AMD Ryzen 7 1800X CPU Review

[palladin9479]

It's not an external interface, the fabric is where the system memory bus is located, think old school FSB.

The memory controller is likely only one stop on the fabric and unless AMD was incredibly dumb, the fabric should be some form of transactional bus with the memory controller having a command buffer to decouple memory latency from the internal bus/fabric/whatever. Using the memory bus for cache snoops and synchronization between CCX would also nuke memory performance - that's what killed shared FSB CPU architectures' performance and forced CPU designers to go with point-to-point links between CPUs in multi-socket configurations ~10 years ago.

In order to get to main memory it needs to go through the fabric and for some reason AMD seems to run them both on the same clock generator. For data to go from one CCX to another it must traverse the fabric, for that same data to go out to a PCIe device it against must traverse this fabric.

https://www.techpowerup.com/231585/amd-ryzen-infinity-fabric-ticks-at-memory-speed

This is why installing higher speed DDR4 memory gives such a large boost in performance, it's directly raising the clock speed of the fabric interconnect between the two CCX units. The two 8MB segments of L3 are kept separate but coherency is synchronized across the fabric.

Then you agree the lower latency RAM is going to have a significant increase in Infinity Fabric performance. Everyone has been testing with 2800-3200MHz and say 16 or 15 CAS. How about 2400 at 10 CAS that would be much faster than the 3200 at 16 or 15 CAS.

....

Latency has nothing to do with that ....

And ... latency is a component of clock speed not a separate number. Latency is just the number of clock ticks between a specific set of commands, clock speed is the number of ticks per second. Thus higher clock speeds with higher latency is the exact same as lower clock speeds with lower latency. Finally the infinity fabric has jack shit to do with the column address select timings on a stick of DRAM sitting over in a slot on the motherboard, it just use's the same clock generator as the memory controller. DDR4-2133 has a 1066Mhz clock speed that is being driven by an external clock generator. That clock line will then also be used by the infinity fabric for it's transactions, so 1066Mhz. Installing DDR4-3000 memory would give a 1500Mhz clock speed to both the DDR4 memory and the infinity fabric, which would raise inter-CCX performance by 40%.

 

[goldstone77]

It's not an external interface, the fabric is where the system memory bus is located, think old school FSB.

The memory controller is likely only one stop on the fabric and unless AMD was incredibly dumb, the fabric should be some form of transactional bus with the memory controller having a command buffer to decouple memory latency from the internal bus/fabric/whatever. Using the memory bus for cache snoops and synchronization between CCX would also nuke memory performance - that's what killed shared FSB CPU architectures' performance and forced CPU designers to go with point-to-point links between CPUs in multi-socket configurations ~10 years ago.

In order to get to main memory it needs to go through the fabric and for some reason AMD seems to run them both on the same clock generator. For data to go from one CCX to another it must traverse the fabric, for that same data to go out to a PCIe device it against must traverse this fabric.

https://www.techpowerup.com/231585/amd-ryzen-infinity-fabric-ticks-at-memory-speed

This is why installing higher speed DDR4 memory gives such a large boost in performance, it's directly raising the clock speed of the fabric interconnect between the two CCX units. The two 8MB segments of L3 are kept separate but coherency is synchronized across the fabric.

Then you agree the lower latency RAM is going to have a significant increase in Infinity Fabric performance. Everyone has been testing with 2800-3200MHz and say 16 or 15 CAS. How about 2400 at 10 CAS that would be much faster than the 3200 at 16 or 15 CAS.

....

Latency has nothing to do with that ....

And ... latency is a component of clock speed not a separate number. Latency is just the number of clock ticks between a specific set of commands, clock speed is the number of ticks per second. Thus higher clock speeds with higher latency is the exact same as lower clock speeds with lower latency. Finally the infinity fabric has jack shit to do with the column address select timings on a stick of DRAM sitting over in a slot on the motherboard, it just use's the same clock generator as the memory controller. DDR4-2133 has a 1066Mhz clock speed that is being driven by an external clock generator. That clock line will then also be used by the infinity fabric for it's transactions, so 1066Mhz. Installing DDR4-3000 memory would give a 1500Mhz clock speed to both the DDR4 memory and the infinity fabric, which would raise inter-CCX performance by 40%.

Hold on now you just admitted it's a component of the speed. And I said 2400MHz RAM at CAS 10 is Faster than 3200MHz RAM at CAS 16 or 15. Do you disagree with that statement. Also, wouldn't it equate to better performance?
Also, check out the bench marks just manipulating timing. https://community.amd.com/message/2790033?et=notification.mention#comment-2790033

 

[Crashman]

Then you agree the lower latency RAM is going to have a significant increase in Infinity Fabric performance. Everyone has been testing with 2800-3200MHz and say 16 or 15 CAS. How about 2400 at 10 CAS that would be much faster than the 3200 at 16 or 15 CAS.

....

Latency has nothing to do with that ....

And ... latency is a component of clock speed not a separate number. Latency is just the number of clock ticks between a specific set of commands, clock speed is the number of ticks per second. Thus higher clock speeds with higher latency is the exact same as lower clock speeds with lower latency. Finally the infinity fabric has jack shit to do with the column address select timings on a stick of DRAM sitting over in a slot on the motherboard, it just use's the same clock generator as the memory controller. DDR4-2133 has a 1066Mhz clock speed that is being driven by an external clock generator. That clock line will then also be used by the infinity fabric for it's transactions, so 1066Mhz. Installing DDR4-3000 memory would give a 1500Mhz clock speed to both the DDR4 memory and the infinity fabric, which would raise inter-CCX performance by 40%.

Hold on now you just admitted it's a component of the speed. And I said 2400MHz RAM at CAS 10 is Faster than 3200MHz RAM at CAS 16 or 15. Do you disagree with that statement. Also, wouldn't it equate to better performance?
Also, check out the bench marks just manipulating timing. https://community.amd.com/message/2790033?et=notification.mention#comment-2790033

Wait, you got your DDR4-2400 to run CAS 10 straight timings? Congrats. With similar effort you could get DDR4-3200 to run at CAS 13, and then it would be both quicker and with more bandwidth than DDR4-2400 CAS 10. 2400/10 = 3200/13.3333

 

[goldstone77]

Then you agree the lower latency RAM is going to have a significant increase in Infinity Fabric performance. Everyone has been testing with 2800-3200MHz and say 16 or 15 CAS. How about 2400 at 10 CAS that would be much faster than the 3200 at 16 or 15 CAS.

....

Latency has nothing to do with that ....

And ... latency is a component of clock speed not a separate number. Latency is just the number of clock ticks between a specific set of commands, clock speed is the number of ticks per second. Thus higher clock speeds with higher latency is the exact same as lower clock speeds with lower latency. Finally the infinity fabric has jack shit to do with the column address select timings on a stick of DRAM sitting over in a slot on the motherboard, it just use's the same clock generator as the memory controller. DDR4-2133 has a 1066Mhz clock speed that is being driven by an external clock generator. That clock line will then also be used by the infinity fabric for it's transactions, so 1066Mhz. Installing DDR4-3000 memory would give a 1500Mhz clock speed to both the DDR4 memory and the infinity fabric, which would raise inter-CCX performance by 40%.

Hold on now you just admitted it's a component of the speed. And I said 2400MHz RAM at CAS 10 is Faster than 3200MHz RAM at CAS 16 or 15. Do you disagree with that statement. Also, wouldn't it equate to better performance?
Also, check out the bench marks just manipulating timing. https://community.amd.com/message/2790033?et=notification.mention#comment-2790033

Wait, you got your DDR4-2400 to run CAS 10 straight timings? Congrats. With similar effort you could get DDR4-3200 to run at CAS 13, and then it would be both quicker and with more bandwidth than DDR4-2400 CAS 10. 2400/10 = 3200/13.3333

If you can do it Crashman go for it! This RAM runs at that speed already I didn't have to do anything. And, my point was much better performance gains could be achieved with lower latency RAM vs. instability 3200-3600MHz RAM.
CORSAIR Dominator Platinum
DDR4 2400 (PC4 19200)
Timing 10-12-12-28
Cas Latency 10
Voltage 1.35V
https://www.newegg.com/Product/Product.aspx?Item=N82E16820233999

P.S. I'm just trying to have an honest conversation. I don't understand the hostility.

 

[palladin9479]

*Slams head on Desk*

Memory timings have absolutely nothing to do with Infinity Fabric, it just happens to use the same clock generator. Memory timings are just the waiting periods between commands in order to ensure a signal has been terminated before putting a second signal down a line. When you send a command signal down a wire, that signal takes time to go all the way and either be processed or be terminated. If you were to attempt to put another signal down that wire while a signal was already present it would cause the two signals to collide and produce an error. To prevent this from happening a set amount of time needs to pass before sending another signal and the computer measures that time in clock ticks.

DDR-2133 has a clock rate of 1066 for 0.938ns per clock cycle. DDR-2400 has 1200 clock rate 0.833ns per clock cycle, at 10 clock cycles your looking at 8.33ns wait in-between signals. DDR-3200 is 1600mhz clock for 0.625ns per clock cycle, at 15 your waiting 9.375ns at 16 is 10ns in between signals. Now after doing some research DDR4-3200 CL16 is very common while DDR4-2400 seems to hover around the 12~14 range. It's all a matter of how fast the individual DRAM chips can process the signal combined with how long that signal path is, DDR4 has a longer signal path then DDR3 did. FYI the general floor of memory latency for system DRAM was about 7ns. Stability with latencies is largely a function of quality of memory and how short you can keep the path to.

And none of that applies to the infinity fabric which merely use's the memory clock for reference.

 

[-Fran-]

I'm having a good laugh, lol.

Cheers!

 

[InvalidError]

Hold on now you just admitted it's a component of the speed. And I said 2400MHz RAM at CAS 10 is Faster than 3200MHz RAM at CAS 16 or 15. Do you disagree with that statement. Also, wouldn't it equate to better performance?

The clock frequency determines the amount of bandwidth available across the infinity fabric and between the memory controller and DIMMs. Latency determines how many cycles the memory controller needs to wait between commands to the DIMMs and has absolutely nothing to do with the infinity fabric, so latency should have next to no effect in cross-CCX performance tests. The only thing the fabric has in common with the memory is using the same clock source and that's why faster RAM frequency has such a direct impact on Ryzen's performance.

Looks like AMD may have underestimated Ryzen's fabric bandwidth requirement by ~30% and will need to either widen it or put it on a separate clock domain in a future revision.

 

[goldstone77]

*Slams head on Desk*

Memory timings have absolutely nothing to do with Infinity Fabric, it just happens to use the same clock generator. Memory timings are just the waiting periods between commands in order to ensure a signal has been terminated before putting a second signal down a line. When you send a command signal down a wire, that signal takes time to go all the way and either be processed or be terminated. If you were to attempt to put another signal down that wire while a signal was already present it would cause the two signals to collide and produce an error. To prevent this from happening a set amount of time needs to pass before sending another signal and the computer measures that time in clock ticks.

DDR-2133 has a clock rate of 1066 for 0.938ns per clock cycle. DDR-2400 has 1200 clock rate 0.833ns per clock cycle, at 10 clock cycles your looking at 8.33ns wait in-between signals. DDR-3200 is 1600mhz clock for 0.625ns per clock cycle, at 15 your waiting 9.375ns at 16 is 10ns in between signals. Now after doing some research DDR4-3200 CL16 is very common while DDR4-2400 seems to hover around the 12~14 range. It's all a matter of how fast the individual DRAM chips can process the signal combined with how long that signal path is, DDR4 has a longer signal path then DDR3 did. FYI the general floor of memory latency for system DRAM was about 7ns. Stability with latencies is largely a function of quality of memory and how short you can keep the path to.

And none of that applies to the infinity fabric which merely use's the memory clock for reference.

Okay, in your opinion which will increase performance more on Ryzen 3200MHz CAS 16 RAM or 2400MHz CAS 10 RAM?

 

[palladin9479]

Hold on now you just admitted it's a component of the speed. And I said 2400MHz RAM at CAS 10 is Faster than 3200MHz RAM at CAS 16 or 15. Do you disagree with that statement. Also, wouldn't it equate to better performance?

The clock frequency determines the amount of bandwidth available across the infinity fabric and between the memory controller and DIMMs. Latency determines how many cycles the memory controller needs to wait between commands to the DIMMs and has absolutely nothing to do with the infinity fabric, so latency should have next to no effect in cross-CCX performance tests. The only thing the fabric has in common with the memory is using the same clock source and that's why faster RAM frequency has such a direct impact on Ryzen's performance.

Looks like AMD may have underestimated Ryzen's fabric bandwidth requirement by ~30% and will need to either widen it or put it on a separate clock domain in a future revision.

I agree, it was probably a bad idea to use the same clock for both. Infinity fabric should have it's own independent clock multiplier from the memory channels, which would prevent this problem. Also it seems that current programs are crossing the CCX boundary far too often, likely a culprit of the OS schedulers not realizing the performance penalty associated with doing so.

Okay, in your opinion which will increase performance more on Ryzen 3200MHz CAS 16 RAM or 2400MHz CAS 10 RAM?

In current state then 3200 memory will produce better system performance, if only because the fabric would then be clocked at 1.5Ghz instead of the 1.2Ghz from the 2400 memory. Now if we were to separate the fabric clock from the memory clock then it gets down to a case of "it depends on application". Some performance profiles require wide data access, databases, web servers, simulations, and other wide tasks, but individual seeks don't need to be returned quickly. Other tasks don't use large memory access but instead have a ton of serial access patterns (games) where the next read is dependent on the value returned from the previous read / write and then access time in ns becomes the dominate factor. As crashman said CL10 2400 is the same as CL13 3200 so your not even comparing the same kind of memory.

 

[InvalidError]

Okay, in your opinion which will increase performance more on Ryzen 3200MHz CAS 16 RAM or 2400MHz CAS 10 RAM?

Since 3200MT/s gives Ryzen 33% more fabric bandwidth, it will likely win most of the time but I'm not seeing anyone running DDR4 at 2400-10.

 

[goldstone77]

Okay, in your opinion which will increase performance more on Ryzen 3200MHz CAS 16 RAM or 2400MHz CAS 10 RAM?

Since 3200MT/s gives Ryzen 33% more fabric bandwidth, it will likely win most of the time but I'm not seeing anyone running DDR4 at 2400-10.

Yeah, it's much more expensive. I would really like to see some testing done. To find the true benefits. Also, RAM that can operate at that low latency might be of the quality to be able to overclock and find a sweet spot between clock speed and latency for performance. RAM I know has very little effect on Intel performance, and people have been more concerned with with higher clock speed, and have forgot about lower timings to increase the responsiveness of a system. Testing would help shed some light on this. Hopefully, more people will realize RAM's impact on Ryzen and Benchmark it for all to see!

 

[InvalidError]

Yeah, it's much more expensive. I would really like to see some testing done. To find the true benefits.

You can get 3200-16 DDR4 for about the same price as common 2400-14, that isn't really a problem. Also, 3200-16 is lower latency than 2400-14: 10ns vs 11.6ns.

 

[goldstone77]

Yeah, it's much more expensive. I would really like to see some testing done. To find the true benefits.

You can get 3200-16 DDR4 for about the same price as common 2400-14, that isn't really a problem. Also, 3200-16 is lower latency than 2400-14: 10ns vs 11.6ns.

You can find 2400 CAS 10 for $169.99
CORSAIR Dominator Platinum 16GB (2 x 8GB) 288-Pin DDR4 SDRAM DDR4 2400
DDR4 2400 (PC4 19200)
Timing 10-12-12-28
Cas Latency 10
Voltage 1.35V
https://www.newegg.com/Product/Product.aspx?Item=N82E16820233999
The more common RAM I've seen people use in testing G.SKILL TridentZ Series 16GB (2 x 8GB) 288-Pin DDR4 SDRAM DDR4 3200
DDR4 3200 (PC4 25600)
Timing 16-18-18-38
Cas Latency 16
Voltage 1.35V
https://www.newegg.com/Product/Product.aspx?Item=N82E16820231929
Corsair is the lowest latency RAM I could find on the market. This is a huge difference in latency, which I would like to see someone benchmark.

 

[InvalidError]

Corsair is the lowest latency RAM I could find on the market. This is a huge difference in latency, which I would like to see someone benchmark.

Once you account for the clock frequency difference, 2400-10 is only 1.67ns (16.7%) faster than 3200-16's 10ns. On the other hand, the fabric is operating 33% faster with the 3200MT/s memory. In the end, you would likely be better off with the cheaper 3200-16 DDR4 in most cases.

Another problem with focusing on DRAM latency on Ryzen is that Ryzen has ~90ns of total latency between the CPU cores and DRAM versus ~70ns for Intel's CPUs. Shaving 1.67ns from 90ns isn't going to be as helpful as shaving the same 1.67ns from 70ns. AMD still has a lot of work to do on its memory controller and fabric.

 

[palladin9479]

Corsair is the lowest latency RAM I could find on the market. This is a huge difference in latency, which I would like to see someone benchmark.

Once you account for the clock frequency difference, 2400-10 is only 1.67ns (16.7%) faster than 3200-16's 10ns. On the other hand, the fabric is operating 33% faster with the 3200MT/s memory. In the end, you would likely be better off with the cheaper 3200-16 DDR4 in most cases.

Another problem with focusing on DRAM latency on Ryzen is that Ryzen has ~90ns of total latency between the CPU cores and DRAM versus ~70ns for Intel's CPUs. Shaving 1.67ns from 90ns isn't going to be as helpful as shaving the same 1.67ns from 70ns. AMD still has a lot of work to do on its memory controller and fabric.

I agree, people often completely forget that most of the latency happens inside the CPU and that tighter timings result in a very small percentage increase in latency over larger bandwidth. Memory hasn't been an issue in a very long time for a reason, lately we're just discussing it because the fabric's clock is tied to the memory and thus adding faster memory essentially overclocks your CPU's fabric.

 

[goldstone77]

Corsair is the lowest latency RAM I could find on the market. This is a huge difference in latency, which I would like to see someone benchmark.

Once you account for the clock frequency difference, 2400-10 is only 1.67ns (16.7%) faster than 3200-16's 10ns. On the other hand, the fabric is operating 33% faster with the 3200MT/s memory. In the end, you would likely be better off with the cheaper 3200-16 DDR4 in most cases.

Another problem with focusing on DRAM latency on Ryzen is that Ryzen has ~90ns of total latency between the CPU cores and DRAM versus ~70ns for Intel's CPUs. Shaving 1.67ns from 90ns isn't going to be as helpful as shaving the same 1.67ns from 70ns. AMD still has a lot of work to do on its memory controller and fabric.

I agree, people often completely forget that most of the latency happens inside the CPU and that tighter timings result in a very small percentage increase in latency over larger bandwidth. Memory hasn't been an issue in a very long time for a reason, lately we're just discussing it because the fabric's clock is tied to the memory and thus adding faster memory essentially overclocks your CPU's fabric.

Hey I just wanted to let you know that some testing results have been posted show the difference of clock frequency vs. latency of RAM with Ryzen. Unlike Intel Ryzen does benefit quite a bit from lower latency. I would really like to see where the sweet spot is between frequency and latency. Like I suggested before 2400 CAS 10 RAM or an overclock of it may result in really beneficial results!
I ran the Gears4 benchmark at the three frequencies again, but with timings of 16-18-18-18-36. I will display the results in a table below. On average, going from 16-18-18-18-36 to 14-16-16-16-36 increased fps by 5% (max 8% at 2133MHz) regardless of frequency.
https://community.amd.com/message/2790538#comment-2790538
https://community.amd.com/servlet/JiveServlet/showImage/2-2790538-119043/pastedImage_0.png

 

[Avro Arrow]

A noble effort but again falls far short of the pre-launch claims ... Deja Vu all over again.

Pre-Launch claims were a 40% increase in IPC over Excavator. The reality turned out to be a 52% increase. Just how is that falling short?

 

[Avro Arrow]

Disappointing gaming benchmarks. Good for other usage. For now, if your on the market the 7700K is looking pretty good.

In BF1, the i7-7700K is running at 100%, it's totally tapped out. It might be better for the next year maybe but that's about it. In the same game, the R7 has that same GTX 1080ti running at 100% while it is only at around 50% usage or more across all sixteen threads. If that test gets redone with Vega 10, expect the R7 to pull away as it will be able to use more of itself to push a Vega while the 7700K can't push anything faster than the 1080ti without a serious OC. Of course, with all the heat problems that the i7-7700Ks are having you probably don't want to do that.

 

[Avro Arrow]

With the exception of this review that manages to keep Intel in the lead in nearly EVERYTHING, you'll find that Hardware Canucks and AnandTech are far more fair.

Every review that I have seen shows Ryzen being quite competitive if not winning in multi-threaded productivity against the i7-6900k and losing practically all gaming benchmarks to the i7-7700k. No surprise there as single-threaded performance is still king in most games. Canuks is no different and Anand omitted gaming benchmarks from its initial review, so its incomplete picture skews perception.

As far as I am concerned, Ryzen delivered pretty much exactly what I was expecting based on leaks and rumors. The most shocking things IMO are the poor memory compatibility at launch and next to no overclocking headroom beyond 4GHz using even high-end conventional cooling.

Actually, with Ryzen, AMD actually delivered MORE than promised (yes, I was also shocked) with a 52% IPC increase over Excavator instead of 40%. Other than that, you're spot-on but the tone of the review and the way it was set up made it literally impossible for Ryzen to get a win ANYWHERE! Ryzen is designed to be the consummate all-rounder. What this review did is EXACTLY what so many review sites did with the Phenom II X4, they put it up against a CPU that it was NEVER aimed at. There should have been THREE CPUs in this review, the R7-1800X, the i7-6900K and the i7-6950X. That's it, nothing more. Those were the two CPUs that AMD was aiming at, those are Intel's flagships. Instead, Alcorn takes the R7 and pits it against the ENTIRE INTEL LINEUP! Let's do a baseball analogy since so many people here are Americans. I guess that Derek Jeter had a disappointing career because he didn't have the most Home Runs, RBIs, stolen bases, double plays, etc. I guess a team that was looking for any one of those qualities wouldn't have wanted him because they'd have found him to be "disappointing". Calling Ryzen disappointing sounds every bit as retarded as that.

 

[Avro Arrow]

A lot of text and nothing behind.... To compare results right, the used hardware and clock rates must be 1:1 the same. But explain this iron law some fanboys. Impossible. 😀

I can benchmark only physical existing things and not prices. Prices comes and go, the hardware will be the same. And dude, we tested all CPUs at stock and one time at the same, fixed clock rate that gave AMD more advantages than Intel.

My 2D tests and also AutoCAD shows, that the windows scheduler has problems with Ryzen and that the latency is a bottleneck in some cases. This is (in my opinion) the follow of a not optimized hardware and software, nothing else. Fast launch, too early. As I wrote earlier.

Nothing behind it even though I posted evidence? Prices come and go... $500 vs. $2000? Are you joking? It's painfully apparent that you're unable to see any fault in anything you do. I've said my piece, some agree, some don't. Make no mistake though, there are people who took the time to read it and now see what I saw.

 

[Avro Arrow]

Your post are tl;dr... Avro...

Is it too much of an impossibility that the NVIDIA TITAN that was used at HardOCP for their gaming test and not the Titan XP that was use for their VR Bench causing that performance difference against Tom's GTX 1080?

And the 3.8Ghz was already explained many times as comparison point between the CPU at the same clock.

If the post is TLDR, then you can't possibly have anything relevant to say about it.

 

[TJ Hooker]

Actually, with Ryzen, AMD actually delivered MORE than promised (yes, I was also shocked) with a 52% IPC increase over Excavator instead of 40%.

Where are you getting the 52% number from? Not saying it's wrong, just wondering where it comes from.

 

[madmatt30]

@Avro Arrow - your commitment is to be applauded mate but tbh I think you're knocking on the door to a house that has been empty awhile - have you been away or something ??

Many of the negative points made have been disproved on other threads , the optimisation is happening slowly but surely & many of the previous comment on this thread have become fairly worthless now.

The bigoted posts ?? & there are a few wildly bigoted members on here who will just choose to ignore any practical evidence you throw at them.
Yup us experienced /older members know who to ignore anyway in that respect.

 

[InvalidError]

Actually, with Ryzen, AMD actually delivered MORE than promised (yes, I was also shocked) with a 52% IPC increase over Excavator instead of 40%.

Where are you getting the 52% number from? Not saying it's wrong, just wondering where it comes from.

He likely got that straight from Lisa Su's (AMD's CEO) Ryzen presentation - it is in at least one of her slides.

 

[jimmysmitty]

With the exception of this review that manages to keep Intel in the lead in nearly EVERYTHING, you'll find that Hardware Canucks and AnandTech are far more fair.

Every review that I have seen shows Ryzen being quite competitive if not winning in multi-threaded productivity against the i7-6900k and losing practically all gaming benchmarks to the i7-7700k. No surprise there as single-threaded performance is still king in most games. Canuks is no different and Anand omitted gaming benchmarks from its initial review, so its incomplete picture skews perception.

As far as I am concerned, Ryzen delivered pretty much exactly what I was expecting based on leaks and rumors. The most shocking things IMO are the poor memory compatibility at launch and next to no overclocking headroom beyond 4GHz using even high-end conventional cooling.

Actually, with Ryzen, AMD actually delivered MORE than promised (yes, I was also shocked) with a 52% IPC increase over Excavator instead of 40%. Other than that, you're spot-on but the tone of the review and the way it was set up made it literally impossible for Ryzen to get a win ANYWHERE! Ryzen is designed to be the consummate all-rounder. What this review did is EXACTLY what so many review sites did with the Phenom II X4, they put it up against a CPU that it was NEVER aimed at. There should have been THREE CPUs in this review, the R7-1800X, the i7-6900K and the i7-6950X. That's it, nothing more. Those were the two CPUs that AMD was aiming at, those are Intel's flagships. Instead, Alcorn takes the R7 and pits it against the ENTIRE INTEL LINEUP! Let's do a baseball analogy since so many people here are Americans. I guess that Derek Jeter had a disappointing career because he didn't have the most Home Runs, RBIs, stolen bases, double plays, etc. I guess a team that was looking for any one of those qualities wouldn't have wanted him because they'd have found him to be "disappointing". Calling Ryzen disappointing sounds every bit as retarded as that.

I prefer CPU reviews that are more complete instead of just picking what you want it to be seen against. It is nice to help others see what they could start at and upgrade to.

If they pit it just against one or two I feel it would be cherry picking, much like the Bulldozer AMD benchmarks that only pitted the CPU against what it won against. Or the people who use to get mad and not want to compare what AMD had to whatever Intels current lineup was but rather older lineups to shine it in a more positive light. I say compare the best mainstream Intel CPU and high end CPU with AMDs best available.

Besides while it is targeting the performance of the higher end i7s it still is within the price range of mainstream i5s and i7s which means there is plenty of people who will consider it if those i5s and i7s beat Ryzen in their specific performance metric, such as if the i7 7700K does better in gaming than the 1800X and cost less, why would they not buy the i7?