Given the abundance of CES 2023 hardware announcements that are expected and the conclusion of AMD’s presentation, now would be a good time to review and analyse the news from the event. So let’s go over everything that’s noteworthy and share our opinions about what we saw.
AMD 3D V-Cache CPUs
Starting with AMD’s new Ryzen 7000 3D V-Cache CPUs, which were just announced this morning. Last year, the well-liked Ryzen 7 5800X3D made its debut. V-Cache significantly increased the L3 cache capacity from 32 to 96 MB by stacking a cache die on top of the CPU chiplet. The Ryzen 7000 Zen 4 processors are now using the same kind of technology.
Prior to Zen 4, AMD had only released one V-Cache processor; now, there are three. The 7800X3D has eight cores and sixteen threads, the 7900X3D has twelve cores and twenty-four threads, and the 7950X3D has sixteen cores and 32 threads. They will all be available in February and are all 120W parts.
There are a few noteworthy aspects of these specifications. Clock speeds come first. There is no decrease in turbo speed with the 7950X3D compared to the 7950X, and both processors have the same maximum clock speed of 5.7 GHz. However, base clocks are slower, and we anticipate all core frequencies to be lower than those of the 7950X in typical applications in order to reduce the TDP from 170W to 120W. The 7900X3D, which has the same maximum frequency of 5.6 GHz as the 7900X, is the same.
The 7800X3D, on the other hand, only lists a maximum turbo clock of 5.0 GHz. This is lower than the Ryzen 7 7700X’s maximum frequency of 5.4 GHz. Although the clock speed reduction is greater with the Zen 4 8-core part compared to Zen 3, the large cache should make up for it, as we saw for the 5800X3D vs. 5800X comparison.
The total cache figures are additionally fascinating. The combined L3 and L2 cache numbers AMD lists for the 7800X3D are accurate. With Zen 4, you get 1MB per core, 32MB of on-chiplet L3 cache, 64MB of 3D V-Cache, and 8MB of L2 cache. This is 104 MB in size. However, we have 140MB and 144MB of total cache listed for the Ryzen 9 components. As a result, it appears that only one of the two CPU core chiplets has V-Cache. The 7950X already has 80MB of total cache, which is made up of 64MB of L3 and 16MB of L2. To increase to 144MB as seen here, only 64MB of additional V-Cache are required, meaning one chiplet must contain one set of V-Cache and the other chiplet must only contain one.
Given that most games refrain from using more than one chiplet on AMD’s dual-chiplet Ryzen 9 parts, that configuration is intriguing and appears to be gaming-optimized. A chiplet that won’t be used for gaming would not benefit greatly from having V-Cache added. The CPU chiplet that contains V-Cache must always be assigned to games, which means that thread scheduling for these parts must be flawless.
Other than that, these new Ryzen 9 V-Cache components seem perfect for users who want to use their computers for both gaming and productivity: a model like the 7950X should provide excellent multi-thread performance while also having top-end gaming speed enhanced by a sizable L3 cache on a single die.
AMD claims that the 7800X3D is between 21 and 30 percent faster than the 5800X3D across a four-game sample at 1080p using high image quality in their benchmarks, which you should definitely take with a huge grain of salt. That’s a sizable gain that ought to propel this component to the top of the gaming rankings.
While the Core i9-13900K has historically been the best gaming part, AMD is claiming that the 7950X3D has a performance advantage of 13 to 24 percent over it. The 5800X3D, 13900K, and Zen 4 are all quite competitive, so these new X3D parts should be about 15% faster than Zen 4 – a similar uplift from V-Cache with the previous generation. However, no direct performance comparisons between Zen 4 and Zen 4 V-Cache were shown.
Hopefully, this will give the AM5 platform, which has been suffering from low sales and interest in comparison to AM4, some life.
New affordable AMD CPUs: Non-X Ryzen 7000
The release of new, reasonably priced processors from AMD and Intel—the 65W non-X Zen 4 parts from AMD and the 13th-gen non-K lineup from Intel—was the next significant CES announcement.
On the AMD side, the Ryzen 9 7900 is essentially a 7900X with lower multi-core clocks and a max clock reduced from 5.6 to 5.4 GHz as the TDP drops from 170W to 65W. Although the 7900X can now be found for about $460, this part will cost $430, down from the 7900X’s $550 launch price.
The Ryzen 7 7700 and Ryzen 5 7600 share a similar story. As both parts decrease in TDP from 105W to 65W and boost clock speeds decrease by 100 or 200 MHz, the TDP drop is not as pronounced in this case.
The 7700 is priced at $330 and the 7600 at $230, which represent $70 savings over the X models’ launch prices. This is a significant improvement. However, unlike the 7900, the 7700X can be found for just $345, while the 7600X has occasionally dropped to $260. As a result, the discounts are less significant in comparison to current prices.
A box cooler, the Wraith Prism for the Ryzen 7 and 9 and a Wraith Stealth for the 7600 are included with the purchase of a non-X model. Box coolers were previously absent from Zen 4 CPUs, but it makes sense that the more affordable budget models would reinstate their inclusion.
New Intel affordable CPUs: Raptor Lake non-K 65W
Additionally, Intel has unveiled a sizable array of new 13th-generation Core non-K CPU models for cost-effective systems. Standard locked models with integrated graphics and F models without graphics are among them. The majority of these parts have a 65W power consumption. There is also the T series, which reduces the TDP even more to 35W, though most of these components are meant for pre-builts.
For the more expensive components in the lineup, the core configuration is the same as Intel’s unlocked Raptor Lake models. Thus, the Core i7-13700, which also applies to the Core i7-13700K, has 8 P cores and 8 E cores. The P core’s clock speed drops from 5.4 to 5.2 GHz, and the E core’s drops from 4.2 to 4.1 GHz.
All core frequencies will be much lower when running in its TDP limited state at the 65W default than the 125W unlocked models, but many motherboards will run these in a power unlimited state by default, lowering the margins and boosting performance.
All of these components are 13th generation, but Intel is mixing silicon from Raptor Lake and Alder Lake; it’s not entirely clear which CPUs use which architecture, though.
Raptor Lake is anticipated for the Core i7 and Core i9 components, with Alder Lake for Core i5s and lower processors. However, that does not imply that the configurations we are seeing are an exact replica of the 12th-gen. For instance, the popular 12400F was only a six-core model with six P cores and no E cores, whereas the newer 13400F is a 10-core processor with six P cores and four E cores. Thus, E cores are now available in the new Core i5 lineup.
The Core i3 models appear to be just faster clocked versions of 12th generation components, but they are still quad-core processors without E cores.Compared to the previous generation, these locked processors are now more expensive. For instance, the Core i5-13400F is currently $196 (RCP), whereas the Core i5-12400F cost $167 at launch and will cost $184 in 2023. The 13400F is actually listed for $210 at Newegg, pitting itself just $20 less than the Ryzen 5 7600 in what should be an interesting competition. RCP is Intel’s “recommended customer price,” which is not a true MSRP.
Similar to the Core i3-13100F, which has an RCP of $109 but costs $125 on Newegg. The Core i9-13900F costs $570, which is exactly the same price as the Ryzen 9 7950X currently available, while the Core i7-13700F costs $380, falling between the Ryzen 7 7700 and Ryzen 9 7900.
Since both AMD and Intel are releasing new models at the same time, the majority of manufacturer-provided benchmarks aren’t that important. While Intel and AMD mainly concentrated on either improvements over the previous generation or comparisons to their rivals’ last-gen parts, it is important to see how these 13th-gen parts compare to the Zen 4 parts at the same price point.
For instance, AMD pointed out that the Ryzen 5 7600 uses 65W of power while the Core i5-13600K consumes 125W. Now that there is a Core i5-13600 at 65W, this is completely irrelevant. Since they couldn’t have tested each other’s upcoming CPUs when creating these charts, benchmarking will be king for these models, and we’ll soon start to see the results.
New laptop hardware: High-end mobile CPUs
In the first quarter of 2023, all three major hardware manufacturers will release new laptop chips; these will include new AMD and Intel CPUs as well as new Nvidia and AMD GPUs. Let’s perform a brief breakdown.
The best silicon available for high-end gaming and creative laptops, starting with CPUs, is provided by both AMD and Intel. The AMD Dragon line includes up to 16 Zen 4 CPU cores, DDR5 memory, and a 5nm manufacturing process.
Known as the Ryzen 7045 HX series, these processors have an I/O die and up to two CPU core chiplets, and they seem to be built on the same silicon as desktop Ryzen 7000 processors. These are designed for laptops with 45W+ TDP or higher, or 55W+ for the top-of-the-line model.
The Ryzen 9 7945HX, which has 16 cores, 32 threads, 80MB of L3+L2 cache, and a boost clock of up to 5.4 GHz, is the top chip in this line. Essentially, a Ryzen 9 7950X with clock speeds slightly lower was power-limited to a laptop form factor. The Ryzen 9 7845HX with 12 cores, Ryzen 7 7745HX with 8 cores, and Ryzen 5 7645HX with 6 cores are three additional models. These are intended to be used in conjunction with discrete graphics. All will be offered beginning in February.
With the Alder Lake HX series, Intel intends to compete at the top end. Bringing the top specifications of the desktop platform to laptops in a more power-constrained configuration is similar to what AMD is doing. The basic configurations of Intel’s additional SKUs are similar and all have a base power rating of 55W, indicating that they are all intended for larger form factor devices.
The Core i9 range’s top model has 24 cores total, divided into 8 performance cores and 16 energy-efficient cores. The Core i9-13900K has the same configuration, but with slightly lower clocks that top out at 5.6 GHz and lower operating frequencies because of the lower TDP and maximum turbo power.
More laptop hardware: Intel and AMD mid-range offerings
The Core i7-13850HX, which has 8P and 12E cores, is followed by two other Core i7s, one of which has 8P and 8E cores, and the other of which has 6P and 8E cores, as we work our way down the stack. Like the split we saw with the new 13th-gen desktop processors, it appears that Intel is using Alder Lake dies starting with the Core i7-13700HX and going down from there.
However, the tiers have changed; the old Core i9-12950HX now functions as the 13700HX with a frequency boost, so this performance should be available at new price points. The Core i5s, which is located further down the stack, has 6P cores and either 8 or 4 E cores, depending on the model.
The H series from both brands is the next tier down from these HX series processors. The Ryzen 7040 series from AMD is a new monolithic APU that combines Zen 4 technology with RDNA 3 integrated graphics and a new AMD XDNA AI engine, all while utilising TSMC’s N4 node, a variation of their 5nm technology.
AMD has made the decision to discontinue the regular H series in favour of the HS series, which is typically designed for 14 and 15-inch laptops that are thinner and more portable. These come in three models: the Ryzen 9 7940HS, Ryzen 7 7840HS, both of which have eight cores, and the Ryzen 5 7640HS, which has six cores. They all have a TDP range of 35 to 45W. The flagship model’s clock speed is increased from 4.8 to 5.2 GHz, and the amount of cache has increased from 16 to 32 MB L3, matching the desktop lineup’s cache size. This is a good improvement for their monolithic APU option.
Additionally, the 7040 series offers up to 12 RDNA 3 compute units, which take the place of the 12 RDNA 2 compute units in the Ryzen 6000 APUs of the previous generation. Additionally, AMD spent some time discussing a new AI engine that it claims is faster and more effective than the Apple M2 AI engine. These Phoenix-codenamed HS series models will be offered beginning in March.
The H series from Intel, which includes a wide variety of processors, is in opposition to this. These are 45W models, whereas AMD is more interested in the 35W range, so we’ll probably see them in models that are more geared toward gaming and in a slightly higher tier. With some models having nearly identical specifications, the Core i7 and Core i9 models primarily use 6P core and 8E core configurations. The Core i5 range, which offers 4P cores, either 8 or 4 E cores, and lower frequencies, is followed by the Core i7-13620H, which halves the E cores as we move down the stack.
The higher end of the mobile CPU stack is heavily segmented, with a variety of options available depending on the precise kind of system an OEM wants to create.
Since the HX series’ CPU die has more cores and fewer integrated graphics than the H and HS series from either brand, it is likely to draw the most interest from people looking for high-end gaming laptops. These series sacrifice core count in favour of a beefier iGPU. The goal is to provide more specialised designs that are more appropriate for the type of system being built; for instance, a gaming laptop with discrete graphics doesn’t require the larger RDNA3 iGPU found in AMD’s Phoenix APU.
Since the majority of the information discussed isn’t pertinent to the state of laptops in 2023, we’re avoiding some of the performance comparisons made by businesses. While Intel compared their components to Ryzen 6000, AMD compared their components to Alder Lake. However, 13th-gen and Zen 4 will face off against one another and will hit the market at the same time. There are a few interesting tidbits in this article comparing hardware from different generations and generational gains in gaming performance, but with laptop benchmarking, we’ll really have to test this for ourselves to see how it works.
As we descend the stack, AMD offers the Ryzen 7035 series, which features models built for 28W thin and light notebooks. AMD is continuing to refer to this series as the U series. These components are an update to AMD’s Ryzen 6000 U-series chips, which used the Rembrandt die; the Rembrandt-R die in the 7035 series uses Zen 3+ CPU cores and up to 12 RDNA2 compute units.
The 13th-generation P series processors from Intel, which resemble the 12th-generation Alder Lake in design, will be the main rivals of the 7035 series for U-series. This may be the reason AMD is concentrating on Rembrandt-R rather than switching to Zen 4. The Core i7-1370P, which has 6P and 8E cores and shares the same architecture as the Core i7-1280P but with slightly faster clock speeds, is the top model in the 13th-generation P series of notebook processors. There are three additional CPUs that support 4P and 8E cores.
The new thin and light processors for this generation weren’t given much attention by Intel or AMD, so it appears that the majority of the exciting laptop releases will be for larger gaming and productivity systems. For a new generation of laptop processors, both companies are using a mix of new silicon and refreshed silicon across their product lines. This is likely the most updated hardware seen in a while.
We believe that both of these lineups are incredibly unclear.
We believe that both of these lineups are incredibly unclear. With different dies and the potential for cross-over between the 7035 and 7040 in the HS and U series, AMD has the 7045 HX series, 7040 HS series, 7035 U series, and 7030 U series.
The HX, H, P, and U series from Intel have a variety of configurations, power targets, and SKUs. This is great for OEMs because they can choose the exact part they need, but it’s terrible for customers because they can’t tell what chip they’re buying and what performance it will provide.
New laptop GPUs: Nvidia RTX 4000 and AMD RDNA3
The introduction of Nvidia’s entire Ada Lovelace RTX 4000 series product stack, which includes the RTX 4090, RTX 4080, RTX 4070, RTX 4060, and RTX 4050, is the big news in the laptop GPU market. These are all laptop GPUs, and despite having identical names to desktop cards (with the addition of “Laptop GPU” at the end), their performance will be inferior to that of their desktop counterparts, as we have seen repeatedly with earlier generations.
Even in comparison, the RTX 4090 Laptop GPU is not even close. It has 16GB of GDDR6 memory on a 256-bit bus and 9728 CUDA cores clocked up to 2.04 GHz in the 150W configuration. The correct desktop RTX 4090 uses 24GB of GDDR6X on a 384-bit bus and has 16384 CUDA cores clocked up to 2.52 GHz. These two components are completely unrelated even though they are both referred to as RTX 4090, which is very deceptive. Even the Ada Lovelace die is different; the RTX 4090 for desktops uses AD102, whereas the RTX 4080 for desktops uses what appears to be AD103.
As a result, the RTX 4080 Laptop GPU is pushed down a notch as well, using AD104, and has specifications that are significantly lower than those of the RTX 4080 Desktop GPU. In fact, the 4080 doesn’t even match the RTX 4070 Ti in terms of CUDA core count.
It goes without saying that the RTX 4090’s 450W TDP and high-power consumption make it impossible to fit into a laptop form factor. That much power is too much for laptops, which typically have power outputs in the 120–150W range. However, Nvidia is just calling it the RTX 4090 rather than changing the name to reflect this and calling it the RTX 4090M or something.
The majority of benchmark “numbers” compared the RTX 4000 to earlier generations, but only when newer GPUs had exclusive access to features like DLSS 3. This comparison was not fair and was therefore useless.
Because on the naming side it just seems to be getting worse with each generation, we do plan to benchmark these and compare them to their desktop models to see what the performance gap is. Yes, there was a performance gap between laptop and desktop back in the RTX 20 series era due to clock speeds and power constraints, but at least you were getting a comparable core configuration. The configurations of today’s laptops and desktops are very different.
Additionally, because Nvidia didn’t provide many accurate performance estimates, it is impossible to predict what these new GPUs will bring to laptop form factors. The majority of benchmark “numbers” compared the RTX 4000 to earlier generations, but only when newer GPUs had exclusive access to features like DLSS 3. This comparison was not fair and was therefore useless. Two games were included that didn’t have this stuff enabled, which results in an improvement in performance of about 50%, which is quite impressive given that laptops can’t use more power than they already do.
Along with improved performance, laptops will soon support DLSS 3 and AV1 video encoding, among other Ada Lovelace benefits. Additionally, Nvidia promoted new Max-Q technologies like lower voltage GDDR6 and on-chip memory that are intended to increase efficiency for these laptop models.
What is known is that on February 8th, laptops costing more than $2,000 will begin to offer the RTX 4080 and RTX 4090 graphics cards.
Starting on February 22, the other RTX 4070 through RTX 4050 models will be offered for $1,000 to $2,000 each.
The AMD side of mobile GPU news isn’t all that exciting because they’re only releasing a few mid-range RDNA3 SKUs, the Radeon RX 7600M XT and RX 7600M along with the S variants.
The 7600M XT strangely becomes the 7700S.
The only real information we have for these Navi 33 designs is 8GB of GDDR6 and a 128-bit memory bus, along with a few unhelpful FP32 numbers.
AMD anticipates that the 7600M XT will perform 20 to 40% better than the RX 6600M, which would put it above Nvidia’s desktop RTX 3060 in terms of power. Beginning in February, these 7600 series GPUs will be accessible.
News From Source link