Overclocking under nitrogen and testing AMD Ryzen 5 PRO 4650G and Ryzen 3 PRO 4350G processors. Comparison of Renoir graphics core with AMD Ryzen 5 3400G, A10-7850K, Intel Core i5-10400 and Core i3-10100

[ad_1]

In the summer of 2020, the Zen 2 architecture also reached the Advanced Processing Unit solutions, in which the integrated video card received an update, and even the segment was replenished with 6-core and 8-core models. In this material, the reader is waiting for a sortie into the wilds of experiments related to both the processor part and the integrated graphics core. The analysis of the architecture has already been done in the previous material.

AMD Ryzen 5 PRO 4650G и Ryzen 3 PRO 4350G

Let’s oppose the APU to its “flagship” predecessor – AMD Ryzen 5 3400G, purebred Matisse without an integrated graphics card – AMD Ryzen 5 3600, a representative of the Kaveri family – AMD A10-7850K with DDR3 memory, as well as counterparts under the blue flag – Intel Core i3-10100 and Intel Core i5-10400. Special guest – Intel UHD Graphics 630 under the hood of Intel Core i9-9900K. Let’s get started!

Test benches and testing methodology

The CBO circuit consisted of a HAILEA HC-300A chiller, a 15-liter liquid tank, two Hydor Seltz 1200 pumps and a CPU EK-Supremacy full Nickel water block.

Software: Microsoft Windows 10 Pro (2004) x64, AMD Chipset Software driver 2.07.14.327 + AMD Radeon Software Adrenalin 2020 edition 20.8.2, MSI Afterburner 4.6.2, RTSS 7.3.0, HWINFO 6.28, Intel DCH 27.20.100.8476 .

Stand AM4:

  • Processor #1: AMD Ryzen 5 3400G;
  • Processor #2: AMD Ryzen 5 3600;
  • Processor #3: AMD Ryzen 3 PRO 4350G;
  • Processor #4: AMD Ryzen 5 PRO 4650G;
  • cooling: CBO circuit;
  • thermal interface: Arctic Cooling MX-4;
  • motherboard: ASUS TUF Gaming B550-Plus Wi-Fi (UEFI 1004, AGESA 1.0.8.0);
  • memory: HyperX Predator HX441C19PB3/8 x2 (2×8 GB, 4266 MHz, CL 19-26-26-45-2T);
  • system storage: Kingston KC2000 250GB (SKC2000M8250G, NVMe);
  • game drive: Samsung 860 QVO 1TB (MZ-76Q1T0BW);
  • power supply: Rosewill Hercules-1600S (1600 W).

Stand FM2+:

  • processor: AMD A10-7850K;
  • cooling: CBO circuit;
  • thermal interface: Arctic Cooling MX-4;
  • motherboard: ASUS Crossblade Ranger (UEFI 1803);
  • memory: HyperX Beast 2x8GB KHX21C11T3FK8/64X (2133 MHz, CL11-12-11-32-2T);
  • graphics card: IGP AMD Radeon R7 Graphics (Specter core, 512 shaders);
  • system storage: Kingston HyperX Predator 480GB (SHPM2280P2H/480G, NVMe M.2/PCI-E);
  • game drive: Samsung 860 QVO 1TB (MZ-76Q1T0BW);
  • power supply: Rosewill Hercules-1600S (1600 W).

Stand LGA 1200:

  • Processor #1: Intel Core i3-10100;
  • Processor #2: Intel Core i5-10400;
  • cooling: CBO circuit;
  • thermal interface: Arctic Cooling MX-4;
  • motherboard: AsRock B460 Phantom Gaming 4 (UEFI 1.40);
  • memory: HyperX Predator HX441C19PB3/8 x2 (2×8 GB, 4266 MHz, CL19-26-26-45-2T);
  • graphics card: IGP Intel UHD Graphics 630;
  • system storage: Kingston KC2000 250GB (SKC2000M8250G, NVMe);
  • game drive: Samsung 860 QVO 1TB (MZ-76Q1T0BW);
  • power supply: Rosewill Hercules-1600S (1600 W).

Stand LGA 1151v2:

  • Processor: Intel Core i9-9900K;
  • cooling: CBO circuit;
  • thermal interface: Arctic Cooling MX-4;
  • motherboard: ASUS Maximus XI Gene (UEFI 1502);
  • memory: HyperX Predator HX441C19PB3/8 x2 (2×8 GB, 4266 MHz, CL19-26-26-45-2T);
  • graphics card: IGP Intel UHD Graphics 630;
  • system storage: Kingston KC2000 250GB (SKC2000M8250G, NVMe);
  • game drive: Samsung 860 QVO 1TB (MZ-76Q1T0BW);
  • power supply: Rosewill Hercules-1600S (1600 W).

Each tested processor was tested at factory frequencies (Stock) and in overclocking mode (OC):

  • Stock Core i3-10100, DDR4-2666 JEDEC, Stock UHD Graphics 630, Stock Cache;
  • Stock Core i5-10400, DDR4-2666 JEDEC, Stock UHD Graphics 630, Stock Cache;
  • Stock Core i9-9900K, DDR4-2666 JEDEC, Stock UHD Graphics 630, Stock Cache;
  • OC Core i9-9900K@5,1 ГГц, AVX offset 0, DDR4-3600 CL12-14-14-28, UHD Graphics 630 @ 1,4 ГГц, Cache 4,8 ГГц;
  • Stock Ryzen 5 3400G, DDR4-2933 JEDEC, Stock Radeon Vega 11, FCLK@1467 Mon;
  • OC Ryzen 5 3400G@4.2GHz, DDR4-3533 CL16-16-16-32 1T, Radeon Vega 11@1650MHz, FCLK@1467MHz;
  • Stock Ryzen 5 3600, DDR4-3200 JEDEC, Stock GeForce GT 1030, FCLK@1600 Mon;
  • OC Ryzen 5 3600@4.3GHz, DDR4-3800 CL16-17-16-32 1T, Stock GeForce GT 1030, FCLK@1900MHz;
  • Stock A10-7850K, DDR3-2133 JEDEC, Stock Radeon R7 Graphics, Stock NB;
  • OC A10-7850K@4.6GHz, DDR3-2133 CL10-11-10-26 1T, Radeon R7 Graphics@1200MHz, HT@2GHz;
  • Stock Ryzen 3 PRO 4350G, DDR4-3200 JEDEC, Stock Radeon Graphics, FCLK@1600 МГц;
  • OC Ryzen 3 PRO 4350G@4.4GHz, DDR4-4400 CL18-20-18-32 1T, Radeon Graphics@2200MHz, FCLK@2200MHz;
  • Stock Ryzen 5 PRO 4650G, DDR4-3200 JEDEC, Stock Radeon Graphics, FCLK@1600 МГц;
  • OC Ryzen 5 PRO 4650G@4.5GHz, DDR4-4400 CL18-20-18-32 1T, Radeon Graphics@2400MHz, FCLK@2200MHz.

The integrated graphics cards Vega 11 and Renoir Radeon Graphics are allocated 4 GB of RAM, the Radeon R7 is 2 GB, and the UHD Graphics 630 lives its own life, regardless of the settings, up to 8 GB of video memory is consumed.

The temperature in the room was 22–25 degrees Celsius. The power consumption of the platform was measured using the Feron TM55 energy meter for 10 minutes in load mode (LinX AMD Edition run for AMD processors, as well as LinX 0.7.0 for Intel) and idle. Fan speed control on all stands is automatic. The frequency potential was studied with a voltage step of 0.025 V, the stability of the system was checked by a complex of loads in the form of mining on the RandomX algorithm, a LinX run (problem size 28432, 10 iterations), as well as passing the HWBOT x265 2.2 Benchmark on the 4K preset. Each benchmark was run five times, the results indicate the arithmetic mean, the same goes for games. The games recorded telemetry (1% low fps + Average fps) using the Riva Tuner Statistics Server. Further specific methods. In CS:GO, a test on the FPS Benchmark workshop map, from the beginning of the test to the third signal on a black screen at the end of the benchmark. In DOTA 2, replay of the final game OG vs Liquid at The International 2019, 19:45-21:15 on automatic camera. In GTA V, the last scene of the benchmark from the waterfall to the black screen after a collision with a truck. DOOM Eternal – The starting position of the player in the Stronghold of Doom with a double-barreled gun in his hands and the weapon view in the center, then move to the end of the room. Crysis 2 – Beginning of the Dead Man Walking mission with a SCARAB equipped, moving forward to the end of the walkie-talkie monologue.

Branded coolers

AMD Ryzen 5 PRO 4650G и Ryzen 3 PRO 4350G

How will these processors feel under a typical bundled AMD Wraith Spire cooler? It’s hot, in a nutshell. 91 degrees peak for the Ryzen 3 PRO 4350G and 93 degrees for the Ryzen 5 PRO 4650G. So providing processors with a more efficient cooler immediately guarantees higher acceleration frequencies for a longer period of time.

AMD Ryzen 5 PRO 4650G и Ryzen 3 PRO 4350G

AMD Ryzen 5 PRO 4650G и Ryzen 3 PRO 4350G

Next, we supply the processors with water cooling.

frequency potential

AMD Ryzen 5 PRO 4650G и Ryzen 3 PRO 4350G

Against the background of APU Picasso, the frequency potential has grown, and at equal frequencies, you can get by with lower voltages. Medium voltages are similar to Zen 2 processors. Processor batches: Ryzen 3 PRO 4350G – ED 2025SUS (Week 25, 2020, Suzhou), Ryzen 5 PRO 4650G – ED 2022SUS (Week 22, 2020, Suzhou). Both can not be called outstanding, however, you can’t call them “potatoes” either, we will consider them good specimens at this point in time.

AMD Ryzen 5 PRO 4650G и Ryzen 3 PRO 4350G

FCLK hit 2 GHz!

This moment, perhaps, is in second place in terms of interest after working with RAM. Yes, now FCLK can reach impressive frequencies of 2200-2266 MHz, but the test platform on hand is not a flagship solution, so stable operation can only be ensured at 2200 MHz. Both tested processors did not conquer the frequency of 2266 MHz, and were unstable at 2233 MHz.

Automation forcibly sets the FCLK = 1/2 MEM_CLK mode as soon as the user uses the XMP profile, or sets the memory frequency above 4000 MHz by itself, and this leads to a performance loss. The next mini-test involved a Ryzen 5 PRO 4650G at a fixed frequency of 4000 MHz, RAM at a frequency of 4400 MHz with CL18-20-18-34 1T timings, an integrated graphics card at factory settings and, in fact, FCLK in 1: 1 mode and 1/2.

AMD Ryzen 5 PRO 4650G и Ryzen 3 PRO 4350G

AMD Ryzen 5 PRO 4650G и Ryzen 3 PRO 4350G

AMD Ryzen 5 PRO 4650G и Ryzen 3 PRO 4350G

AMD Ryzen 5 PRO 4650G и Ryzen 3 PRO 4350G

AMD Ryzen 5 PRO 4650G и Ryzen 3 PRO 4350G

AMD Ryzen 5 PRO 4650G и Ryzen 3 PRO 4350G

Noticeable losses are observed in mining, as well as working with archives. The rest almost does not suffer from the cut frequency. Do not be lazy to check the FCLK mode of operation, it takes almost no time!

Mining? Seriously?

Believe it or not, IGP Renoir can even mine cryptocurrency! But please don’t rush to invest in non-obvious solutions, because when you see such power on the same Ethash algorithm…

AMD Ryzen 5 PRO 4650G и Ryzen 3 PRO 4350G

… everyone will cause a local shortage of APUs, and an unjustified rise in the price of these processors, dooming themselves to hatred from those who are not in the know.

The power of the IGP

If you describe the situation concisely – the APU has a very suitable modern budget graphics core, something like the AMD Radeon HD 7850 (in factory mode) with the ability to “bite off” up to 16 GB of the video buffer for your needs.

AMD Ryzen 5 PRO 4650G и Ryzen 3 PRO 4350G

It seems to be great, but it seems to be pointless, although, thanks to the features of the new architecture, the maximum memory frequency is now around 4566 MHz (predecessors were content with 3866 MHz). This is based on operation in the FCLK:MEM = 1:1 mode. It is clear that the architecture can squeeze out all 5000 MHz, but these will be real corn frequencies, which will be proved by testing with liquid nitrogen. It is also worth noting the violent frequency nature of the IGP, it was possible to squeeze out a stable 2400 MHz on the core at a voltage of 1.2 V. When you increase to 1.3 V, you can count on 2500 MHz, however, as in all previous APUs, a much greater increase is provided by working with RAM, not core clock.

Renoir’s integrated graphics cards differ from each other despite identical Radeon Graphics names. And in my personal opinion, this is the wrong decision, because it was easier to navigate by common nouns, like Vega 3, Vega 11 from the previous generation. The main differences lie in the computational units (CPU Model = ROP/TMU/Shaders, Core Clock):

  • Ryzen 3 PRO 4350G = 8/24/384, 1700 MHz;
  • Ryzen 5 PRO 4650G = 8/32/448, 1900 MHz;
  • Ryzen 7 PRO 4750G = 8/32/512, 2100 MHz;
  • Ryzen 5 3400G = 16/44/704, 1400 MHz.

It is immediately striking that the new IGPs in all cases have fewer pixel pipelines, texture mappers and shader processors than their predecessor in the face of Vega 11. It is also worth paying attention to the difference between Ryzen 3 PRO 4350G and Ryzen 5 PRO 4650G – the junior solution has less TMU. But the IGP Ryzen 7 PRO 4750G did not increase the texture units against the background of the Ryzen 5 4650G, therefore the graphics of the older model are not as different from the six-core as we would like.

As practice shows, even the IGP Ryzen 3 Pro 4350G takes 2100 MHz on the core without any problems, so the frequencies are easy to compensate, but we will find out the increase from the computing units right now.

Comparison of IGP Renoir Radeon Graphics

In this mini-test, IGP Ryzen 3 Pro 4350G, as well as IGP Ryzen 5 4650G, take part in equal conditions on the same stand. The RAM frequency was 3600 MHz, XMP profile with CL18 timings, both processors were fixed at 4000 MHz, the IGP core frequency was 2100 MHz.

AMD Ryzen 5 PRO 4650G и Ryzen 3 PRO 4350G

AMD Ryzen 5 PRO 4650G и Ryzen 3 PRO 4350G

AMD Ryzen 5 PRO 4650G и Ryzen 3 PRO 4350G

AMD Ryzen 5 PRO 4650G и Ryzen 3 PRO 4350G

AMD Ryzen 5 PRO 4650G и Ryzen 3 PRO 4350G

It’s safe to say that the difference is minimal when we mean tasks with a focus on the graphics card. Of course, the best processor in the complex gives a more noticeable gap, and guarantees better overclocking potential as a more successful instance. It is better to choose based on the number of processor cores for the desired tasks. The author will again repeat the obvious: “Work with RAM, not the IGP itself!”.

LN2, aka liquid nitrogen. Renoir@XOC

First off, great news for nitrogen-breathing enthusiasts! FCLK now doesn’t go crazy around -50 degrees, which means PCI Express processor lines remain stable and there is no need to unfreeze the platform for a successful start. In addition, you can forget about the FCLK 1400-1500 MHz safe limit, which is relevant for Matisse processors, now you can work at FCLK 2200 MHz even at -100 degrees. For those who are not in the know, I remind you that the more processor cores on board, the more the overall performance ended up being cut due to the forced operation of FCLK at 1500 MHz, not to mention the 800-900 MHz range.

AMD Ryzen 5 PRO 4650G и Ryzen 3 PRO 4350GAMD Ryzen 5 PRO 4650G и Ryzen 3 PRO 4350G

The built-in graphics card is also notably scalable in frequency and works stably at 2800 MHz! Even 2900 MHz with partial stability, but it was not possible to break through 3000 MHz. Vega 11 could only handle 2000 MHz. Here is the Fire Strike result from the Ryzen 5 PRO 4650G:

AMD Ryzen 5 PRO 4650G и Ryzen 3 PRO 4350G

Why, even the discrete GeForce GTX 760 or Radeon R9 270 are left behind!

The processor apparently does not have CBB. CB is in the region of -110…120 degrees, which corresponds to a typical AM4 APU. The frequency limit is 5300 MHz, the same Ryzen 5 3600 squeezed out 5424 MHz.

Unfortunately, the lack of L3 cache spoils all the fun, even with the improved FCLK with memory, the processor is not able to outperform the Ryzen 5 3600, so in competitive overclocking it will be interesting only in terms of IGP.

AMD Ryzen 5 PRO 4650G и Ryzen 3 PRO 4350G

[ad_2]

Leave a Reply

Your email address will not be published. Required fields are marked *