Review and testing Xeon E5-2699 v3: 18-core monster in actual games and applications

We don’t want to go around with the traditional introduction this time, and the reason is only five words: eighteen cores, thirty-six threads. We present to your attention the ultimate Haswell, and at the same time the crown of engineering philosophy of 2014. Ladies and gentlemen, we have Xeon E5-2699 v3 – one of the most powerful 22-nanometer processors in existence!

But first, we still have to take a look at the prices of the processors presented in this article:

An 18-core, 36-thread processor is priced cheaper than the new Core i3-10100F? But that’s the plain truth. In recent months, virtually all of the Xeon E5 v3 series chips have lost significant price. Up to the point of obscenity: A 6-core 2620 v3 can be bought for $1 on local online flea markets, while a 12-core 2670 v3 is going for just $7!

What about your 18-core? – you may ask. We got our copy of 2699 v3 for a modest $45. The purchase was made not on the classic AliExpress, but at a trusted local seller, who sold this processor not without his markup.

Thus, it can be stated that it is possible to buy Xeon E5-2699 v3 chip at a more attractive price. But we did not take any risks and stopped at a guaranteed reliable store.

CPU

S-Spec of the test sample SR1XD. This is a release version of the CPU for workstations in tray version.

2699 v3 is a bit wider than its LGA 2011 v3 socket brethren, which can be seen with the naked eye on the following photos:

However, it doesn’t mean that Xeon E5-2699 v3 has any specific requirements to the board socket. This chip can be installed into the vast majority of LGA 2011 v3-compatible motherboards without any problems. Ideally, there should be no problems with launching it, but there are still some nuances. We will talk more a little bit later.

CPU specifications

The Intel Xeon E5-2699 v3 chip is based on an 18-core, 36-thread, 22nm revision R2 die based on the Intel Haswell-EP architecture. This CPU configuration is the maximum possible for mass server chips of the third version. Thus, we are facing a top representative of its generation (there is also Xeon E5 2696 v3, but this beast is much rarer, plus, it is probably custom).

This “pebble” is capable of working in both single-processor and dual-processor configurations, but in our case the first variant is used.

2699 v3 has 256 KB of L2 cache per core and a total, truly gigantic for its time 45 MB L3 cache. This CPU is designed for installation in motherboards with LGA 2011 v3 socket, its nominal frequency is equal to 2300 MHz, but with the help of Turbo Boost technology it can increase up to 3600 MHz for one core, and up to 2800 MHz for all 18 cores, 36 threads. We would also like to note that Xeon E5-2699 v3 supports the entire current set of processor instructions, including AVX2 and FMA3.

Xeon E5 2699 v3 speed of integrated memory controller (IMC) and L3 cache, as well as in the whole series of v3 chips, is not tied to the frequency of cores and uses its own divider. In our case it is 3000 MHz.

As for the supported RAM frequency, officially the maximum RAM speed with which 2699 v3 can function is limited to four-channel DDR4-2133 MHz. And unfortunately, it is impossible to change it by overclocking or other methods.

The TDP level of this chip is set at 145 watts. And judging by our measurements, it is not difficult to achieve such power consumption figures at all. And even without using highly specialized benchmarks like LinX with AVX instructions.

In games with good multithreading optimization the power consumption of E5-2699 v3 stays at the level of 90-120 watts, and in professional software the chip often falls within its TDP limit.

Please note: E5-2699 v3 requires special attention to its cooling. Coolers with four or more heat pipes are suitable. If you are building a system for round-the-clock work (rendering, coding and so on), it is better to buy a tower-type cooling system with six heat pipes.

In addition to the processor itself, take care of sufficient cooling of your motherboard’s power supply circuits! Even quite good motherboards with decent VRM (for Chinese devices) can hardly handle E5-2699 v3!

Undervolting and Unlock Turbo-Boost

Obviously, no one will use such processor in stock, especially since most Chinese motherboards already come with Unlock Turbo-Boost in the BIOS as a basic configuration. If you are unlucky and your motherboard doesn’t have this “luxury” from the factory, the Internet is full of all kinds of instructions and ready-made BIOS builds with Unlock Turbo-Boost embedded for almost any board. We won’t stop here in detail and will go straight to the processor undervolt.

In our case we managed to achieve stability at the following voltages: CPU cores -80mv (in games and simple computing our sample was able to function stably at -100mv), L3-cache -70mv, memory controller -60mv.

In this simple way we slightly increased performance in complex calculations, such as rendering using AVX2 instructions, but at the same time significantly increased the speed of rendering frames in games.

So why is that? It’s quite simple. Let’s take for example the current version of Cinebench R24, which engine is built with AVX2 in mind. It’s quite logical that without Unlock Turbo-Boost, the CPU frequency at multithreaded rendering is only 2800 MHz. What about after UTB? Only 2900 MHz. That is, there is an increase in frequency, but it is very far from the potential 3600 MHz.

The reason for such a modest frequency increase is banal: the processor is within its own TDP limit of 145 watts. And this is not enough to make all 36 threads function at 3600 MHz.

Of course, there is a possibility to bypass this limit (disable SVID/FIVR in BIOS settings). But the problem is that this trick works well on a conditional E5-2670 v3, or E5-2660 v3, but not on the high-end 2699 v3 (instead of 2900, it will be 3000 MHz). Besides TDP limit, processors also have EDP (Electrical Design Point) limit, which is essentially a consumption limit in terms of current limit. And EDP cannot be bypassed in any way.

Plus, since the CPU does start to consume much more power, disabling SVID/FIVR can harm your motherboard’s power supply circuitry. In the case of the 2699 v3, we recorded crossing the 180W threshold. And let’s face it: not every Chinese motherboard can handle such TDP. Therefore, in our humble opinion, it is better not to use SVID/FIVR disabling in a bundle with Xeon E5-2699 v3 chip.

Heavy load has been dealt with. And what about games after Unlock Turbo-Boost? You know – everything is quite good. Often, the CPU frequency stays at 3400-3600 MHz, gravitating closer to the last figure. Even in Starfield it didn’t fall below 3500 MHz, which is an excellent result.

Hyper-Threading

If you have read our article about testing Xeon E5-2670 v3, you probably remember that the technology of simultaneous processing of two threads on one core (Hyper-Threading) in case of multi-core solutions can play a cruel joke with the final CPU performance in poorly optimized programs (often in games).

However, those conclusions were made on the basis of CPU tests on Windows 10, and at the start of this benchmark session we switched to the current Microsoft Windows 11 product. We won’t bore you – nothing has changed. Thus, there is a strong possibility that the problem lies not in the OS, but in the Haswell hardware scheduler built into the processor, but we won’t say whether this is actually the case.

Anyway, after a number of tests we found out that in some applications, as well as in the vast majority of games, the Xeon E5-2699 v3 demonstrates relatively weak results with active Hyper-Threading technology (and it is this technology that allows an 18-core processor to have a mind-blowing 36 threads). And the reason here is far from limitations in the thermal or electrical package. The problem lies in the difficulty of distributing a small load on a huge number of executive blocks, and as a consequence, the slowdown of this or that weakly-optimized program.

For this reason, we decided to test the processor in two modes again: with active Hyper-Threading technology (hereinafter referred to as HT-ON) and without it (hereinafter referred to as HT-OFF). Thus, on the graphs you can see two versions of Xeon E5-2699 v3 chip at once:

• Xeon E5-2699 v3@2900-3600 MHz (18/36) UTB, HT-ON – is a full processor with active multi-threading technology, or 18 cores/36 threads, as well as Unlock Turbo-Boost.
• Xeon E5-2699 v3@3000-3600 MHz (18/18) UTB, HT-OFF – version with deactivated multi-threading technology, in other words “pure 18 cores ”+ Unlock Turbo-Boost.

You’ve probably already noticed that the minimum frequencies of the two versions are slightly different. This is not a mistake, in heavy loads with AVX2, the variant without Hyper-Threading is able to hold 3000 MHz for all cores, while the full-fledged CPU is only 2900 MHz. But otherwise (games, simple software, everyday tasks and so on), the frequencies of both versions are completely the same. For example, in the Starfield which is the hardest for processors, both versions operate at 3500 MHz.

Test bench, software and CPU settings

Resizable BAR has been activated on platforms that have BIOS support for it.

Test bench

• Intel CPU: Core i5-4670K, Core i3-10100F, Core i5-10400F, Core i7-11700K, Core i3-12100F, Core i5-12400F, Xeon E5-1620, Xeon E5-2690, Xeon E5-2630 v3, Xeon E5-2670 v3, Xeon E5-2699 v3, Xeon E5-2640 v4;
• AMD CPU: Ryzen 5 1600X, Ryzen 7 1800X, Ryzen 5 3600;
• CPU cooling: Cooler Master Hyper 212 Black Edition (RR-212S-20PK-R1);
• LGA 1150 RAM: 2 x 8 GB HyperX Genesis Na’Vi Edition (KHX16C9C2K2/8);
• LGA 1200, LGA 1700 and AM4 RAM: 2 x 8 GB Corsair Vengeance RGB PRO (CMW16GX4M2C3600C18, Micron E-Die chips);
• LGA 2011 RAM: 4 x 8 GB Micron MT18KSF1G72PZ-1G6E1HI (D9PQL chips);
• LGA 2011 v3 RAM: 4 x 4 GB G.SKILL DDR4 F4-2400C15S-4GNT (Hynix MFR chips);
• LGA 1150 Motherboard: ASRock Z87 Pro4;
• LGA 1200 Motherboard: GIGABYTE Z490 AORUS ELITE AC for Core i7-11700K, ASUS PRIME B560M-A for Core i3-10100F and Core i5-10400F;
• LGA 1700 Motherboard: MSI PRO Z690-A DDR4 (MS-7D25);
• LGA 2011 Motherboard: DELL T3610 (09M8Y8);
• LGA 2011 v3 Motherboard: Kllisre X99-D8 (AD12) with modified BIOS (Unlock Turbo Boost, Undervolt, as well as unlocked timings control);
• AM4 Motherboard: ASUS TUF GAMING B450M-PRO II;
• Graphics card: PALIT GAMEROCK GeForce RTX 3090 24 GB (~1900/19000 MHz, Power Limit 113%);
• SSD: 2 x KINGSTON SUV400S37120G 120.0 GB (AMD/Intel Windows 11), SAMSUNG 870 EVO 1 TB (Games/Applications);
• PSU: Chieftec GPS-1250C.

Software

• Operating system: Windows 11 Pro x64 with latest updates as of February 2024. Core isolation/memory integrity is disabled;
• Graphics card drivers: NVIDIA GeForce 551.23 WHQL;
• FPS measurement software: MSI Afterburner 4.6.5;
• Games: testing was performed on current versions of games as of February 2024;
• Game settings: testing was performed at the highest possible graphics settings in 1080p resolution.

Processors Settings

• Intel CPUs:
  • Core i5-4670K@3400-3800MHz, Dual Channel DDR3@1600MHz (motherboard auto: 9-9-9-27);
  • Core i5-4670K@4500MHz, UnCore@4100 MHz, Dual Channel DDR3@2133MHz (11-12-12-28), vCore voltage – 1.4v, VCCIN/VRIN voltage – 1.9v, CPU Cache – 1.32v, SA voltage – 1.25v, DDR3 voltage – 1.69v;
  • Core i3-10100F@3600-4300 MHz, Dual Channel DDR4@2667 MHz (motherboard auto: 16-18-18-36);
  • Core i3-10100F@3600-4300 MHz, Dual Channel DDR4@4000 MHz (16-21-21-42), IO voltage – 1,2v, SA voltage – 1.2v, DDR4 voltage — 1.420v;
  • Core i5-10400F@2900-4300 MHz, Dual Channel DDR4@2667 MHz (motherboard auto: 16-18-18-36);
  • Core i5-10400F@2900-4300 MHz, Dual Channel DDR4@4000 MHz (16-21-21-42), IO voltage – 1,2v, SA voltage – 1.2v, DDR4 voltage — 1.420v;
  • Core i7-11700K@3600-5000 MHz, Dual Channel DDR4@3200 MHz (motherboard auto: 16-18-18-36);
  • Core i7-11700K@4800 MHz, UnCore@4400 MHz, Dual Channel DDR4@3600 MHz (16-19-19-38) GEAR 1, vCore voltage – 1.33v, mem-IO voltage – 1.2v, SA voltage – 1.23v, DDR4 voltage – 1.36v;
  • Core i3-12100F@3300-4300 MHz, Dual Channel DDR4@3200 MHz (motherboard auto: 18-19-19-39) GEAR 1;
  • Core i3-12100F@3300-4300 MHz, Dual Channel DDR4@3600 MHz (16-19-19-38) GEAR 1;
  • Core i5-12400F@2500-4400 MHz, Dual Channel DDR4@3200 MHz (motherboard auto:18-19-19-39) GEAR 1;
  • Core i5-12400F@2500-4400 MHz, Dual Channel DDR4@3600 MHz (16-19-19-38) GEAR 1;
  • Xeon E5-1620@3600-3800 MHz, Quad-Channel DDR3@1600 MHz (motherboard auto: 11-11-11-28);
  • Xeon E5-2690@2900-3800 MHz, Quad-Channel DDR3@1600 MHz (motherboard auto: 11-11-11-28);
  • Xeon E5-2630 v3@3200 MHz (Unlock Turbo Boost is labeled as UTB on the graphs), vCore -60mv, UnCore -50mv, SA -50mv, Quad-Channel DDR4@1866 MHz (10-10-10-24);
  • Xeon E5-2670 v3@3100 MHz (Unlock Turbo Boost is labeled as UTB on the graphs), vCore -50mv, UnCore -50mv, SA -50mv, Quad-Channel DDR4@2133 MHz (12-12-12-35);
  • Xeon E5-2699 v3@2900-3600 MHz, Hyper-Threading – ON, 18 cores / 36 threads, (Unlock Turbo Boost is labeled as UTB on the graphs), vCore -80mv, UnCore -70mv, SA -60mv, Quad-Channel DDR4@2133 MHz (12-12-12-35);
  • Xeon E5-2699 v3@3000-3600 MHz, Hyper-Threading – OFF, 18 cores / 18 threads (Unlock Turbo Boost is labeled as UTB on the graphs), vCore -80mv, UnCore -70mv, SA -60mv, Quad-Channel DDR4@2133 MHz (12-12-12-35);
  • Xeon E5-2640 v4@2400-3400 MHz, Quad-Channel DDR4@2133 MHz (12-12-12-28).
• AMD CPUs:
  • Ryzen 5 1600X@3600-4000 MHz, Dual Chanel DDR4@2667 MHz (motherboard auto: 18-18-18-44);
  • Ryzen 5 1600X@4000 MHz, Dual Chanel DDR4@3600 MHz (16-19-16-38), vCore voltage – 1.42v, SOC voltage — 1.075v, cLDO VDDP voltage – 0.975v, DDR4 voltage — 1.350v;
  • Ryzen 7 1800X@3600-4000 MHz, Dual Chanel DDR4@2667 MHz (motherboard auto: 18-18-18-44);
  • Ryzen 7 1800X@3900 MHz, Dual Chanel DDR4@3600 MHz (16-19-16-38), vCore voltage – 1.4v, SOC voltage — 1.1v, cLDO VDDP voltage – 1.025v, DDR4 voltage — 1.350v;
  • Ryzen 5 3600@3600-4200 MHz, Dual Chanel DDR4@3200 MHz (motherboard auto: 22-22-22-53) Infinity Fabric@1600 MHz (1:1);
  • Ryzen 5 3600@4325 MHz, Dual Chanel DDR4@3733 MHz (16-19-16-38) Infinity Fabric@ 1866 MHz (1:1), vCore voltage – 1.39v, SOC voltage — 1.125v, cLDO VDDP voltage – 1.075v, VDDG CCD voltage – 1.025v, VDDG IOD voltage – 1.025v, DDR4 voltage — 1.380v;

You can learn more about almost all the tested processors from our full-fledged materials or by visiting the blogs section, where you can find small notes.

We would like to remind you about the methodology of processor testing: each benchmark/application/game was run five times, then we calculated the average index among the five runs and recorded it in the final results. All games and applications were installed on SSD.

Test results

Data archiving

7-Zip

The most popular free file archiver 7-Zip supports several compression algorithms and many data formats, including its own format 7z with a highly efficient compression algorithm LZMA. This archiver is able to use a huge number of processor threads, as well as not badly responds to increasing the frequency of RAM.

File compressing

For quite a long time we haven’t had a processor capable of beating Xeon E5-2670 v3 in data archiving. Even an overclocked Core i7-11700K, on which we pinned certain hopes, failed to cope with this task. Nevertheless, the leadership of the 12-core couldn’t last forever. The 18-core, 36-thread E5-2699 v3 outperformed its little brother by an impressive 44%!

Disabling Hyper-Threading expectedly reduced the performance of the tested CPU, but even in this case, the advantage over the nearest rival reaches 20%.

File decompressing

Low-frequency chips are often not particularly strong at decompressing files, but this is not the case with the 2699 v3. The 36-threaded variant managed to beat the i7-11700K by 17%, and the pure 18-core actually equaled the $300 CPU.

3D rendering

Blender

Blender is a versatile, free and open source 3D creation software package. It supports the entire 3D modeling cycle – from rigging, animation, simulation, rendering, compositing and motion tracking, to video editing and game creation.

We perform a CPU rendering speed test using Blender’s built-in Cycles engine and a BMW scene.

If you use the Cycles engine built into Blender, the Xeon E5-2699 v3 will provide you with tangible savings in time spent on the final render (about 11%). And this is in comparison with a much more expensive and overclocked Core i7-11700K. But the most amusing here is another thing: besides the fact that 36-thread Haswell turns out to be more productive, it also consumes significantly less power compared to 16-thread Rocket Lake (145 vs. 180-210 watts).

Without multithreading technology, E5-2699 v3 loses about 20% of performance. Even so, its results are not far from the basic i7-11700K.

Corona 10 Benchmark

Corona 10 Benchmark is based on the Corona 10 rendering engine, popular among professionals. This engine is available for rendering scenes in 3ds Max and Cinema 4D. The benchmark evaluates rendering speed on CPUs using proprietary technologies.

In Corona 10, however, the 2699 v3 chip does not shine. Of course, considering the price of the processor we can say that the results are decent, but in connection with the previously mentioned achievements of the 36-thread monster – 8% gapfrom the stock 11700K looks unconvincing.

Without Hyper-Threading, the 18-core Xeon E5-2699 v3 is only able to compete with the 6-core, 12-thread Core i5-12400F.

Cinebench R23

A recently outdated but still relevant version of the benchmark of the very popular editor for computer 3D animation, modeling, simulation and rendering Cinema 4D.

Singlecore score

Single-threaded test Cinebench R23 is not of much interest to us, because obviously, the results of 2699 v3 will not differ much from other Haswell chips, but multi-threaded test is another matter.

Multicore score

The 36-thread chip regains its supremacy after its defeat in Corona. Even though it failed to beat the overclocked Core i7-11700K, it still beat its stock version.

Without simultaneous processing of two threads per core, our 18-core is once again only able to pull in a Core i5-12400F.

Cinebench R24

The newest version of the benchmark of Cinema 4D, a very popular 3D animation, modeling, simulation and rendering editor. Starting from this release, the main CPU test requires AVX2 instruction set support by the processor to run.

Singlecore score

Multicore score

In the current version of Cinebench we were in for a surprise. Developers do not eat their bread in vain, thanks to which optimization for multi-core solutions became even better. In this release, 36-thread 2699 v3 outperforms both the stock and overclocked 11700K version. That’s an amazing result!

But that’s not all, the pure 18-core performs on par with the base 16-threaded Rocket Lake, and quite noticeably outperforms the 12-threaded Core i5-12400F!

If you’re using Maxon software and haven’t yet upgraded to the current Cinema 4D, apparently now is the time to do so.

V-Ray 6 Benchmark

Current version of the V-Ray render engine benchmark. Various versions of this product are available for many solutions, including 3DS Max, Maya, Cinema 4D, Blender and Unreal Engine.

V-Ray 6 was not so favorable to Xeon E5-2699 v3. In this rendering engine the 36-threaded chip failed to show ultimatum performance. It was only slightly ahead of Core i5-12400F, and without HT the situation is even worse.

xNormal – baking texture maps

External programs for baking textures are not so popular anymore, but xNormal is still used by tens of thousands of game developers. And all thanks to its very user-friendly interface and a lot of useful features.

Normal Map render

Ambient Occlusion Map render

In xNormal, the Xeon E5-2699 v3 is at it again. The performance of the 36-threaded version matches the overclocked Core i7-11700K, while the 18-core variant falls between the stock 16-threaded Rocket Lake and 12-threaded Alder Lake.

Overall performance

CPU-z Benchmark

CPU-z information utility benchmark can tell us little about the real performance of a CPU and is rather optional at the moment. However, according to a long tradition, we conduct tests in this discipline as well.

Singlecore score

Multicore score

Just as we write in every article – CPU-z Benchmark in no way reflects the real CPU performance. Based on the utility’s internal test, Xeon E5-2699 v3 is faster than 11700K by an impressive 25% in the multithreaded test.

Geekbench 6

This benchmark measures single-core and multi-core CPU performance in a multitude of task types, from checking email to taking photos and playing music, or all of these at the same time. In addition, Geekbench 6 measures performance in new CPU applications such as augmented reality and machine learning.

Singlecore score

Multicore score

Geekbench 6 is the first to report a performance degradation when using all 36 threads of the Xeon E5-2699 v3. The 18-core version of the chip is only 2% faster, but the very fact that additional threads harm the final speed of the processor is a bit disappointing.

On top of that, the E5-2699 v3 in general does not cause much excitement in this benchmark. The performance of the high-end Haswell is between the overclocked Ryzen 5 3600 and Core i5-12400F, which is not bad in principle, but it doesn’t reach the level of Core i7-11700K.

Video encoding

HandBrake

HandBrake is a free, open source program for converting video from almost any format to a number of modern, widely supported codecs such as AV1, H265, H264 and many others.

AV1 codec

H265 codec

Video transcoding using AV1 and H265 codecs is clearly not the Xeon E5-2699 v3’s strongest point. In both cases, the speed of the ultimatum Haswell is on the level of Ryzen 5 3600, and significantly falls short of 16-thread Rocket Lake.

In addition, we again see a slight drop in performance when using a full 36-threaded processor.

Internet browsing

JetStream 2.1

JetStream 2.1 is a set of JavaScript and WebAssembly benchmarks focused on the most modern browsers and web applications. According to our observations, most of the subtests inside the benchmark are capable of using up to 4 threads, which is quite typical for web surfing.

Speedometer 2.1

Speedometer tests the responsiveness of web applications in the browser. The benchmark simulates user actions to add, finalize, and delete items on a to-do list using several TodoMVC examples. Some call the DOM API directly from ECMAScript 5 (ES5), ECMASCript 2015 (ES6), ES6 transpiled into ES5, and Elm transpiled into ES5. Others use one of eleven popular JavaScript frameworks: React, React with Redux, Ember.js, Backbone.js, AngularJS, (new) Angular, Vue.js, jQuery, Preact, Inferno, and Flight.

Kraken 1.1

Kraken is a JavaScript performance benchmark created by Mozilla that measures the speed of execution of several different test cases taken from real applications and libraries. The subtests include audio processing using the DSP.js library, image filtering operations, JSON parsing, and cryptographic operations.

Working with Internet browsers is also not a distinctive feature of such a multi-core solution. If your professional employment is associated with Internet surfing, it is better to choose a more modern high-frequency processor.

Software test analysis

According to our research, the 36-thread variant of the Xeon E5-2699 v3 is more promising in terms of software performance. Even when taking into account a small victory in video encoding and partial superiority in Internet surfing, the final productivity of the 18-core version is significantly lower. This is especially noticeable in rendering 3D images.

Games, synthetics

3DMark Time Spy

Time Spy is a DirectX 12 benchmark for Windows 10/11 gaming PCs. Thanks to the engine built from the ground up, Time Spy supports all the new DirectX 12 API features: asynchronous computing, AMD Crossfire and NVIDIA SLI, as well as multithreading.

In the first near-game benchmark we observe a 15% drop in performance when using Xeon E5-2699 v3 with Hyper-Threading. Moreover, it is quite important to note that the decrease in productivity is recorded not only in the graphics subtest, but also in the CPU subtest.

Games

Assassins Creed Valhalla

The actual major part of the Assassins Creed franchise. The game is built on Ubisoft’s Anvil engine, and is capable of fully utilizing up to 12 compute threads. Partial 16-24 threads.

In Valhalla, the performance of both versions of Xeon E5-2699 v3 is at a decent level. Nevertheless, the variant without HT outperforms the full-fledged chip by 18% and performs almost on par with Ryzen 5 3600 and i3-10100F.

Baldur’s Gate 3

One of the best turn-based RPGs of modern times, and the best game of last year. Baldur’s Gate 3 is based on Divinity 4.0 Engine, which can utilize up to 16 threads, but the engine works best with 12-threaded “stones”. In addition, it should be noted that the project has an extremely high level of detail, because of which it often accesses RAM. As a consequence – Baldur’s Gate 3 “likes” a capacious third-level cache and high-frequency RAM.

Our test scene is in the city of Baldur’s Gate, which is the third act of the game.

The third part of the Baldur’s Gate franchise is the hardest project for any CPU, and unfortunately, the high-end Haswell is no exception. Xeon E5-2699 v3 is unable to produce the cherished 60 FPS in densely populated locations, plus the minimum recorded FPS often drops below 35 frames per second.

Disabling HT allows you to gain an additional 15% of performance, which certainly makes gameplay slightly more comfortable, but it does not change the situation dramatically.

Cyberpunk 2077

The latest creation of CD Projekt RED studio is based on the REDengine 4 engine and is able to use more than 16 CPU threads. Also, the game responds very positively to the increase in RAM frequency and reduce lag.

Ray Tracing – ON, Path Tracing – ON

Cyberpunk 2077 with active ray tracing runs great on both processor variants, but disabling Hyper-Threading allows for truly amazing performance and achieves the results of the overclocked Core i5-10400F and Ryzen 5 3600.

Ray Tracing – OFF, Path Tracing – OFF

Without ray tracing, the performance of both versions increases proportionally, but their position in the graphs does not change.

However, it is impossible not to note the following: thanks to the titanic work of CD Projekt RED to optimize its project for multi-core processors, the minimum recorded FPS on Xeon E5-2699 v3 is significantly higher than any of the representatives of the Core i3 line.

Far Cry 6

This project is based on Dunia Engine v2. All its very long life, since the second Far Cry, this engine was famous for its poor optimization for multithreaded CPUs. According to our observations, the game is able to work more or less adequately with eight threads. Some chips with more than 10 cores on board and with SMT/HT technology (simultaneous processing of two or more threads on one core) may lose performance in this game. However, this does not apply to all existing CPUs.

Ray Tracing – ON

FarCry 6 is not able to adequately handle multi-core CPUs, making the 36-threaded Xeon E5-2699 v3 effectively equal to its little brother 2670 v3. However, this is correctable. When Hyper-Threading is disabled, the 18-core processor demonstrates much better results and even reaches the performance of Core i3-10100F.

Ray Tracing – OFF

Disabling ray tracing has an extremely positive effect on FPS, thanks to which 2699 v3 manages to outperform the stock i3-10100F and R5 3600. But RT deactivation had no effect on the balance of power between HT-ON and HT-OFF versions of the processor.

Marvel’s Spider-Man Miles Morales

Former PlayStation console exclusive Marvel’s Spider-Man Miles Morales, is a relatively digestible PC port. In general, Insomniac Games’ own engine is able to utilize multi-core CPUs as intended, but the stability of games based on it suffers greatly and leaves much to be desired.

Ray Tracing – ON

The Xeon E5-2699 v3 is great for playing Marvel’s Spider-Man Miles Morales with ray tracing activated. The performance of the 36-thread variant is similar to the Xeon E5-2670 v3 chip, while the pure 18-core variant manages to put up a fight against the overclocked Ryzen 5 3600 and the stock Core i3-12100F.

Ray Tracing – OFF

In less heavy mode, both 2699 v3 variants slightly lose ground, due to which they can only compete with stock variants of modern CPUs.

Starfield

Starfield is a new game from the revered guru Todd Howard. The project is based on a modified version of Creation Engine 2, which was used in Fallout 4. However, the key word here is “refined”. Bethesda studio did not lie in their statements. Of course, the current version of Creation Engine can be criticized for many things (disgusting work with drives, poor optimization for current video cards and so on), but the developers have seriously optimized their code for multithreaded processors.

Our test scene is located in the center of the city of New Atlantis. For FPS measurement we use the section from the landing pad to the central square.

The updated Creation Engine 2 reacted extremely favorably to the Xeon E5-2699 v3. Even the 36-threaded version of the chip shows itself extremely well and has no trouble dealing with E5-2670 v3. The 18-core version, in its turn, actually doesn’t feel competition with the overclocked i5-10400F, and is located right after Core i5-12400F and Core i7-11700K.

The only thing I would like to mention is that we had very high expectations when testing E5-2699 v3 in Starfield. Taking into account the peculiarity of Creation Engine 2 (excellent optimization for multi-core processors), we thought that this game would be able to unlock the potential of 36 threads. However, during the experiments it turned out that Starfield is able to load up to 24 threads. However, there is nothing to blame the developers for, they worked hard on their product.

The Witcher 3 Next-Gen Update 4.04

The Witcher 3 updated version moved to an improved version of the REDengine 3 engine with support for Ray Tracing, but the project was not ready for such complex changes and as a result, the game became much more demanding than the same Cyberpunk 2077.

Next-Gen Update is extremely CPU-dependent, and it cannot adequately utilize powerful CPUs with eight or more cores. The Witcher 3 runs maximally efficiently on 6-core, 12-thread chips. In addition, REDengine 3 responds very well to high-frequency RAM.

Ray Tracing – ON

Next-Gen Update of the third Witcher with ray tracing runs on the 36-threaded version of Xeon E5-2699 v3 just disgusting. The minimum recorded FPS is at a depressing 23 frames, which is the performance level of an old Xeon E5-2690 and a stock Ryzen 5 1600X. The game literally microfreezes, which makes it hard to call such gameplay even minimally acceptable.

But fortunately this is not a verdict. Disabling Hyper-Threading contributes to a significant increase in both minimum and average FPS. The 18-core catches up and even slightly overtakes the stock Core i5-10400F and Ryzen 5 3600 without any problems.

Ray Tracing – OFF

Without the notorious Ray Tracing, the performance of the 36-threaded version evens out somewhat, and its results are now comparable to those of the stock i5-10400F, while the 18-core variant gets to i3-12100F.

Watch Dogs Legion

The third part of the Watch Dogs franchise is most likely based on the improved Disrupt 2 engine. Since the release of the first installment, the projects of this series of games worked quite well with multi-core CPUs. Legion is no exception. The game is able to utilize more than 20 threads without any problems, and also reacts positively to high-frequency memory.

Ray Tracing – ON

Legion works relatively well with 24 threads, but 36 is too much for it. Nevertheless, even in this case both variants of 2699 v3 do quite well with the game when ray tracing is activated.

Ray Tracing – OFF

Without Ray Tracing, the FPS increases quite significantly, but the balance of power does not change.

RPCS3, Red Dead Redemption

RPCS3 is the most popular PlayStation 3 console emulator. The software perfectly parallelizes shader compilation on a huge number of threads, but when it comes to the game emulation itself, the quality of load distribution drops significantly. Of course, everything depends on the specific game, but according to our observations actual CPUs with 6-8 cores and high frequency perform best.

Besides, you can get a noticeable performance boost from such instruction sets as AVX-512 and TSX. However, in case of the latter, sometimes some projects are unstable.

Emulator settings for Red Dead Redemption: SPU block size – Mega, ZCULL accuracy – Relaxed, Write color buffers – On, Sleep timers accuracy – As Host, RSX FIFO accuracy – Atomic, resolution 1080p.

In the PlayStation 3 console emulator, the additional cores/threads had almost no effect on the performance of the Xeon E5-2699 v3 relative to its younger brothers. The only thing that made the 36-threaded chip still managed to overpower them was the frequency.

However, it should be noted that loading games and shader compilation on Xeon E5-2699 v3 is much faster than on conditional E5-2670 v3.

Game test analysis

Just like in the article about Xeon E5-2670 v3, we have to state that when using Hyper-Threading technology such a multi-threaded processor is not able to demonstrate its full potential of gaming performance.

If you are considering Xeon E5-2699 v3 as a budget gaming CPU, it is mandatory to deactivate Hyper-Threading in your motherboard BIOS. This way, it will increase the final performance by 15 to 25%, and as a result, it will provide you with high, but more importantly, stable frame rates.

Conclusions

If you are a happy owner of the LGA 2011 v3 platform and a low-end CPU, or you are looking at it, it seems that the time to buy a high-end Xeon E5-2699 v3 has come. The price of such a powerful product is at the same level as the entry-level chips like Core i3 or Ryzen 3, while the performance sometimes reaches the heights of much more expensive Core i5-12400F and Core i7-11700K.

Without a doubt, the gaming performance of the E5-2699 v3 often falls short of the same overclocked Core i5-10400F, or Ryzen 5 3600, but disabling HT actually corrects this oversight.

In fact, everything is quite simple with E5-2699 v3: if you prioritize heavy software like archivers or 3D modeling packages, it is better to leave Hyper-Threading activated. If you are a gamer – deactivate it.

Xeon E5-2699 v3 is an excellent processor and evokes only positive emotions. But you should keep in mind the following: such a CPU needs a board with a decent VRM, which is able to provide the 145-watt chip with stable power supply and high-quality cooling.