AMD Threadripper 2950X Performance Review
- Thread starter SKYMTL
- Start date
Gaming Performance (Battlefield 1 / COD: IW)
Battlefield 1 will likely become known as one of the most popular multiplayer games around but it also happens to be one of the best looking titles around. It also happens to be extremely well optimized with even the lowest end cards having the ability to run at high detail levels.
In this benchmark we use a runthough of The Runner level after the dreadnought barrage is complete and you need to storm the beach. This area includes all of the game’s hallmarks in one condensed area with fire, explosions, debris and numerous other elements layered over one another for some spectacular visual effects.
The latest iteration in the COD series may not drag out niceties like DX12 or particularly unique playing styles but it nonetheless is a great looking game that is quite popular.
This benchmark takes place during the campaign’s Operation Port Armor wherein we run through a sequence combining various indoor and outdoor elements along with some combat.
Battlefield 1
Battlefield 1 will likely become known as one of the most popular multiplayer games around but it also happens to be one of the best looking titles around. It also happens to be extremely well optimized with even the lowest end cards having the ability to run at high detail levels.
In this benchmark we use a runthough of The Runner level after the dreadnought barrage is complete and you need to storm the beach. This area includes all of the game’s hallmarks in one condensed area with fire, explosions, debris and numerous other elements layered over one another for some spectacular visual effects.
Call of Duty: Infinite Warfare
The latest iteration in the COD series may not drag out niceties like DX12 or particularly unique playing styles but it nonetheless is a great looking game that is quite popular.
This benchmark takes place during the campaign’s Operation Port Armor wherein we run through a sequence combining various indoor and outdoor elements along with some combat.
Gaming Performance (Deus Ex / DOOM)
Deus Ex titles have historically combined excellent storytelling elements with action-forward gameplay and Mankind Divided is no difference. This run-through uses the streets and a few sewers of the main hub city Prague along with a short action sequence involving gunplay and grenades.
Not many people saw a new Doom as a possible Game of the Year contender but that’s exactly what it has become. Not only is it one of the most intense games currently around but it looks great and is highly optimized. In this run-through we use Mission 6: Into the Fire since it features relatively predictable enemy spawn points and a combination of open air and interior gameplay.
Deus Ex – Mankind Divided
Deus Ex titles have historically combined excellent storytelling elements with action-forward gameplay and Mankind Divided is no difference. This run-through uses the streets and a few sewers of the main hub city Prague along with a short action sequence involving gunplay and grenades.
Doom
Not many people saw a new Doom as a possible Game of the Year contender but that’s exactly what it has become. Not only is it one of the most intense games currently around but it looks great and is highly optimized. In this run-through we use Mission 6: Into the Fire since it features relatively predictable enemy spawn points and a combination of open air and interior gameplay.
Gaming Performance (GTA V / Overwatch)
In GTA V we take a simple approach to benchmarking: the in-game benchmark tool is used. However, due to the randomness within the game itself, only the last sequence is actually used since it best represents gameplay mechanics.
Overwatch happens to be one of the most popular games around right now and while it isn’t particularly stressful upon a system’s resources, its Epic setting can provide a decent workout for all but the highest end GPUs. In order to eliminate as much variability as possible, for this benchmark we use a simple “offline” Bot Match so performance isn’t affected by outside factors like ping times and network latency.
Grand Theft Auto V
In GTA V we take a simple approach to benchmarking: the in-game benchmark tool is used. However, due to the randomness within the game itself, only the last sequence is actually used since it best represents gameplay mechanics.
Overwatch
Overwatch happens to be one of the most popular games around right now and while it isn’t particularly stressful upon a system’s resources, its Epic setting can provide a decent workout for all but the highest end GPUs. In order to eliminate as much variability as possible, for this benchmark we use a simple “offline” Bot Match so performance isn’t affected by outside factors like ping times and network latency.
System Power Consumption
I don’t typically dedicate a whole page to power consumption but there’s a pretty substantial story lurking behind the numbers you see below and how they directly relate to TDP claims from both Intel and AMD. Without getting too technical, the way these two companies go about measuring TDP is fundamentally different from one another.
What you need to know is that TDP values are a universally poor way to determine actual power consumption for end users since they are simply thermal design guidelines that are given to system integrators. As I say in every review, TDP is not actual power consumption so don’t take it as such.
As both Intel and AMD recommend, the best way to measure true power deltas between processors is via a simple (yet calibrated) power meter plugged into the wall outlet. That’s exactly what we do but add in a controlled 120V power input to eliminate voltage irregularities from impacting the results.
System Power Consumption
I don’t typically dedicate a whole page to power consumption but there’s a pretty substantial story lurking behind the numbers you see below and how they directly relate to TDP claims from both Intel and AMD. Without getting too technical, the way these two companies go about measuring TDP is fundamentally different from one another.
What you need to know is that TDP values are a universally poor way to determine actual power consumption for end users since they are simply thermal design guidelines that are given to system integrators. As I say in every review, TDP is not actual power consumption so don’t take it as such.
As both Intel and AMD recommend, the best way to measure true power deltas between processors is via a simple (yet calibrated) power meter plugged into the wall outlet. That’s exactly what we do but add in a controlled 120V power input to eliminate voltage irregularities from impacting the results.
Precision Boost Overdrive - A NEW Way To Overclock?
The road to a stable overclock is often paved with the tears of failed attempts, hours of frustration and the loss of some key features. A great example of this can be seen with Intel’s newest Skylake-X HEDT parts which have the capability to reach impressive frequencies through overclocking but to achieve the highest possible clock speed, many essential power saving and efficiency-focused elements have to be disabled. The result is often extreme current requirements and heat production. On AMD Theradripper systems, getting those higher speeds means disabling SenseMi elements like Precision Boost.
While AMD still allows for manual overclocking of Threadripper 2 processors, they’ve instituted a “middle ground” between manual overclocking and stock settings. Called Precision Boost Overdrive (PBO), it still allows for higher performance but does so without sacrificing Precision Boost 2 and XFR 2.
PBO basically takes the limits dictated by Precision Boost 2 for package power, thermal design current and electrical design current and increases them by a preset amount. This means the Precision Boost curve you saw earlier in this review moves upwards in a dynamic fashion so all cores can achieve higher clocks. The result can be up to a 13% jump in multi threaded throughput of more depending on how far you push this setting.
All of this sounds great on paper but before I get too much further into this section, there needs to be a mention about warranty. Since PBO effectively allows the CPU to run beyond its design parameters, engaging it will void your warranty. So take everything that follows with a word of warning: even though Precision Boost Overdrive is easy to enable, only you can determine if the benefits are really worthwhile.
In order to enable PBO, simply open up AMD’s Ryzen Master software suite and switch the Control Mode button from Auto to Precision Boost Overdrive. You’ll be prompted to accept a voided warranty (yes, AMD supposedly has ways to log these things within the processor) and then if its your wish, higher performance should be right at your fingertips.
After that, you’ll need to modify the three values directly below the Auto / PBO / Manual switch. These allow you to increase the following values (taken from AMD’s review guide):
Package Power Tracking (“PPT”):The PPT threshold is the allowed socket power consumption permitted across the voltage rails supplying the socket. Applications with high thread counts, and “heavy” threads, can encounter PPT limits that can be alleviated with a raised PPT limit.
Thermal Design Current (“TDC”):The maximum current that can be delivered by a specific motherboard’s voltage regulator configuration after warming to a steady-state condition through continuous operation.
Electrical Design Current (“EDC”):The maximum current that can be delivered by a specific motherboard’s voltage regulator configuration in a peak (“spike”) condition for a short period of time.
Each of these values will need to be individually modified in order for Performance Boost Overdrive to further boost clock frequencies. Now according to AMD, each motherboard manufacturer will have upper limits in place which are detected by Ryzen Master so folks can’t go absolutely nuts with these settings. However, in my experience with the Zenith Extreme, there weren’t any evident caps in place.
The great thing about Ryzen Master is that it tells you exactly what’s going on at a given point. For example, I moved PPT and TDC to 220 each (defaults of 180W) while EDC was turned to 280. This represented a significant current increase and remained within the limits I figured I’d stay within. There was of course some more headroom but having the CPU package consume upwards of 220W when under full load was plenty for this demonstration.
What became evident is that the Power Package Tracking will quickly become a clock speed bottleneck long before TDC or EDC will. As you can see in the image above, at full load it became the limiting factor in overall performance.
But does this all add up to improved benchmark results? Well lets find out with a few tests run on our 2950X…..the results are certainly interesting.
Let’s start off with a simple all-tread load scenario which in this case is AIDA64’s Stress Test. Generally PBO seems to have allowed us to increase overall the all core boost state frequencies by about 5% or 150MHz according to CPU-Z. However, what you can’t see by these screenshots is the core speed rapidly fluctuates upwards and downwards around the 37.5x multiplier without PBO enabled and 39.25x with it on. Essentially this led to an average frequency speedup of about 10% over course of the 10-minute test.
Single and lightly threaded workloads also see an improvement but it is closer to 4% on average. This is likely because Precision Boost 2 was already operating close to its maximum efficiency when in stock mode due to the power, thermal and voltage headroom lower load scenarios impart upon the CPU.
Across our synthetic tests, a pattern emerged as well. The multi threaded workloads of Cinebench and WPrime allowed the 2950X to deliver noticeably improved results with PBO enabled. Meanwhile, the single threaded scenarios in each application didn’t show any benefit at all.
The same can be said about some real world programs. Premier which doesn’t tend to overly stress all 16 cores -part of its workloads are shunted towards the GPU through CUDA- sees very little improvement while Handbrake and Blender do show some minor speedups.
Finally in gaming where lightly threaded scenarios rule, PBO offers nothing in the way of additional performance. In a way this points towards how well AMD has been able to maximize their clock speeds if Threadripper is operating below its inbred limits. However, it also means that Precision Boost Overdrive will have limited uses for people who aren’t working with higher level applications that stress 8 or more cores.
Precision Boost Overdrive - Details & Benchmarks
The road to a stable overclock is often paved with the tears of failed attempts, hours of frustration and the loss of some key features. A great example of this can be seen with Intel’s newest Skylake-X HEDT parts which have the capability to reach impressive frequencies through overclocking but to achieve the highest possible clock speed, many essential power saving and efficiency-focused elements have to be disabled. The result is often extreme current requirements and heat production. On AMD Theradripper systems, getting those higher speeds means disabling SenseMi elements like Precision Boost.
While AMD still allows for manual overclocking of Threadripper 2 processors, they’ve instituted a “middle ground” between manual overclocking and stock settings. Called Precision Boost Overdrive (PBO), it still allows for higher performance but does so without sacrificing Precision Boost 2 and XFR 2.
PBO basically takes the limits dictated by Precision Boost 2 for package power, thermal design current and electrical design current and increases them by a preset amount. This means the Precision Boost curve you saw earlier in this review moves upwards in a dynamic fashion so all cores can achieve higher clocks. The result can be up to a 13% jump in multi threaded throughput of more depending on how far you push this setting.
All of this sounds great on paper but before I get too much further into this section, there needs to be a mention about warranty. Since PBO effectively allows the CPU to run beyond its design parameters, engaging it will void your warranty. So take everything that follows with a word of warning: even though Precision Boost Overdrive is easy to enable, only you can determine if the benefits are really worthwhile.
In order to enable PBO, simply open up AMD’s Ryzen Master software suite and switch the Control Mode button from Auto to Precision Boost Overdrive. You’ll be prompted to accept a voided warranty (yes, AMD supposedly has ways to log these things within the processor) and then if its your wish, higher performance should be right at your fingertips.
After that, you’ll need to modify the three values directly below the Auto / PBO / Manual switch. These allow you to increase the following values (taken from AMD’s review guide):
Package Power Tracking (“PPT”):The PPT threshold is the allowed socket power consumption permitted across the voltage rails supplying the socket. Applications with high thread counts, and “heavy” threads, can encounter PPT limits that can be alleviated with a raised PPT limit.
Thermal Design Current (“TDC”):The maximum current that can be delivered by a specific motherboard’s voltage regulator configuration after warming to a steady-state condition through continuous operation.
Electrical Design Current (“EDC”):The maximum current that can be delivered by a specific motherboard’s voltage regulator configuration in a peak (“spike”) condition for a short period of time.
Each of these values will need to be individually modified in order for Performance Boost Overdrive to further boost clock frequencies. Now according to AMD, each motherboard manufacturer will have upper limits in place which are detected by Ryzen Master so folks can’t go absolutely nuts with these settings. However, in my experience with the Zenith Extreme, there weren’t any evident caps in place.
The great thing about Ryzen Master is that it tells you exactly what’s going on at a given point. For example, I moved PPT and TDC to 220 each (defaults of 180W) while EDC was turned to 280. This represented a significant current increase and remained within the limits I figured I’d stay within. There was of course some more headroom but having the CPU package consume upwards of 220W when under full load was plenty for this demonstration.
What became evident is that the Power Package Tracking will quickly become a clock speed bottleneck long before TDC or EDC will. As you can see in the image above, at full load it became the limiting factor in overall performance.
But does this all add up to improved benchmark results? Well lets find out with a few tests run on our 2950X…..the results are certainly interesting.
Let’s start off with a simple all-tread load scenario which in this case is AIDA64’s Stress Test. Generally PBO seems to have allowed us to increase overall the all core boost state frequencies by about 5% or 150MHz according to CPU-Z. However, what you can’t see by these screenshots is the core speed rapidly fluctuates upwards and downwards around the 37.5x multiplier without PBO enabled and 39.25x with it on. Essentially this led to an average frequency speedup of about 10% over course of the 10-minute test.
Single and lightly threaded workloads also see an improvement but it is closer to 4% on average. This is likely because Precision Boost 2 was already operating close to its maximum efficiency when in stock mode due to the power, thermal and voltage headroom lower load scenarios impart upon the CPU.
Across our synthetic tests, a pattern emerged as well. The multi threaded workloads of Cinebench and WPrime allowed the 2950X to deliver noticeably improved results with PBO enabled. Meanwhile, the single threaded scenarios in each application didn’t show any benefit at all.
The same can be said about some real world programs. Premier which doesn’t tend to overly stress all 16 cores -part of its workloads are shunted towards the GPU through CUDA- sees very little improvement while Handbrake and Blender do show some minor speedups.
Finally in gaming where lightly threaded scenarios rule, PBO offers nothing in the way of additional performance. In a way this points towards how well AMD has been able to maximize their clock speeds if Threadripper is operating below its inbred limits. However, it also means that Precision Boost Overdrive will have limited uses for people who aren’t working with higher level applications that stress 8 or more cores.
Manual Overclocking Results
Manual overclocking on Ryzen-based systems has been a game a hit or miss but the Zen+ architecture is supposed to improve things a bit. For the most part, I really didn’t see all that much in the way of additional overclocking headroom with the Threadripper 2950X. Back when the 1950X launched I was able to push that chip to about 4GHz on all cores without any fuss but it ended up being hard-capped to that speed. Considering the Base Clock was 3.4GHz, 600MHz wasn’t anything to complain about though.
The 2950X behaves much the same way as that first generation CPU but there’s a bit more headroom vis a vie its 100MHz higher base clock. In this case I was able to boost the all-core frequency to a constant 4.2GHz on air cooling with my trusty Noctua NH-U14S. That was paired up with a memory speed of 3200MHz at relatively tight timings.
To achieve that, it was simply a matter of setting VCore to 1.3V, SOC voltage to 1.11V and the memory modules to their rated 1.35V. Note that Ryzen systems tend to under-volt memory by a pretty substantial margin so make sure your overclocking process includes changing the RAM voltage in the motherboard BIOS.
As is usual for Ryzen and Threadripper-based systems, I simply received hard-locks whenever the system was under load at any frequency above 4.2GHz. 4.3GHz booted into Windows without any issue but stress testing brought about an immediate crash regardless of whether the multiplier or BCLK was used. Some additional investigation is needed –particularly within AMD’s Ryzen Master software- but for the purposes of this review, I’m going to just move on to benchmarks for now.
Starting off with the ubiquitous Cinebench program, there’s obviously a good looking increase in benchmark scores for the multi core test, above even those posted by Precision Boost Overdrive. On the other hand, since the 4.2GHz overclock doesn’t overcome AMD’s stated maximum Boost Clock of 4.4GHz, the single core score is actually lower than stock. That bodes quite poorly for gaming as well.
Premier Pro sees a nice speedup as well and while it may not look like much on paper, if you extrapolate that 5 seconds of time savings over a longer project (this one is just 60 seconds), it could have a substantial impact.
With both Skylake-X and Threadripper or any modern architecture for that matter, if you can’t raise an all-core overclock above a CPU’s Boost frequency, you’ll likely receive diminishing returns. Particularly if the game in question doesn’t take full advantage of multiple threads. This is somewhat the case here since Battlefield does distribute CPU load across 4-6 threads but that’s not enough for the 4.2GHz overclock to realize any net benefits.
Manual Overclocking Results
Manual overclocking on Ryzen-based systems has been a game a hit or miss but the Zen+ architecture is supposed to improve things a bit. For the most part, I really didn’t see all that much in the way of additional overclocking headroom with the Threadripper 2950X. Back when the 1950X launched I was able to push that chip to about 4GHz on all cores without any fuss but it ended up being hard-capped to that speed. Considering the Base Clock was 3.4GHz, 600MHz wasn’t anything to complain about though.
The 2950X behaves much the same way as that first generation CPU but there’s a bit more headroom vis a vie its 100MHz higher base clock. In this case I was able to boost the all-core frequency to a constant 4.2GHz on air cooling with my trusty Noctua NH-U14S. That was paired up with a memory speed of 3200MHz at relatively tight timings.
To achieve that, it was simply a matter of setting VCore to 1.3V, SOC voltage to 1.11V and the memory modules to their rated 1.35V. Note that Ryzen systems tend to under-volt memory by a pretty substantial margin so make sure your overclocking process includes changing the RAM voltage in the motherboard BIOS.
As is usual for Ryzen and Threadripper-based systems, I simply received hard-locks whenever the system was under load at any frequency above 4.2GHz. 4.3GHz booted into Windows without any issue but stress testing brought about an immediate crash regardless of whether the multiplier or BCLK was used. Some additional investigation is needed –particularly within AMD’s Ryzen Master software- but for the purposes of this review, I’m going to just move on to benchmarks for now.
Starting off with the ubiquitous Cinebench program, there’s obviously a good looking increase in benchmark scores for the multi core test, above even those posted by Precision Boost Overdrive. On the other hand, since the 4.2GHz overclock doesn’t overcome AMD’s stated maximum Boost Clock of 4.4GHz, the single core score is actually lower than stock. That bodes quite poorly for gaming as well.
Premier Pro sees a nice speedup as well and while it may not look like much on paper, if you extrapolate that 5 seconds of time savings over a longer project (this one is just 60 seconds), it could have a substantial impact.
With both Skylake-X and Threadripper or any modern architecture for that matter, if you can’t raise an all-core overclock above a CPU’s Boost frequency, you’ll likely receive diminishing returns. Particularly if the game in question doesn’t take full advantage of multiple threads. This is somewhat the case here since Battlefield does distribute CPU load across 4-6 threads but that’s not enough for the 4.2GHz overclock to realize any net benefits.
Last edited:
Conclusion; A Timely Evolution
About a year ago during the launch of Threadripper I was in the process of being impressed yet again by how far AMD had come with the Zen architecture. Before it, there were unfulfilled expectations, a successive line of disappointing launches and a real lack of direct competition for Intel. Whereas AMD was in a struggle for their own survival before Ryzen and Threadripper, just a year later it seems Intel are the ones now struggling to find a solution to the rising Zen-based tide.
The introduction of Zen+ first on Ryzen 2 and now Threadripper 2 may not be the bolt out of the blue which defined those first generation CPUs but it is yet another strong showing. In about a year, AMD has shown they can take an already-strong chip design and improve it in some key ways, resulting in a net positive impact on performance and efficiency. While I can’t talk to the 2990WX’s positioning (we’re hoping to receive one of those in the coming weeks, aligning with a complete update to our testing suite) these steps have benefitted the 2950X in a pretty substantial way.
One paper at least, the Zen+ microarchitecture has allowed the 2950X to run faster and in our tests at least, also more efficiently. That in itself is quite impressive since this CPUs benchmark standing improved over its predecessor but overall power consumption decreased a bit. Naturally, the 7W separating current and previous generation x950 processors is well within the margin of error but its still noteworthy in the grand scheme of things.
Speaking of performance, I ended up seeing improvements over the 1950X that ranged anywhere from 4% up to 12% which happens to align perfectly with AMD’s own internal numbers. While this won’t lead to current Threadripper owners rushing to upgrade, it does make this version of Threadripper all that more competitive against Skylake-X. And that folks is what this revolutionary approach is meant to accomplish; provide another stepping stone closer to absolute parity with or leadership over Intel’s HEDT lineup.
While the 2950X does improve on the 1950X's single threaded and lightly threaded performance, this is still one area where Intel reigns supreme. This shows up in particularly stark contrast within games. However, its also important to remember that as resolution increases so too do the demands upon the GPU, pushing the CPUs importance further into the background. Indeed, we've seen instances where an old Sandy Bridge processor easily stood up against a Coffee Lake CPU when higher resolutions were used.
AMD will be the first to say that if you are first and foremost a gamer, look towards the less expensive Ryzen CPUs and X370 platform. Along with Coffee Lake CPUs, they offer the ultimate value for in-game performance whereas Threadripper and Skylake-X do offer great performance but their costs are stratospheric by comparison. HEDT platforms like X299 and X399 are all about their ability to offer parallel operations WHILE gaming without missing a beat. Parallelized rendering, gameplay and streaming are their forte.
Perhaps a more direct comparison is in order too since the 2950X's natural competitor on the Intel side is the 16-core, 32-thread 7960X. Its an interesting slugfest between these two beastly chips but in the end, the AMD chip lost more than it won, particularly in real world applications. However, the $800(!) premium for that Skylake-X chip is a bridge many won't want to (and shouldn't) cross.
Intel’s issue remains one of pricing, particularly when it comes to the high end desktop market. While they do win in some key metrics, many took their lack of response to Threadripper’s higher value quotient as both a sign of confidence and arrogance. And now with the 2950X and eventual 2920X being introduced at even lower price points while offering higher performance, one has to wonder if we’ll finally see some movement. Make no mistake about it, Skylake-X remains a compelling solution but not at the i9 lineup’s current price.
AMD also happens to be (still!) winning the platform war with X399 continuing to provide a solution that has a long list of enviable, appealing characteristics for creators and enthusiasts alike. AMD knew damn well they didn’t need to change anything in this respect since X299 has been looking long in the tooth for the better part of 18 months now. However the inclusion of StoreMI could be a godsend, particularly for those who want to take advantage of the large capacity quad channel memory kits roaming the market.
The last thing I wanted to mention is overclocking. I actually didn’t spend all that much time on it but in this case I was able to get the 2950X bustling along at about 4.2GHz on all cores by using the silent but extremely potent Noctua NH-U14S. Temperatures remained well below 80C throughout torture testing. The end result was a marked improvement in every benchmark category but the somewhat limited amount of headroom still points towards AMD doing everything they can to maximize out-of-box frequencies.
On the flip side of that equation, AMD deserves credit for Precision Boost Overdrive. It is an intuitive albeit warranty-nuking feature that can easily grant a boost for everyday applications while retailing key efficiency-driving features native to Zen+. It is however somewhat limited in scope since I saw very few benefits for lightly threaded applications. If you are constantly coming close to maximizing the number of threads on these new CPUs, they its certainly something you should look at using instead of going straight to multiplier and voltage overclocking.
A roadmap can’t turn on a dime but it feels like Intel is slowly but surely putting their ducks in a line for yet another thrust directly at AMD’s heart. However, with the likes of the Ryzen and Threadripper 2000-series, it also seems that AMD is now extremely well positioned to take pretty much anything Intel can throw at them. The Threadripper 2950X maybe be nothing more than an evolutionary step forward but it proves that after a successive string of launches, AMD is doing everything but sitting on their laurels. As such, this is one incremental update that feels like so much more.
Conclusion; A Timely Evolution
About a year ago during the launch of Threadripper I was in the process of being impressed yet again by how far AMD had come with the Zen architecture. Before it, there were unfulfilled expectations, a successive line of disappointing launches and a real lack of direct competition for Intel. Whereas AMD was in a struggle for their own survival before Ryzen and Threadripper, just a year later it seems Intel are the ones now struggling to find a solution to the rising Zen-based tide.
The introduction of Zen+ first on Ryzen 2 and now Threadripper 2 may not be the bolt out of the blue which defined those first generation CPUs but it is yet another strong showing. In about a year, AMD has shown they can take an already-strong chip design and improve it in some key ways, resulting in a net positive impact on performance and efficiency. While I can’t talk to the 2990WX’s positioning (we’re hoping to receive one of those in the coming weeks, aligning with a complete update to our testing suite) these steps have benefitted the 2950X in a pretty substantial way.
One paper at least, the Zen+ microarchitecture has allowed the 2950X to run faster and in our tests at least, also more efficiently. That in itself is quite impressive since this CPUs benchmark standing improved over its predecessor but overall power consumption decreased a bit. Naturally, the 7W separating current and previous generation x950 processors is well within the margin of error but its still noteworthy in the grand scheme of things.
Speaking of performance, I ended up seeing improvements over the 1950X that ranged anywhere from 4% up to 12% which happens to align perfectly with AMD’s own internal numbers. While this won’t lead to current Threadripper owners rushing to upgrade, it does make this version of Threadripper all that more competitive against Skylake-X. And that folks is what this revolutionary approach is meant to accomplish; provide another stepping stone closer to absolute parity with or leadership over Intel’s HEDT lineup.
While the 2950X does improve on the 1950X's single threaded and lightly threaded performance, this is still one area where Intel reigns supreme. This shows up in particularly stark contrast within games. However, its also important to remember that as resolution increases so too do the demands upon the GPU, pushing the CPUs importance further into the background. Indeed, we've seen instances where an old Sandy Bridge processor easily stood up against a Coffee Lake CPU when higher resolutions were used.
AMD will be the first to say that if you are first and foremost a gamer, look towards the less expensive Ryzen CPUs and X370 platform. Along with Coffee Lake CPUs, they offer the ultimate value for in-game performance whereas Threadripper and Skylake-X do offer great performance but their costs are stratospheric by comparison. HEDT platforms like X299 and X399 are all about their ability to offer parallel operations WHILE gaming without missing a beat. Parallelized rendering, gameplay and streaming are their forte.
Perhaps a more direct comparison is in order too since the 2950X's natural competitor on the Intel side is the 16-core, 32-thread 7960X. Its an interesting slugfest between these two beastly chips but in the end, the AMD chip lost more than it won, particularly in real world applications. However, the $800(!) premium for that Skylake-X chip is a bridge many won't want to (and shouldn't) cross.
Intel’s issue remains one of pricing, particularly when it comes to the high end desktop market. While they do win in some key metrics, many took their lack of response to Threadripper’s higher value quotient as both a sign of confidence and arrogance. And now with the 2950X and eventual 2920X being introduced at even lower price points while offering higher performance, one has to wonder if we’ll finally see some movement. Make no mistake about it, Skylake-X remains a compelling solution but not at the i9 lineup’s current price.
AMD also happens to be (still!) winning the platform war with X399 continuing to provide a solution that has a long list of enviable, appealing characteristics for creators and enthusiasts alike. AMD knew damn well they didn’t need to change anything in this respect since X299 has been looking long in the tooth for the better part of 18 months now. However the inclusion of StoreMI could be a godsend, particularly for those who want to take advantage of the large capacity quad channel memory kits roaming the market.
The last thing I wanted to mention is overclocking. I actually didn’t spend all that much time on it but in this case I was able to get the 2950X bustling along at about 4.2GHz on all cores by using the silent but extremely potent Noctua NH-U14S. Temperatures remained well below 80C throughout torture testing. The end result was a marked improvement in every benchmark category but the somewhat limited amount of headroom still points towards AMD doing everything they can to maximize out-of-box frequencies.
On the flip side of that equation, AMD deserves credit for Precision Boost Overdrive. It is an intuitive albeit warranty-nuking feature that can easily grant a boost for everyday applications while retailing key efficiency-driving features native to Zen+. It is however somewhat limited in scope since I saw very few benefits for lightly threaded applications. If you are constantly coming close to maximizing the number of threads on these new CPUs, they its certainly something you should look at using instead of going straight to multiplier and voltage overclocking.
A roadmap can’t turn on a dime but it feels like Intel is slowly but surely putting their ducks in a line for yet another thrust directly at AMD’s heart. However, with the likes of the Ryzen and Threadripper 2000-series, it also seems that AMD is now extremely well positioned to take pretty much anything Intel can throw at them. The Threadripper 2950X maybe be nothing more than an evolutionary step forward but it proves that after a successive string of launches, AMD is doing everything but sitting on their laurels. As such, this is one incremental update that feels like so much more.
Last edited by a moderator:
