no increase in I/O latency due to traveling through the cache chip. We will follow up for more details.

Obviously, any increase in memory latency is undesirable. However, we expect this to have minimal impact on overall performance due to the larger slab of L3 cache minimizing access to main memory in some scenarios. Let’s see how the performance looks in game on the next page.