Discussion AMD Ryzen MegaThread! FAQ and Resources

[juanrga]

Ryzen has better 1% and .1%. This because it has more cores, more resources. Look at the 5930k. Its going to make game play feel much smoother without all those low frame drops. Which is ultimately going to give you a better gaming experience!

http://www.gamersnexus.net/game-bench/2182-fallout-4-cpu-benchmark-huge-performance-difference

Hum! The i5-4690K with less resources and lower clocks got higher 0.1% than the i7-4790k. Their i5-4690K got higher 0.1% than the i5-6600K. That graph is clearly invalid.

Also we don't know if the i7-5930K has better 0.1% because it has more cores or because it has a better memory subsystem.

Juanrga GamersNexus.net does an amazing job with Benchmarks, and they really know their stuff on the technical side! Calling it an invalid graph, because you think one CPU should perform better than another is a little ridiculous. Anyone watching benchmarks for a good number of years knows that things don't always happen as you expect them too. If you read the commentary section for the YouTube video he discusses it.
barondorado
"4690k outperforming the 6600k lel"
Gamers Nexus
"Happened to the 6700K vs 4790K in some games also. Seems to be a latency hangup."
Tony DoMonte
"anandtech covers it in the skylake review as to why. i dont remember the specifics but you can check it out."
https://www.youtube.com/watch?v=1WjEWrDyPmk&feature=youtu.be

The i7 has the same Haswell muarch, but more cache, more threads, and higher clocks. It is weird that it got lower score. If it got a lower score then something was wrong: methodology or hardware. Just saying hey looks the i5 was faster than the i7 doesn't solve the problem. Same with the i5 Skylake vs the i5 Haswell.

This logic applies to any rational measurement. We don't do measurements and publish anything we measure. There are hundred reasons that a measurement can go wrong. That is why the results of measurements have to be checked before publishing. Outliers and blatantly wrong results cannot be published.

Moreover, this rationality is even more needed in this special case. The argument made was that the R5 would provide a smoothing gaming experience because it has more execution resources than i5. We cannot then link to a bench where a i5 gets higher 0.1% than an i7, if only for consistency of the argument.

 

[juanrga]

The only issue I see with juan's benchmarks is that they are from the first reviews of Ryzen, which lack the most recent patches and bios applied. Ryzen is still not as good as Intel in gaming, but the real difference should be measured again today, otherwise we will keep arguing over old numbers.

The second performance graph in my former post

http://www.tomshardware.co.uk/forum/id-3327589/amd-ryzen-megathread-faq-resources/page-33.html#19628916

is from a recent review that includes all the last patches/upgrades, including the last AGESA/BIOS 1.0.0.4 with the latency reductions. Things have changed in the low single digit percent since the first reviews, because no magic patch/BIOS was going to change RyZen muarch. I have been saying since launch, that no magic 'fix' will improve RyZen average performance by 10--20%. Patches can affect one or two games that were broken, as it happened with AoTs, but individual patches will not change the average gaming behavior. And BIOS updates can change situation only by 5% or so, because a ~6ns latency reduction is not going to do anything when your latency gap was about one order of magnitude higher than that.

RyZen problem is structural and it will be not solved, if it is, until Zen2.

 

[juanrga]

Stutter is a latency problem when one or more frames take more time to render. Overclocking RAM reduces access latency and this can eliminate stutter if the bottleneck was memory. Check the graphs, specially the i3 numbers. Average increased by 25%, but minimum increased a huge 59%. which reduces the possibility of stuttering.

Timings reduce latency, Mhz increase effective bandwidth. Isn't that the rule of thumb? Something that is bandwidth starved might not be latency sensitive at all. If you increase bandwidth, then you're talking about something different.

And it has nothing to do with what I said anyway. Increasing effective FPS (min or max) does not decrease the *chance* of stutter, it actually increases it! That is why it's important to measure how the frame generation is distributed along a benchmark run and then inversely correlate the information you get.

Stutter is a problem that is not due to latency (cas level) or bandwidth (MT/s), it is due to the CPU not providing the data quick enough due to overall workload. RAM latency has virtually nothing to do with that at all, which is why most CPUs see single digit performance gains even making massive leaps in RAM frequency or cas level.

There are some games on very dated engines that are extremely single core sensitive that see higher gains, like FO4 for example, or Arma3, but beyond those outliers...going from DDR4-2400 to DDR4-3600 (50% increase in MT/s) only yields something like 6-8% performance gain on Intel systems, even if you keep the cas level the same (total RAM latency).

I am not discussing CAS latency, but frame latency. If a CPU (with a given GPU) does 100 frames per second, then it takes 0.01 seconds to draw each frame on average. The frame total latency T is the sum of a compute contribution plus a memory contribution. If the game is compute-bound, then T ~ T_comp. If the game is memory bound, then T ~ T_mem.

This T_mem is given by the time to access the data plus the time required to move the needed data to the CPU,

T_mem = T_access + T_transp

increasing memory clocks reduces both latencies, because the increased bandwidth reduces T_transp, and because CAS latencies are relative, not absolute, and increasing clocks reduces T_access through the formula given by goldstone77.

Stuttering is the reflection of some instantaneous higher-than-average bottleneck that increases the frame latency above the average.

The game tested by jdwii is memory bound and the more probable cause for stuttering is some instantaneous higher-than-average bottleneck when accessing memory. Increasing memory clocks reduces the bottleneck, which affects both average latencies and the possibility of stuttering. That is also why minimum framerates got much higher increase than average framerates when going from 1333MHz to 2400MHz.

Things go to other side on compute-bound games. In those games stuttering is generated because at some instant there are not enough compute resources and one or more frames take more time to render (higher-than-average frame latency). Stuttering in this case is solved by increasing the compute resources, e.g. moar cores or OC the CPU.

You do realize RAM latency and bandwidth have nothing to do with frame latency, or VRAM bandwidth, right?

Not only I have just explained how memory latency can affect frame latency on memory-bound games, but I provided benches showing how frame latency is reduced by up 70% on certain games when RAM is clocked higher.

We are talking IPC in that chart, not clockspeed gaps stemming from process differences. A 20% gap in clockspeed generating a 20% gap in performance means IPC is basically the same.

Happy to educate you about the graph

Then the graph is even more useless, because we cannot extract IPC from comparing 4-core to 8-core. If we take this ordering

we obtain a score of 139.50 for an 7700k underclocked at the R7 1700x levels. And this score is only 4% higher, but not the IPC gap. To get the IPC gap we have to compare the (4-core/8-threads) 7700k with the (4-core/8-threads) R5 1500/1400 or compare the (8-core/16-threads) 6900k with some of the (8-core/16-threads) R7 models...

We can see the R5 1500X/1400 are more close to Sandy/Ivy than to Kabylake.

 

[Arc911]

Ryzen seems to perform better with high end cards than it does with low end/mid range cards.
It needs a discussion as said by this reviewer here : https://youtu.be/BjoagsOxMn0?t=6m48s

Also this review says the same : https://youtu.be/kVS0NJyWZus?t=45s

Why does Ryzen seem to perform better with high end cards when ram is oc ed vs when using low end cards n non oc ed ram ?
Is it the Infiniy Fabric ? Or is it just same with intel ?
Can someone please shed some light on this

{p.s im aware that these reviwers are not well known but this does bring up a theory so as to speak}

 

[jaymc]

Stock Madness: AMD Shares Crater 25% In A Day.

https://www.youtube.com/watch?v=xPAlcMpyjgU

 

[cdrkf]

Ryzen seems to perform better with high end cards than it does with low end/mid range cards.
It needs a discussion as said by this reviewer here : https://youtu.be/BjoagsOxMn0?t=6m48s

Also this review says the same : https://youtu.be/kVS0NJyWZus?t=45s

Why does Ryzen seem to perform better with high end cards when ram is oc ed vs when using low end cards n non oc ed ram ?
Is it the Infiniy Fabric ? Or is it just same with intel ?
Can someone please shed some light on this

{p.s im aware that these reviwers are not well known but this does bring up a theory so as to speak}

So I watched that vid- so the 'ryzen performing better with high end card' is just that the G4560 is bottlenecking the high end card so it can't go any faster. The 4 core 8 thread R5 1400 has a lot more power behind it so it can push the 1080ti up to faster frame rates.

When you switch to the RX480 in quite a few of those charts you see the G4560 and R5 1400 sitting close to each other- this isn't Ryzen performing badly but the fact that at 1080 ultra settings the RX480 is hitting it's limit (so there is no gap). Ryzen still performs well (there are a few tests which are more cpu heavy where it's still way out in front).

The key is however getting balance- there's no point pairing a Ryzen 7 1800x with a RX550 entry level gpu, just the same as there is no point pairing the GTX 1080ti with the Pentium G4560. With a slow gpu (e.g. gtx 1050, RX 460 class) the G4560 is often enough to keep the gpu fed and running flat out. If you are looking for a higher end build with something like a GTX 1060, RX480 or higher than you'll probably need more cpu grunt like an i5 or Ryzen 5 cpu.

The only result in there that was out of place was where the G4560 was ahead of the R5 with the RX480- although I only saw that once and it's worth noting the *minimum* frame rates where better with the R5.

As for ram- Ryzen will perform better with faster ram, although the same is also true for Intel cpu's. However again if you have an entry level graphics board that will often become the limiting factor meaning you wont see much (if any) scaling from memory.

 

[8350rocks]

Spoiler

Stutter is a latency problem when one or more frames take more time to render. Overclocking RAM reduces access latency and this can eliminate stutter if the bottleneck was memory. Check the graphs, specially the i3 numbers. Average increased by 25%, but minimum increased a huge 59%. which reduces the possibility of stuttering.

Timings reduce latency, Mhz increase effective bandwidth. Isn't that the rule of thumb? Something that is bandwidth starved might not be latency sensitive at all. If you increase bandwidth, then you're talking about something different.

And it has nothing to do with what I said anyway. Increasing effective FPS (min or max) does not decrease the *chance* of stutter, it actually increases it! That is why it's important to measure how the frame generation is distributed along a benchmark run and then inversely correlate the information you get.

Stutter is a problem that is not due to latency (cas level) or bandwidth (MT/s), it is due to the CPU not providing the data quick enough due to overall workload. RAM latency has virtually nothing to do with that at all, which is why most CPUs see single digit performance gains even making massive leaps in RAM frequency or cas level.

There are some games on very dated engines that are extremely single core sensitive that see higher gains, like FO4 for example, or Arma3, but beyond those outliers...going from DDR4-2400 to DDR4-3600 (50% increase in MT/s) only yields something like 6-8% performance gain on Intel systems, even if you keep the cas level the same (total RAM latency).

I am not discussing CAS latency, but frame latency. If a CPU (with a given GPU) does 100 frames per second, then it takes 0.01 seconds to draw each frame on average. The frame total latency T is the sum of a compute contribution plus a memory contribution. If the game is compute-bound, then T ~ T_comp. If the game is memory bound, then T ~ T_mem.

This T_mem is given by the time to access the data plus the time required to move the needed data to the CPU,

T_mem = T_access + T_transp

increasing memory clocks reduces both latencies, because the increased bandwidth reduces T_transp, and because CAS latencies are relative, not absolute, and increasing clocks reduces T_access through the formula given by goldstone77.

Stuttering is the reflection of some instantaneous higher-than-average bottleneck that increases the frame latency above the average.

The game tested by jdwii is memory bound and the more probable cause for stuttering is some instantaneous higher-than-average bottleneck when accessing memory. Increasing memory clocks reduces the bottleneck, which affects both average latencies and the possibility of stuttering. That is also why minimum framerates got much higher increase than average framerates when going from 1333MHz to 2400MHz.

Things go to other side on compute-bound games. In those games stuttering is generated because at some instant there are not enough compute resources and one or more frames take more time to render (higher-than-average frame latency). Stuttering in this case is solved by increasing the compute resources, e.g. moar cores or OC the CPU.

You do realize RAM latency and bandwidth have nothing to do with frame latency, or VRAM bandwidth, right?

Not only I have just explained how memory latency can affect frame latency on memory-bound games, but I provided benches showing how frame latency is reduced by up 70% on certain games when RAM is clocked higher.

We are talking IPC in that chart, not clockspeed gaps stemming from process differences. A 20% gap in clockspeed generating a 20% gap in performance means IPC is basically the same.

Happy to educate you about the graph

Then the graph is even more useless, because we cannot extract IPC from comparing 4-core to 8-core. If we take this ordering

we obtain a score of 139.50 for an 7700k underclocked at the R7 1700x levels. And this score is only 4% higher, but not the IPC gap. To get the IPC gap we have to compare the (4-core/8-threads) 7700k with the (4-core/8-threads) R5 1500/1400 or compare the (8-core/16-threads) 6900k with some of the (8-core/16-threads) R7 models...

We can see the R5 1500X/1400 are more close to Sandy/Ivy than to Kabylake.

1.) Benchmarks from memory bound games do not show a general trend. There are only about 3-4 popular games that are heavily memory bound, and you posted the benchmarks for those. The other 1,000,000+ games are different from those 4.

2.) You are insinuating that somehow system memory is going to impact the draw time of the GPU producing frames? I could understand talking about VRAM, but there is no logical connection to support system memory impacting GPU draw times. System memory would also have no impact on CPU issuing draw calls either, since the resources for the game would be loaded into cache.

3.) If you are calculating IPC, trying to extrapolate from multiple cores is apples and oranges. IPC is a per core measurement. So compare single core benchmarks...there are many out there. Why are you trying to construe multi-core throughput as being somehow equivalent to IPC? There are different terms for these things for specific reasons. Single core IPC benchmarks are available, those show the IPC gap. Multi-core throughput benchmarks are available, and those show the difference in throughput. Which is it you are trying to accomplish?? I am confused because it appears to me you are trying to convert throughput into IPC, and that is a fool's errand because results will be wildly inaccurate.

 

[jdwii]

^ With my own testing i am trying my best to figure out the IPC of this chip or to keep some members in this forum from shaking their head performance per cycle haha.

It's not easy i'll say that much, finding tests that run on just 1 thread or one core

https://browser.geekbench.com/v4/cpu/2748993

Ryzen 1700 3.7Ghz 2933mhz ram Amd Balance
My Geekbench run
Single core
4233
Multi-core
23216

This is the test at performance mode instead of Amd's latest Balance mode
https://browser.geekbench.com/v4/cpu/2749164
Single Core 4263
Multi-core 23502

Dolphin benchmark gets 503 seconds with Amd balance on instead of performance which gets 490 seconds. So just to get that little tiny bit more performance i'm keeping my system at performance instead of Amd's balance

 

[jdwii]

Spoiler

Ryzen seems to perform better with high end cards than it does with low end/mid range cards.
It needs a discussion as said by this reviewer here : https://youtu.be/BjoagsOxMn0?t=6m48s

Also this review says the same : https://youtu.be/kVS0NJyWZus?t=45s

Why does Ryzen seem to perform better with high end cards when ram is oc ed vs when using low end cards n non oc ed ram ?
Is it the Infiniy Fabric ? Or is it just same with intel ?
Can someone please shed some light on this

{p.s im aware that these reviwers are not well known but this does bring up a theory so as to speak}

So I watched that vid- so the 'ryzen performing better with high end card' is just that the G4560 is bottlenecking the high end card so it can't go any faster. The 4 core 8 thread R5 1400 has a lot more power behind it so it can push the 1080ti up to faster frame rates.

When you switch to the RX480 in quite a few of those charts you see the G4560 and R5 1400 sitting close to each other- this isn't Ryzen performing badly but the fact that at 1080 ultra settings the RX480 is hitting it's limit (so there is no gap). Ryzen still performs well (there are a few tests which are more cpu heavy where it's still way out in front).

The key is however getting balance- there's no point pairing a Ryzen 7 1800x with a RX550 entry level gpu, just the same as there is no point pairing the GTX 1080ti with the Pentium G4560. With a slow gpu (e.g. gtx 1050, RX 460 class) the G4560 is often enough to keep the gpu fed and running flat out. If you are looking for a higher end build with something like a GTX 1060, RX480 or higher than you'll probably need more cpu grunt like an i5 or Ryzen 5 cpu.

The only result in there that was out of place was where the G4560 was ahead of the R5 with the RX480- although I only saw that once and it's worth noting the *minimum* frame rates where better with the R5.

As for ram- Ryzen will perform better with faster ram, although the same is also true for Intel cpu's. However again if you have an entry level graphics board that will often become the limiting factor meaning you wont see much (if any) scaling from memory.

https://www.youtube.com/watch?v=U4k_ErEg-FU&t=431s

With higher-end GPUs ryzen really benefits from faster ram sped in some cases by 30%

 

[cdrkf]

Spoiler

Ryzen seems to perform better with high end cards than it does with low end/mid range cards.
It needs a discussion as said by this reviewer here : https://youtu.be/BjoagsOxMn0?t=6m48s

Also this review says the same : https://youtu.be/kVS0NJyWZus?t=45s

Why does Ryzen seem to perform better with high end cards when ram is oc ed vs when using low end cards n non oc ed ram ?
Is it the Infiniy Fabric ? Or is it just same with intel ?
Can someone please shed some light on this

{p.s im aware that these reviwers are not well known but this does bring up a theory so as to speak}

So I watched that vid- so the 'ryzen performing better with high end card' is just that the G4560 is bottlenecking the high end card so it can't go any faster. The 4 core 8 thread R5 1400 has a lot more power behind it so it can push the 1080ti up to faster frame rates.

When you switch to the RX480 in quite a few of those charts you see the G4560 and R5 1400 sitting close to each other- this isn't Ryzen performing badly but the fact that at 1080 ultra settings the RX480 is hitting it's limit (so there is no gap). Ryzen still performs well (there are a few tests which are more cpu heavy where it's still way out in front).

The key is however getting balance- there's no point pairing a Ryzen 7 1800x with a RX550 entry level gpu, just the same as there is no point pairing the GTX 1080ti with the Pentium G4560. With a slow gpu (e.g. gtx 1050, RX 460 class) the G4560 is often enough to keep the gpu fed and running flat out. If you are looking for a higher end build with something like a GTX 1060, RX480 or higher than you'll probably need more cpu grunt like an i5 or Ryzen 5 cpu.

The only result in there that was out of place was where the G4560 was ahead of the R5 with the RX480- although I only saw that once and it's worth noting the *minimum* frame rates where better with the R5.

As for ram- Ryzen will perform better with faster ram, although the same is also true for Intel cpu's. However again if you have an entry level graphics board that will often become the limiting factor meaning you wont see much (if any) scaling from memory.

https://www.youtube.com/watch?v=U4k_ErEg-FU&t=431s

With higher-end GPUs ryzen really benefits from faster ram sped in some cases by 30%

Yes ofc- my point was though that it wasn't a case of Ryzen performing poorly with a lower end gpu- it's just that in that case the gpu is the bottleneck so host processor won't really change the benchmark much (same goes for memory). With a high end gpu the bottleneck will shift back towards the cpu / ram so more to gain.

 

[jaymc]

MSI's Best Secret Of Ryzen trip to AMD in Austin!

https://www.youtube.com/watch?v=gAVA4khyWg8

 

[jdwii]

Man why don't people talk about how cool ryzen runs this is crazy i almost don't trust the temps

http://s23.postimg.org/trg2jh61n/Temps.png

This is on my Kraken X61 with fans set to 25% only. Voltage on ryzen is set to 1.225V 3.7Ghz, These temps are after 1 hour of benchmarks and 2 hours of gaming in fallout 4. Temp in my room is 68F

Most of the time temps in fallout 4 were at 32C or so. Like man that is half the temp my Haswell chip was running at with the same cooler. Intel please go back to your old way of putting the heat spreader on the CPU!

With my wraith cooler getting only 10-15C higher temps i must say i think Ryzen will do well in laptops they need to make some 65 watt 4 cores with SMT and a integrated GPU for a laptop. Or just bring a low clock 6 core part of ryzen for laptops.

 

[-Fran-]

What caught my eye was the reported core at 3.8Ghz where all others were at 3.7Ghz. So it has per-core speed? Even more, per logical core speed? How does it even work? lol

Cheers!

 

[gamerk316]

2.) You are insinuating that somehow system memory is going to impact the draw time of the GPU producing frames? I could understand talking about VRAM, but there is no logical connection to support system memory impacting GPU draw times. System memory would also have no impact on CPU issuing draw calls either, since the resources for the game would be loaded into cache.

Remember textures need to get into main memory before they get to the CPU. As a general rule, so much is pre-loaded nowadays where its really hard for a legitimate memory bottleneck to occur in gaming (the HDD is far more likely to be the culprit if data isn't in memory yet), but it certainly can happen under certain situations. At the end of the day, if the GPU is sitting around waiting for data, yeah, that's going to affect latency.

 

[gamerk316]

Man why don't people talk about how cool ryzen runs this is crazy i almost don't trust the temps

http://s23.postimg.org/trg2jh61n/Temps.png

This is on my Kraken X61 with fans set to 25% only. Voltage on ryzen is set to 1.225V 3.7Ghz, These temps are after 1 hour of benchmarks and 2 hours of gaming in fallout 4. Temp in my room is 68F

Most of the time temps in fallout 4 were at 32C or so. Like man that is half the temp my Haswell chip was running at with the same cooler. Intel please go back to your old way of putting the heat spreader on the CPU!

With my wraith cooler getting only 10-15C higher temps i must say i think Ryzen will do well in laptops they need to make some 65 watt 4 cores with SMT and a integrated GPU for a laptop. Or just bring a low clock 6 core part of ryzen for laptops.

As has been noted, AMD chose a mobile based process for Ryzen, which is playing a large part of keeping power/temps down at stock settings. Start to approach 4 GHz though, and power/temp draw increases at a fairly decent clip, again due to the process chosen. I'm not surprised Ryzen is fairly cool at stock.

 

[Rogue Leader]

Everyone,

Its that time again, time for a groupwide warning. I want you ALL to take a look through the last 2 pages of posts. All I see is calling out, aggressive responses, and even worse, it seems some people have taken to moving the goal posts around just to make their argument valid and use that to call people out. Now what I SHOULD do is delete all this stuff but I'm going to let it live as the discussion is otherwise productive.

That however ends with this post. So read my warning above, consider what you are writing and know that we are watching, and if I continue to see this, you will at a minimum see your posts deleted. At worst you won't be here any longer. You all know we have done it before in this thread, I don't want to have to do it again but I will.

 

[juanrga]

We are talking IPC in that chart, not clockspeed gaps stemming from process differences. A 20% gap in clockspeed generating a 20% gap in performance means IPC is basically the same.

Happy to educate you about the graph

I found the source for the chart. It is a tweet from AMD

https://twitter.com/AMDRyzen/status/858028371141419008

The original is an animated GIF and the last part of the GIF contains the specifications. It is not IPC measurement but 8-core stock vs 4-core stock. The 1080p gaming part of the chart is not from AMD measurements, but apparently obtained from third parties

Note that the information is contradictory. One asterisk in the chart means Premium Gaming and two asterisks in the chart mean 1080p gaming, but in the specifications, the two-asterisk footnote mentions "Premium Gaming" as well.

I have absolutely no idea of how AMD got the 8% gap in the chart. Other reviews give ~20% gap

Even one of the sources used by AMD gives 13% gap

Note that PCGamer uses geometric means instead arithmetic means, and this must be closing the gap.

 

[jdwii]

^^^ Something is terribly wrong with those results either slow memory or a bias selection of games. It's important to note that you can't get IPC from games unless you disable one CCX(even then its hard to do). From my own testing i'm seeing ryzen close to haswell in most cases in single core in others its better then ivy-sandy but fails to beat haswell like in dolphin emulator.

http://www.kitguru.net/components/cpu/luke-hill/amd-ryzen-5-1500x-4c8t-cpu-review/

This is my favorite review as they include both haswell and sandy. They also OC sandy and ryzen to 3.9Ghz.

Others should take a look at this as well, but even then it should be noted that games do not understand ryzen as two quad cores.

 

[juanrga]

^^^ Something is terribly wrong with those results either slow memory or a bias selection of games. It's important to note that you can't get IPC from games unless you disable one CCX(even then its hard to do). From my own testing i'm seeing ryzen close to haswell in most cases in single core in others its better then ivy-sandy but fails to beat haswell like in dolphin emulator.

http://www.kitguru.net/components/cpu/luke-hill/amd-ryzen-5-1500x-4c8t-cpu-review/

This is my favorite review as they include both haswell and sandy. They also OC sandy and ryzen to 3.9Ghz.

Others should take a look at this as well, but even then it should be noted that games do not understand ryzen as two quad cores.

Those reviews are using stock RAM. Kitguru is using OC RAM, which I find a bit weird and unfair: 3200MHz Ryzen vs 2400Mhz Haswell!! Not to mention that their Haswell chip was an engineering sample.

Despite all that, the conclusions of the kitguru reviews aren't very different from rest of reviews:

Gaming performance for Ryzen is still off the pace set by Kaby Lake and other Intel architectures. That point is still true for Ryzen 5 in its four-core, eight-thread guise. There were instances where the Core i5 was significantly faster when paired with a powerful graphics card.

Switch to a 4K gaming resolution and you’ll struggle to notice a performance difference between Ryzen 5 1500X and Core i5-7400. However, even at 1440P when paired with a GTX 1070, the multiplier-locked Kaby Lake i5 is a little quicker.

Worth noting, though, is Ryzen 5 1500X’s spare CPU capacity whilst gaming thanks to its SMT implementation. Those four additional threads compared to a Core i5 can be used to power background activities such as game streaming. This is a task that will result in definite slow-down on the Core i5-7400 as it tended to stick close to 100% utilisation when delivering its gaming results.

Power draw on the 1500X is noticeably higher than that of the Core i5-7400. Disproportionally so, too, as the increase in power draw is larger than the performance boost obtained by opting for AMD’s eight-thread chip. With that said, the levels are far from concerning and they are not going to add significant cost to an energy bill.

Thermal performance was good, even with the included Wraith Spire cooler (which is a potent heatsink fan). We dialled in 1.40V for the overclock and were met by temperatures well below 80C. Core i5-7400 runs cooler, despite its physically inferior stock cooler, thanks largely to lower power draw.

Our conclusion for the Ryzen 5 1500X is somewhat mixed. If you have a hard budget of less than £200 for a CPU, the 4C8T AMD part offers superb value for money in computational workloads. However, Intel’s Kaby Lake competitors are faster if all you do is play games.

If you have a mixed variety of workloads which include gaming and productivity, Ryzen 5 1500X puts itself forward as a solid choice. This is thanks to its balance between productivity capabilities and good enough gaming performance when paired with a sensible graphics card for buyers in this market.

Unlike the Ryzen 5 flagship 1600X that is overshadowed by its cheaper, overclockable sibling, the 1500X is worth its $20 price increase over the 1400. You get noticeably higher out-of-the-box frequencies (especially for XFR), double the shared L3 cache, and a better CPU cooler.

As such, we have no problem giving the 1500X a positive recommendation if you use your system for more than just gaming. Significant challenges will, however, be thrown the 1500X’s way if there is any price reduction on the £30 more expensive six-core Ryzen 5 1600.

 

[jdwii]

Ryzen must be tested with 2933-3200mhz memory in fairness sake i say let all the platforms be tested with the fastest memory but ryzen's performance tanks pretty bad(20-30% In cases) in games when used with high-end GPU's with slower memory.

https://www.youtube.com/watch?v=XOsYOASddeo&t=689s

But anyways that review at kitguru seems to be my favorite as it actually tests ryzen with high-speed memory and it tests sandy 4 core 8 thread at a ryzen 4 core 8 thread.

Part of me wish's Amd just made normal 4-6-8 core processors then things would be easier to test.

 

[salgado18]

Ryzen must be tested with 2933-3200mhz memory in fairness sake i say let all the platforms be tested with the fastest memory but ryzen's performance tanks pretty bad(20-30% In cases) in games when used with high-end GPU's with slower memory.

https://www.youtube.com/watch?v=XOsYOASddeo&t=689s

But anyways that review at kitguru seems to be my favorite as it actually tests ryzen with high-speed memory and it tests sandy 4 core 8 thread at a ryzen 4 core 8 thread.

Part of me wish's Amd just made normal 4-6-8 core processors then things would be easier to test.

My wish is AMD made the 4-core parts with only one CCX, even if delayed by half a year or so.

 

[jdwii]

Though i'm not that big of a fan of joker productions as i do see him as a fanboy that doesn't mean i don't trust his results in some cases he really needs to be more neutural but hey whatever

https://www.youtube.com/watch?v=wPwtlki0FVo

Here is a video of the upcoming prey game and the 1700 is beating a 6800K when the 6800K OC 4.3Ghz and ryzen is at 4.0Ghz.

Also if anyone is interested read extremetech news on what Intel said about temps on the 7700K once OC. Won't post the link here as its not Amd related. But i hope Amd doesn't switch over to TIM based in the future to save money. Intel should really consider going back to the older way even more so since their design overclocks so well.

Edit lol Joker's video went down typical haha

 

[Gon Freecss]

CCX inter-communication doesn't seem to be the problem.

 

[jdwii]

CCX inter-communication doesn't seem to be the problem.

I've been reading and one possible theory is that the PCI-E controller for the dGPU is kind of terrible and needs high performance ram

 

[jdwii]

https://www.overclock3d.net/news/cpu_mainboard/am4_motherboards_will_soon_be_receiving_a_new_agesa_microcode_update/1

YES now perhaps this will fix my ram issue ha ha