• Competitor rules

    Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.

Ryzen IPC

Another issue about how convoluted the whole benchmarking websites/youtubers is in relation with the 7700K and why it performs so much better enjoying it's high clocks than the Ryzen or SkylakeX, is obvious on the video bellow. Look closely at the core by core load between Nvidia and AMD. Same fps between Vega & 1080 (irrelevant to this thread), and same 100% utilization, but focus on how differently the CPU core loads act.
(this is the first ever benchmark run using AMD card against 1080 in such video).
The games usage with the AMD card, spread more evenly across the cores compared to Nvidia one, which clogs one thread at 90%+ some times.
Also observe the overall CPU usage on the NV cards is almost double that of the AMD cards. Still FPS and detail is the same.

And no, I am not "pro AMD Ryzen" or "pro Vega". It shows why SkylakeX and Ryzen perform so poorly against the 7700K on all benchmarks up to now, were common denominator is the GTX1080/1080Ti. (aka NV drivers).
The 7820X shouldn't perform poorly against the 7700K as benchmarks show. On the contrary at 4.5Ghz should be crushing it and the Ryzen 7 or the 5960X/6900K also, let alone the Ryzen 5.
Same applies to the 7900X which is great CPU, even if very overpriced.

 
jrwagh333 said:
Intersting to see you are running Win8 - Any issues installing? As I heard that when installing Window 7, keyboard/mouse does not work unless in PS2 port
Click to expand...

I'm on windows 10 mate, cinebench just reads it as 8

Scougar said:
Wish DDR4 pricing would come down, it really is silly at the moment.
Click to expand...

I dont expect it to get better anytime soon either.
 
Panos said:
Another issue about how convoluted the whole benchmarking websites/youtubers is in relation with the 7700K and why it performs so much better enjoying it's high clocks than the Ryzen or SkylakeX, is obvious on the video bellow. Look closely at the core by core load between Nvidia and AMD. Same fps between Vega & 1080 (irrelevant to this thread), and same 100% utilization, but focus on how differently the CPU core loads act.
(this is the first ever benchmark run using AMD card against 1080 in such video).
The games usage with the AMD card, spread more evenly across the cores compared to Nvidia one, which clogs one thread at 90%+ some times.
Also observe the overall CPU usage on the NV cards is almost double that of the AMD cards. Still FPS and detail is the same.

And no, I am not "pro AMD Ryzen" or "pro Vega". It shows why SkylakeX and Ryzen perform so poorly against the 7700K on all benchmarks up to now, were common denominator is the GTX1080/1080Ti. (aka NV drivers).
The 7820X shouldn't perform poorly against the 7700K as benchmarks show. On the contrary at 4.5Ghz should be crushing it and the Ryzen 7 or the 5960X/6900K also, let alone the Ryzen 5.
Same applies to the 7900X which is great CPU, even if very overpriced.

Click to expand...

You had me intrigued with this, so I paused the video at 3 random points and added up all the individual core usages and divided them by 8 to give me a total CPU usage. At all 3 points the cpu usage with vega is lower.
Excuse my pathetic excuse of a chart but I quickly made this.

upload my pictures

I have no idea what this means and what would happen to vega if it had a higher cpu utilisation but its certainly interesting.
 
gavinh87 said:
You had me intrigued with this, so I paused the video at 3 random points and added up all the individual core usages and divided them by 8 to give me a total CPU usage. At all 3 points the cpu usage with vega is lower.
Excuse my pathetic excuse of a chart but I quickly made this.

upload my pictures

I have no idea what this means and what would happen to vega if it had a higher cpu utilisation but its certainly interesting.
Click to expand...

On all scenes all cards were running at 100% so even if someone pushed more CPU power on the Vega Fe, couldn't perform better.
Howerver while FPS of the Vega FE were constantly in par of the 1080, the Nvidias are using more CPU power, especially on some single cores.
Where those cores were running at 90% on the same scene the VegaFe was at 50%.

Consider a Ryzen CPU at best is running at 4Ghz, mostly 3.9. That 90% @ 4.5Ghz just clogs the related thread is working on in a Ryzen 4Ghz CPU. Hence the lower fps on benchmarks.

If it was spreading evenly across multiple threads like it does with the Vega FE, the perf could have been in par of the 4Ghz Ryzen to the 4.5+ 7700K.

But I bet next week, when such benchmarks start flooding, would make every "benchmark" up to now used NV card irrelevant.
As it happened with all the Ryzen benchmarks from pre-Agesa 1006. Benchmarking the Ryzen 7 with 2666 ram in March (because couldn't run higher), produces completely different results by a hefty 15-20% compared to 3200 or 3466 today after Agesa 1006.
Similarly applies to the SkylakeX CPUs, but that has to do with the mesh optimisation, the NV drivers are lacking.
 
Last edited:
Panos said:
On all scenes all cards were running at 100% so even if someone pushed more CPU power on the Vega Fe, couldn't perform better.
Howerver while FPS of the Vega FE were constantly in par of the 1080, the Nvidias are using more CPU power, especially on some single cores.
Where those cores were running at 90% on the same scene the VegaFe was at 50%.

Consider a Ryzen CPU at best is running at 4Ghz, mostly 3.9. That 90% @ 4.5Ghz just clogs the related thread is working on in a Ryzen 4Ghz CPU. Hence the lower fps on benchmarks.

If it was spreading evenly across multiple threads like it does with the Vega FE, the perf could have been in par of the 4Ghz Ryzen to the 4.5+ 7700K.

But I bet next week, when such benchmarks start flooding, would make every "benchmark" up to now used NV card irrelevant.
As it happened with all the Ryzen benchmarks from pre-Agesa 1006. Benchmarking the Ryzen 7 with 2666 ram in March (because couldn't run higher), produces completely different results by a hefty 15-20% compared to 3200 or 3466 today after Agesa 1006.
Similarly applies to the SkylakeX CPUs, but that has to do with the mesh optimisation, the NV drivers are lacking.
Click to expand...

Well hopefully this will make nvidia take a look at how their driver performs. Otherwise a lot of people with multicore CPU's will be going to AMD.
 
gavinh87 said:
Looking around I found this. With the 1800x having 4 of its cores disabled

upload multiple image

http://www.zolkorn.com/reviews/amd-ryzen-7-1800x-vs-intel-core-i7-7700k-mhz-by-mhz-core-by-core/3/
Click to expand...

If that's with all cores disabled in a ccx, it's going to be faster than normal because it won't have the latency caused by using infinity fabric to communicate between ccxs.

In a similar way to how the ipc of bulldozer could be improved by disabling one core per module so the resources weren't shared.
 
pandem0nium said:
If that's with all cores disabled in a ccx, it's going to be faster than normal because it won't have the latency caused by using infinity fabric to communicate between ccxs.

In a similar way to how the ipc of bulldozer could be improved by disabling one core per module so the resources weren't shared.
Click to expand...

The Raven Ridge APUs look to be using one CCX.
 
pandem0nium said:
If that's with all cores disabled in a ccx, it's going to be faster than normal because it won't have the latency caused by using infinity fabric to communicate between ccxs.

In a similar way to how the ipc of bulldozer could be improved by disabling one core per module so the resources weren't shared.
Click to expand...

Yeah this is on one ccx. I suppose this could be seen as a future case if AMD can improve the speeds of the IF so it doesn't become a bottleneck anymore.
 
Last edited:
Scougar said:
What is LL, Low Latency? so like 12 or 14 CAS or something? (Because that jump on the green bar is quite significant)
Click to expand...

gavinh87 said:
Yeah low latency. It will be CL14, I dont think anyone has managed CL12 on anything higher than 3200.
If you can manage 3466 CL14 there is no point go for 3600, bigger improvements come from tightening the subs.
Click to expand...

3200MHz LL is CL12 and 3466 LL is CL14. It is possible with Samsung B-Die and came out when The Stilt over at oc.net released a graph showing the differences before started including them in their community blog.

https://community.amd.com/community...emory-oc-showdown-frequency-vs-memory-timings

Worth giving 3200MHz CL12 a go if you can't get any higher on the memclk.
 
CAT-THE-FIFTH said:
The Raven Ridge APUs look to be using one CCX.
Click to expand...
Which makes the talk of an 8c/16t APU rather silly. It's basically a Ryzen CPU with one of the CCXs replaced with an IGP. However, when they move to 7 nm and have 6-core CCXs and may have even moved to Navi for the IGP, their APUs will kick all sorts of ass. People will be enjoying affordable 6c/12t APUs with great performance and looking back on the early 2010s Intel dominated era simply shaking their heads.
 
pandem0nium said:
Great stuff.
Click to expand...

DragonQ said:
Which makes the talk of an 8c/16t APU rather silly. It's basically a Ryzen CPU with one of the CCXs replaced with an IGP. However, when they move to 7 nm and have 6-core CCXs and may have even moved to Navi for the IGP, their APUs will kick all sorts of ass. People will be enjoying affordable 6c/12t APUs with great performance and looking back on the early 2010s Intel dominated era simply shaking their heads.
Click to expand...

Well,I had a think about - AMD was talking about a second 14NM generation,and the single leak of Raven Ridge was a ULV CPU with a 2000 series moniker and is 4C/8T:

https://videocardz.com/71232/amd-ryzen-5-2500u-with-radeon-vega-graphics-spotted

So methinks it must be that second 14NM generation with some core improvements,and all the new AMD APUs in the past tended to debut with improved cores - examples being Llano which had the last iteration of the K10 and Trinity which was the first example of the Piledriver core to be demoed.

Providing it still has the same amount of cache,and can clock around the same as the current Ryzen CPUs,the CPU side might actually be reasonably decent,since it won't have the CCX latency issues.
 
DragonQ said:
Depends how soon their second generation 14 nm process, whatever that is, is ready. Still won't be more than 4c/8t though.
Click to expand...

It won't be,but I kind of said that ages ago too - its the whole reason why they have the CCX design,so its easier for them to add custom units. The thing is AMD is most likely going to have a better IGP too than Intel,so if these come to desktop these should eventually replace the Ryzen 3 and lower end Ryzen 5 CPUs,so if anything AMD will be in a better place than they are now.

If you look at the A12 9800,I am actually shocked how reasonable the performance is:
https://www.youtube.com/watch?v=fw11VXVOZ3k

That is with a relatively anaemic CPU section,and limited memory bandwidth. Now add two more generations of IGP improvement,a better memory controller and a significantly better CPU section. If they do come to desktop and are priced well,it will be a compelling product,especially if AMD can get them into prebuilt PCs.
 
You must log in or register to reply here.
Back
Top Bottom