https://www.geeks.fyi/ Mon, 22 Jun 2026 08:38:18 +0000 MyBB https://www.geeks.fyi/showthread.php?tid=22070 Tue, 16 Jun 2026 06:01:35 +0000 harlan4096]]> https://www.geeks.fyi/showthread.php?tid=22070 Quote:AMD quietly launches new Zen2/Zen+ APUs: Ryzen 7 4700LE arrives as OEM-only AM4 desktop CPU with no iGPU

Older Renoir and Picasso designs return in 2026.



AMD has quietly added three processors to its official product pages. The list includes Ryzen 7 4700LE for desktops and two Ryzen 3000U mobile parts, Ryzen 5 3501U and Ryzen 3 3100U.

4700LE: Zen2 CPURyzen 7 4700LE is listed as an OEM-only desktop processor. Ryzen 7 4700LE is based on Renoir and uses Zen 2 CPU cores. AMD gives it 8 cores and 16 threads, a 3.6 GHz base clock and up to 4.2 GHz boost. The CPU has a 65W default TDP and uses the AM4 socket, but AMD lists no integrated graphics. A discrete graphics card is required.

3000U: Zen+ CPUsThe two Ryzen 3000U parts are listed under laptop processors, although AMD’s page shows “laptops, desktops” as the form factor. Ryzen 5 3501U has 4 cores and 8 threads, a 2.1 GHz base clock and up to 3.7 GHz boost. Ryzen 3 3100U is a 2-core and 2-thread model with a 1.9 GHz base clock and up to 3.2 GHz boost. That’s a Zen+ CPUs, which were first introduced in 2019. 

Continue Reading... ]]> Quote:AMD quietly launches new Zen2/Zen+ APUs: Ryzen 7 4700LE arrives as OEM-only AM4 desktop CPU with no iGPU

Older Renoir and Picasso designs return in 2026.



AMD has quietly added three processors to its official product pages. The list includes Ryzen 7 4700LE for desktops and two Ryzen 3000U mobile parts, Ryzen 5 3501U and Ryzen 3 3100U.

4700LE: Zen2 CPURyzen 7 4700LE is listed as an OEM-only desktop processor. Ryzen 7 4700LE is based on Renoir and uses Zen 2 CPU cores. AMD gives it 8 cores and 16 threads, a 3.6 GHz base clock and up to 4.2 GHz boost. The CPU has a 65W default TDP and uses the AM4 socket, but AMD lists no integrated graphics. A discrete graphics card is required.

3000U: Zen+ CPUsThe two Ryzen 3000U parts are listed under laptop processors, although AMD’s page shows “laptops, desktops” as the form factor. Ryzen 5 3501U has 4 cores and 8 threads, a 2.1 GHz base clock and up to 3.7 GHz boost. Ryzen 3 3100U is a 2-core and 2-thread model with a 1.9 GHz base clock and up to 3.2 GHz boost. That’s a Zen+ CPUs, which were first introduced in 2019. 

Continue Reading... ]]> https://www.geeks.fyi/showthread.php?tid=22022 Tue, 02 Jun 2026 06:18:47 +0000 harlan4096]]> https://www.geeks.fyi/showthread.php?tid=22022 Quote:Xe3P also listed for “Next Gen PC”



The GPU architecture roadmap shown at Computex 2026 also lists Xe3P under “Next Gen PC.” This is separate from Panther Lake, which already uses Xe3 graphics in the Core Ultra 300 series.

What Xe3p is

Intel confirmed Xe3p powers Crescent Island data-center GPU, which was today confirmed to support up to 480GB memory. This is not the game card though. Meanwhile, the new slide mentions PC.

What the wording does not confirm are discrete Xe3P graphics cards for gamers. It more likely refers to future client processors, perhaps Nova Lake variants with Xe3P-based graphics. Intel has confirmed Xe3P for Crescent Island, but it has not announced a gaming graphics card based on this architecture.

Continue Reading... ]]> Quote:Xe3P also listed for “Next Gen PC”



The GPU architecture roadmap shown at Computex 2026 also lists Xe3P under “Next Gen PC.” This is separate from Panther Lake, which already uses Xe3 graphics in the Core Ultra 300 series.

What Xe3p is

Intel confirmed Xe3p powers Crescent Island data-center GPU, which was today confirmed to support up to 480GB memory. This is not the game card though. Meanwhile, the new slide mentions PC.

What the wording does not confirm are discrete Xe3P graphics cards for gamers. It more likely refers to future client processors, perhaps Nova Lake variants with Xe3P-based graphics. Intel has confirmed Xe3P for Crescent Island, but it has not announced a gaming graphics card based on this architecture.

Continue Reading... ]]> https://www.geeks.fyi/showthread.php?tid=22021 Tue, 02 Jun 2026 06:17:40 +0000 harlan4096]]> https://www.geeks.fyi/showthread.php?tid=22021 Quote:NVIDIA confirms RTX Spark PCs with Blackwell GPU, 20-core Grace CPU and DLSS support

NVIDIA is turning RTX Spark into a Windows PC platform



NVIDIA has officially announced RTX Spark, a new Windows PC platform built around a Grace Blackwell superchip. The announcement confirms that NVIDIA is entering the consumer PC processor segment with a full system-on-chip design, not just a discrete GPU.

The chip was announced with Microsoft at GTC Taipei. NVIDIA says RTX Spark is designed for AI agents, content creation and gaming. The platform will appear in slim Windows laptops and compact desktop PCs this fall.

RTX Spark uses a Blackwell RTX GPU with 6144 CUDA cores and fifth-generation Tensor Cores with FP4 support. NVIDIA lists up to 1 petaFLOP of FP4 AI performance. The GPU is connected through NVLink-C2C to a 20-core Grace CPU, which NVIDIA says was custom-built with MediaTek.

Continue Reading... ]]> Quote:NVIDIA confirms RTX Spark PCs with Blackwell GPU, 20-core Grace CPU and DLSS support

NVIDIA is turning RTX Spark into a Windows PC platform



NVIDIA has officially announced RTX Spark, a new Windows PC platform built around a Grace Blackwell superchip. The announcement confirms that NVIDIA is entering the consumer PC processor segment with a full system-on-chip design, not just a discrete GPU.

The chip was announced with Microsoft at GTC Taipei. NVIDIA says RTX Spark is designed for AI agents, content creation and gaming. The platform will appear in slim Windows laptops and compact desktop PCs this fall.

RTX Spark uses a Blackwell RTX GPU with 6144 CUDA cores and fifth-generation Tensor Cores with FP4 support. NVIDIA lists up to 1 petaFLOP of FP4 AI performance. The GPU is connected through NVLink-C2C to a 20-core Grace CPU, which NVIDIA says was custom-built with MediaTek.

Continue Reading... ]]> https://www.geeks.fyi/showthread.php?tid=22018 Mon, 01 Jun 2026 08:00:59 +0000 harlan4096]]> https://www.geeks.fyi/showthread.php?tid=22018 AMD extends AM5 socket support through 2029 with future Ryzen CPUs]]> AMD extends AM5 socket support through 2029 with future Ryzen CPUs]]> https://www.geeks.fyi/showthread.php?tid=22017 Mon, 01 Jun 2026 08:00:00 +0000 harlan4096]]> https://www.geeks.fyi/showthread.php?tid=22017 Quote:Intel confirms 160GB Crescent Island design, partners may go up to 480GB



Intel has shared more details on Crescent Island at Computex 2026. The GPU is based on the Xe3P architecture and is aimed at AI inference workloads rather than gaming.

Intel previously confirmed Crescent Island with 160GB of LPDDR5X memory. According to the latest report from Computex, this is the reference configuration. Board partners may be able to build cards with up to 480GB of LPDDR5X memory.

Continue Reading... ]]> Quote:Intel confirms 160GB Crescent Island design, partners may go up to 480GB



Intel has shared more details on Crescent Island at Computex 2026. The GPU is based on the Xe3P architecture and is aimed at AI inference workloads rather than gaming.

Intel previously confirmed Crescent Island with 160GB of LPDDR5X memory. According to the latest report from Computex, this is the reference configuration. Board partners may be able to build cards with up to 480GB of LPDDR5X memory.

Continue Reading... ]]> https://www.geeks.fyi/showthread.php?tid=22016 Mon, 01 Jun 2026 07:58:37 +0000 harlan4096]]> https://www.geeks.fyi/showthread.php?tid=22016 Quote:Intel Xeon 6+ brings 288 E-cores to the existing Xeon 6900 platform

First Intel 18A data center CPU launches June 1.



Intel has launched its Xeon 6+ processor series, also known as Clearwater Forest. This is Intel’s first data center CPU built with compute tiles on Intel 18A, and the new family focuses on high core density rather than P-core performance.

The top SKU is the Xeon 6990E+, a 288-core processor based on next-gen Darkmont E-cores. Intel lists two power versions of this model, one at 450W and another at 330W. Both keep the same 288-core count, 576MB of L3 cache, 12-channel DDR5-8000 support and 96 PCIe Gen5 lanes.

Continue Reading... ]]> Quote:Intel Xeon 6+ brings 288 E-cores to the existing Xeon 6900 platform

First Intel 18A data center CPU launches June 1.



Intel has launched its Xeon 6+ processor series, also known as Clearwater Forest. This is Intel’s first data center CPU built with compute tiles on Intel 18A, and the new family focuses on high core density rather than P-core performance.

The top SKU is the Xeon 6990E+, a 288-core processor based on next-gen Darkmont E-cores. Intel lists two power versions of this model, one at 450W and another at 330W. Both keep the same 288-core count, 576MB of L3 cache, 12-channel DDR5-8000 support and 96 PCIe Gen5 lanes.

Continue Reading... ]]> https://www.geeks.fyi/showthread.php?tid=22014 Sun, 31 May 2026 10:19:20 +0000 harlan4096]]> https://www.geeks.fyi/showthread.php?tid=22014 Quote:AMD Radeon RX 9070 GRE gets global launch with 12GB memory



VideoCardz has learned that AMD is launching the Radeon RX 9070 GRE globally on June 1. The card will carry a $549 starting price, which puts it directly into the same MSRP slot as the Radeon RX 9070 non-XT at launch.

The Radeon RX 9070 GRE is not a new GPU. AMD introduced this model last year as a China-focused Radeon 9000 series card. It uses a cut-down Navi 48 GPU with 48 Compute Units and 3,072 Stream Processors, paired with 12GB of GDDR6 memory on a 192-bit bus. AMD lists memory bandwidth at up to 432 GB/s and board power at 220W.

Same price as the RX 9070 at launchThe $549 price is the most interesting part of the global launch. This is the same MSRP AMD announced for the Radeon RX 9070, which has 16GB of memory and a wider 256-bit bus. Interestingly, AMD’s own comparison slide now lists the RX 9070 as “starting at $619”, so it is not clear whether the non-XT model is being moved up, reduced in availability, or simply compared using current retail pricing rather than launch MSRP. The PCPartPicker confirms that RX 9070 is now sold for $599+.

Continue Reading... ]]> Quote:AMD Radeon RX 9070 GRE gets global launch with 12GB memory



VideoCardz has learned that AMD is launching the Radeon RX 9070 GRE globally on June 1. The card will carry a $549 starting price, which puts it directly into the same MSRP slot as the Radeon RX 9070 non-XT at launch.

The Radeon RX 9070 GRE is not a new GPU. AMD introduced this model last year as a China-focused Radeon 9000 series card. It uses a cut-down Navi 48 GPU with 48 Compute Units and 3,072 Stream Processors, paired with 12GB of GDDR6 memory on a 192-bit bus. AMD lists memory bandwidth at up to 432 GB/s and board power at 220W.

Same price as the RX 9070 at launchThe $549 price is the most interesting part of the global launch. This is the same MSRP AMD announced for the Radeon RX 9070, which has 16GB of memory and a wider 256-bit bus. Interestingly, AMD’s own comparison slide now lists the RX 9070 as “starting at $619”, so it is not clear whether the non-XT model is being moved up, reduced in availability, or simply compared using current retail pricing rather than launch MSRP. The PCPartPicker confirms that RX 9070 is now sold for $599+.

Continue Reading... ]]> https://www.geeks.fyi/showthread.php?tid=21987 Thu, 21 May 2026 10:38:08 +0000 harlan4096]]> https://www.geeks.fyi/showthread.php?tid=21987 Quote:AMD Ryzen AI Max+ PRO 495 confirmed

AMD confirms Ryzen AI Max 400 “Gorgon Halo” supports up to 192GB memory and 160GB VRAM



AMD has confirmed that its Ryzen AI Max 400 series, codenamed “Gorgon Halo,” will support up to 192GB of unified memory. The same platform will also allow up to 160GB to be allocated as VRAM, according to AMD’s own presentation slide.

The confirmation came alongside AMD’s Ryzen AI Halo workstation presentation. That system uses the current Ryzen AI Max+ 395, also known as Strix Halo, with 128GB of LPDDR5x-8000 memory. We will cover the Ryzen AI Halo system itself in a separate article.

Strix and Gorgon Halo differencesThe new Ryzen AI Max 400 slide lists up to 16 Zen 5 CPU cores and 32 threads, up to 5.2 GHz CPU boost clock, up to 40 RDNA 3.5 GPU Compute Units and up to 3.0 GHz GPU boost clock. AMD also lists up to 55 TOPS from the XDNA 2 NPU, compared to 50 TOPS on current Ryzen AI Max+ 395 systems.

AMD will also upgrade graphics (as suggested by the SKU names), with clocks boosted up to 100 MHz it seems. This is known from previous leak, mind you.

AMD has not confirmed to The Register when exactly the MAX+ 395 variant of the Ryzen AI Halo box will be available. 

Continue Reading... ]]> Quote:AMD Ryzen AI Max+ PRO 495 confirmed

AMD confirms Ryzen AI Max 400 “Gorgon Halo” supports up to 192GB memory and 160GB VRAM



AMD has confirmed that its Ryzen AI Max 400 series, codenamed “Gorgon Halo,” will support up to 192GB of unified memory. The same platform will also allow up to 160GB to be allocated as VRAM, according to AMD’s own presentation slide.

The confirmation came alongside AMD’s Ryzen AI Halo workstation presentation. That system uses the current Ryzen AI Max+ 395, also known as Strix Halo, with 128GB of LPDDR5x-8000 memory. We will cover the Ryzen AI Halo system itself in a separate article.

Strix and Gorgon Halo differencesThe new Ryzen AI Max 400 slide lists up to 16 Zen 5 CPU cores and 32 threads, up to 5.2 GHz CPU boost clock, up to 40 RDNA 3.5 GPU Compute Units and up to 3.0 GHz GPU boost clock. AMD also lists up to 55 TOPS from the XDNA 2 NPU, compared to 50 TOPS on current Ryzen AI Max+ 395 systems.

AMD will also upgrade graphics (as suggested by the SKU names), with clocks boosted up to 100 MHz it seems. This is known from previous leak, mind you.

AMD has not confirmed to The Register when exactly the MAX+ 395 variant of the Ryzen AI Halo box will be available. 

Continue Reading... ]]> https://www.geeks.fyi/showthread.php?tid=21986 Thu, 21 May 2026 10:36:13 +0000 harlan4096]]> https://www.geeks.fyi/showthread.php?tid=21986 Quote:AMD EPYC 8005 server CPUs for telco



AMD has now released its EPYC 8005 “Sorano” server CPU series, following the earlier announcement around MWC 2026. We already covered Sorano in February, but AMD has now shared more details.

EPYC 8005 is not using Zen 5c, despite being the successor to EPYC 8004 “Siena,” which used Zen 4c. AMD moved this lineup to full Zen 5 cores. This also increases L3 cache and raises clock speeds, with current SKUs reaching up to 4.5 GHz. 

Full Zen 5 instead of Zen 5cThe top model is the EPYC 8635P with 84 cores, 168 threads, 384MB of L3 cache and a 225W default TDP. AMD’s own blog says the new series scales from 8 to 84 Zen 5 cores and targets single-socket systems for edge, telco and cloud storage deployments.

Continue Reading... ]]> Quote:AMD EPYC 8005 server CPUs for telco



AMD has now released its EPYC 8005 “Sorano” server CPU series, following the earlier announcement around MWC 2026. We already covered Sorano in February, but AMD has now shared more details.

EPYC 8005 is not using Zen 5c, despite being the successor to EPYC 8004 “Siena,” which used Zen 4c. AMD moved this lineup to full Zen 5 cores. This also increases L3 cache and raises clock speeds, with current SKUs reaching up to 4.5 GHz. 

Full Zen 5 instead of Zen 5cThe top model is the EPYC 8635P with 84 cores, 168 threads, 384MB of L3 cache and a 225W default TDP. AMD’s own blog says the new series scales from 8 to 84 Zen 5 cores and targets single-socket systems for edge, telco and cloud storage deployments.

Continue Reading... ]]> https://www.geeks.fyi/showthread.php?tid=21971 Wed, 13 May 2026 11:39:03 +0000 harlan4096]]> https://www.geeks.fyi/showthread.php?tid=21971 Quote:AMD Ryzen PRO 9000 expands with 3D V-Cache SKUs: Ryzen 9 PRO 9965X3D and Ryzen 7 PRO 9755X3D



AMD has expanded its Ryzen PRO 9000 desktop workstation lineup with six new Zen 5 processors. The list now includes Ryzen 9 PRO 9965X3D, Ryzen 9 PRO 9965, Ryzen 9 PRO 9955, Ryzen 7 PRO 9755X3D, Ryzen 7 PRO 9755 and Ryzen 5 PRO 9655.

This is a larger update than the first Ryzen PRO 9000 release. We previously covered the original three models, Ryzen 9 PRO 9945, Ryzen 7 PRO 9745 and Ryzen 5 PRO 9645, which were all 65W parts. The new models raise the lineup to 16 cores and introduce 120W and 170W options.

The Ryzen 9 PRO 9965X3D is the new top model. It has 16 Zen 5 cores, 32 threads, a boost clock up to 5.5 GHz, 128MB of L3 cache with 3D V-Cache and a 170W TDP. Basically, this is the PRO-system version of the Ryzen 9 9950X3D, although AMD uses a different name to separate it from the consumer desktop SKU.

Continue Reading... ]]> Quote:AMD Ryzen PRO 9000 expands with 3D V-Cache SKUs: Ryzen 9 PRO 9965X3D and Ryzen 7 PRO 9755X3D



AMD has expanded its Ryzen PRO 9000 desktop workstation lineup with six new Zen 5 processors. The list now includes Ryzen 9 PRO 9965X3D, Ryzen 9 PRO 9965, Ryzen 9 PRO 9955, Ryzen 7 PRO 9755X3D, Ryzen 7 PRO 9755 and Ryzen 5 PRO 9655.

This is a larger update than the first Ryzen PRO 9000 release. We previously covered the original three models, Ryzen 9 PRO 9945, Ryzen 7 PRO 9745 and Ryzen 5 PRO 9645, which were all 65W parts. The new models raise the lineup to 16 cores and introduce 120W and 170W options.

The Ryzen 9 PRO 9965X3D is the new top model. It has 16 Zen 5 cores, 32 threads, a boost clock up to 5.5 GHz, 128MB of L3 cache with 3D V-Cache and a 170W TDP. Basically, this is the PRO-system version of the Ryzen 9 9950X3D, although AMD uses a different name to separate it from the consumer desktop SKU.

Continue Reading... ]]> https://www.geeks.fyi/showthread.php?tid=21954 Tue, 12 May 2026 07:04:11 +0000 harlan4096]]> https://www.geeks.fyi/showthread.php?tid=21954 Quote:AMD Radeon RX 9050 uses 2048 cores but lower clocks than RX 9060 XT



AMD is preparing another Radeon RX 9000 desktop graphics card. According to information obtained by VideoCardz, the company is working on the Radeon RX 9050, a lower-clocked Navi 44 model positioned below the Radeon RX 9060 series.

The Radeon RX 9050 is not using the same core count as the Radeon RX 9060. Instead, the card is listed with Navi 44 and 2,048 Stream Processors, matching the Radeon RX 9060 XT. This means the card uses the full Navi 44 XT core configuration, rather than the 1,792-core Navi 44 XL design used by the Radeon RX 9060. Keep in mind, the specs are preliminary and only single-sourced.


Specs from one of AMD’s AIBs

Continue Reading... ]]> Quote:AMD Radeon RX 9050 uses 2048 cores but lower clocks than RX 9060 XT



AMD is preparing another Radeon RX 9000 desktop graphics card. According to information obtained by VideoCardz, the company is working on the Radeon RX 9050, a lower-clocked Navi 44 model positioned below the Radeon RX 9060 series.

The Radeon RX 9050 is not using the same core count as the Radeon RX 9060. Instead, the card is listed with Navi 44 and 2,048 Stream Processors, matching the Radeon RX 9060 XT. This means the card uses the full Navi 44 XT core configuration, rather than the 1,792-core Navi 44 XL design used by the Radeon RX 9060. Keep in mind, the specs are preliminary and only single-sourced.


Specs from one of AMD’s AIBs

Continue Reading... ]]> https://www.geeks.fyi/showthread.php?tid=21946 Fri, 08 May 2026 07:32:12 +0000 harlan4096]]> https://www.geeks.fyi/showthread.php?tid=21946 Quote:AMD Instinct MI350P launches with 4.6 PFLOPS MXFP4 compute



AMD has launched the Instinct MI350P, a new PCIe accelerator based on the CDNA 4 architecture. The card is designed for air-cooled servers and targets AI inference deployments that do not require a full OAM platform.

The MI350P is the PCIe version of AMD’s MI350 series. It has 128 Compute Units, 8,192 Stream Processors and 512 Matrix Cores. AMD lists the peak engine clock at 2.2 GHz. Basically half of the MI350X/355X specs.

Continue Reading... ]]> Quote:AMD Instinct MI350P launches with 4.6 PFLOPS MXFP4 compute



AMD has launched the Instinct MI350P, a new PCIe accelerator based on the CDNA 4 architecture. The card is designed for air-cooled servers and targets AI inference deployments that do not require a full OAM platform.

The MI350P is the PCIe version of AMD’s MI350 series. It has 128 Compute Units, 8,192 Stream Processors and 512 Matrix Cores. AMD lists the peak engine clock at 2.2 GHz. Basically half of the MI350X/355X specs.

Continue Reading... ]]> https://www.geeks.fyi/showthread.php?tid=21943 Thu, 07 May 2026 10:38:08 +0000 harlan4096]]> https://www.geeks.fyi/showthread.php?tid=21943 Quote:Qualcomm Snapdragon 6 Gen 5 brings Wi-Fi 7 and Bluetooth 6.0 to mid-range phones



Qualcomm has announced two new mobile platforms for lower-cost Android phones, the Snapdragon 6 Gen 5 and Snapdragon 4 Gen 5. Both chips are built for commercial devices launching in the second half of 2026 from OEMs including Honor, OPPO, realme and REDMI. The Snapdragon 6 Gen 5 is expected in Honor and REDMI devices, while Snapdragon 4 Gen 5 will appear first in OPPO, realme and REDMI phones.

Snapdragon 6 Gen 5

Snapdragon 6 Gen 5 is the higher-end of the two. The chip uses a 4nm process and an eight-core Kryo CPU with four performance cores up to 2.6 GHz and four efficiency cores up to 2.0 GHz. Qualcomm claims up to 21% higher Adreno GPU performance, 20% faster app launches, 18% less screen stutter and 8% better power efficiency compared with Snapdragon 6 Gen 4.
 

Source: Qualcomm

Continue Reading... ]]> Quote:Qualcomm Snapdragon 6 Gen 5 brings Wi-Fi 7 and Bluetooth 6.0 to mid-range phones



Qualcomm has announced two new mobile platforms for lower-cost Android phones, the Snapdragon 6 Gen 5 and Snapdragon 4 Gen 5. Both chips are built for commercial devices launching in the second half of 2026 from OEMs including Honor, OPPO, realme and REDMI. The Snapdragon 6 Gen 5 is expected in Honor and REDMI devices, while Snapdragon 4 Gen 5 will appear first in OPPO, realme and REDMI phones.

Snapdragon 6 Gen 5

Snapdragon 6 Gen 5 is the higher-end of the two. The chip uses a 4nm process and an eight-core Kryo CPU with four performance cores up to 2.6 GHz and four efficiency cores up to 2.0 GHz. Qualcomm claims up to 21% higher Adreno GPU performance, 20% faster app launches, 18% less screen stutter and 8% better power efficiency compared with Snapdragon 6 Gen 4.
 

Source: Qualcomm

Continue Reading... ]]> https://www.geeks.fyi/showthread.php?tid=21935 Mon, 04 May 2026 06:44:18 +0000 harlan4096]]> https://www.geeks.fyi/showthread.php?tid=21935 Quote:AMD Ryzen AI Max+ PRO 495 benchmark lists HP system with 192GB memory



AMD’s Ryzen AI Max 400 series, codenamed Gorgon Halo, is expected to be a refresh of Strix Halo rather than a new architecture. The series first leaked in January, with early reports listing Zen 5 CPU cores, RDNA 3.5 graphics and the same FP11 platform. The main changes were listed as higher clocks and likely faster memory support.

Radeon 8065S

A new PassMark entry now lists the Ryzen AI Max+ PRO 495 with Radeon 8065S. This appears to be the professional variant of the Ryzen AI Max+ 495. The chip is still a 16-core, 32-thread design, matching the Ryzen AI Max+ 395. The earlier leak listed a 5.2 GHz CPU boost and a 3.0 GHz integrated GPU clock, compared with 5.1 GHz and 2.9 GHz on Ryzen AI Max+ 395.

Currently PassMark lists only one sample. The Ryzen AI Max+ PRO 495 scores 57,525 points in CPU Mark and 4,293 points in single-thread. For comparison, Ryzen AI Max+ 395 is listed at 55,163 points and 4,161 points. That puts the new chip around 4% ahead in multi-thread and around 3% ahead in single-thread in this database.

Continue Reading... ]]> Quote:AMD Ryzen AI Max+ PRO 495 benchmark lists HP system with 192GB memory



AMD’s Ryzen AI Max 400 series, codenamed Gorgon Halo, is expected to be a refresh of Strix Halo rather than a new architecture. The series first leaked in January, with early reports listing Zen 5 CPU cores, RDNA 3.5 graphics and the same FP11 platform. The main changes were listed as higher clocks and likely faster memory support.

Radeon 8065S

A new PassMark entry now lists the Ryzen AI Max+ PRO 495 with Radeon 8065S. This appears to be the professional variant of the Ryzen AI Max+ 495. The chip is still a 16-core, 32-thread design, matching the Ryzen AI Max+ 395. The earlier leak listed a 5.2 GHz CPU boost and a 3.0 GHz integrated GPU clock, compared with 5.1 GHz and 2.9 GHz on Ryzen AI Max+ 395.

Currently PassMark lists only one sample. The Ryzen AI Max+ PRO 495 scores 57,525 points in CPU Mark and 4,293 points in single-thread. For comparison, Ryzen AI Max+ 395 is listed at 55,163 points and 4,161 points. That puts the new chip around 4% ahead in multi-thread and around 3% ahead in single-thread in this database.

Continue Reading... ]]> https://www.geeks.fyi/showthread.php?tid=21880 Tue, 21 Apr 2026 06:58:35 +0000 harlan4096]]> https://www.geeks.fyi/showthread.php?tid=21880 Quote:Half the memory bus, half the bandwidth



HKEPC has published what appears to be the first hands-on performance test of DDR5 HUDIMM, a cut-down memory design that keeps only one 32-bit sub-channel instead of the two 32-bit sub-channels used by standard DDR5 UDIMM modules. The site says the test was carried out with ASUS support and used a BIOS that recognizes HUDIMM operation.

This is not a retail HUDIMM kit review. HKEPC says it simulated the format by masking one 32-bit sub-channel on a normal DDR5 DIMM with tape, then booting the system with HUDIMM-capable firmware. In the demo, the BIOS still identified the module as a 16GB G.Skill stick, but the information page showed 8GB effective capacity, while CPU-Z reportedly detected the memory as running in “1x 32-bit” mode.

Continue Reading... ]]> Quote:Half the memory bus, half the bandwidth



HKEPC has published what appears to be the first hands-on performance test of DDR5 HUDIMM, a cut-down memory design that keeps only one 32-bit sub-channel instead of the two 32-bit sub-channels used by standard DDR5 UDIMM modules. The site says the test was carried out with ASUS support and used a BIOS that recognizes HUDIMM operation.

This is not a retail HUDIMM kit review. HKEPC says it simulated the format by masking one 32-bit sub-channel on a normal DDR5 DIMM with tape, then booting the system with HUDIMM-capable firmware. In the demo, the BIOS still identified the module as a 16GB G.Skill stick, but the information page showed 8GB effective capacity, while CPU-Z reportedly detected the memory as running in “1x 32-bit” mode.

Continue Reading... ]]>