AMD THREADRIPPER VS INTEL SKYLAKE X

[zornyan]

Those games are newer releases are they not? Games that will use more cores more effectively. Fallout 4, arma, dayz are poorly optimized biased for single thread as pointed out before on multiple occasions.

Those games that run better largely run better because of the clock advantage that Intel has over ryzen but like I said, ryzen is a few months old on it's first iteration, Intel have been refining theirs for many years. Clock for clock ryzen holds it's own and is cheaper than intel equivalents.

In your opinion will games going forward continue to utilize fewer cores or or more cores? Are intel moving towards more cores or stopping at 4 cores?

amdahl's law kind of goes against this whole 'games will run better in the future on more cores' it is extremely extremely difficult to parrallize multiple threads in games, even in games that currently appear to be well multi threaded such as battlefield 1, they are still dominates by 1 main thread that requires higher ipc/clock speeds.

as we could both agree, technically if everything did run parallel then a 4ghz 4 core should be equal to a 2ghz 8 core in some of these games, and yet we see the 8 core will run much lower fps and struggle where as the 4 core will be much more efficient and give better performance.

I'm not saying in years to come it won't be like that, but 4 cores is more than enough for now and for years to come, I would imagine a 6 core becoming necessary around 2018/19 and 8 cores maybe 2/3 years after that, it's not as easy for them to just program "more cores"

even look at a game like the witcher 3, it utilisied 8 threads, yet the 1800x is some 40+ fps lowercompared to the 7700k, even a 6900k which is close in IPC to kabylake is slower than the 7700k, when all facts tell us it *should* be quicker.

 

[zornyan]

My personal comp is indeed 980 however only a single now. And the systems on the Ryzen and 7700k are for machines I have built for friends. Some are 1080 and others 1080Ti.

There are a few videos on YouTube showing the same performance as I've stated and these are on latest agesa updates.

I've tested this on 6 different systems whilst setting up with same windows 10 build as it's an ISO I made to make sure it setup right.

My 4790k I still use is about 5% slower to compare.

Resolution I am getting around 5fps different at 1080 and it reducing to 1-2fps different at 1440p and 4K

so we can disregard your "my 6700k at 4.8ghz isn't any better than 6800k at 4ghz" comment then because you're massively gpu bound so the cpu is totally irrelevant?

 

[Doogles]

I'm thinking the next 2500k will be a 8 core Skylake-X chip, as in that if you get one when they launch it will do you for the next 6 years.

 

[Curlyriff]

so we can disregard your "my 6700k at 4.8ghz isn't any better than 6800k at 4ghz" comment then because you're massively gpu bound so the cpu is totally irrelevant?

That wasn't me. I've stated which GPUs I've used with both Ryzen and Intel CPU for friends builds.

They weee benched and overclocked. All the Intel systems are using the Noctua D15, the Ryzen are using the stock cooler. All had 32GB RAM because most of them play games, stream and watch YouTube etc at same time. Only reason there is even two Intel systems I built is because they didn't want to go AMD and that's fair enough.

I don't mind what anyone picks but at least be factual with figures and performance.

 

[Doobedoo]

amdahl's law kind of goes against this whole 'games will run better in the future on more cores' it is extremely extremely difficult to parrallize multiple threads in games, even in games that currently appear to be well multi threaded such as battlefield 1, they are still dominates by 1 main thread that requires higher ipc/clock speeds.

as we could both agree, technically if everything did run parallel then a 4ghz 4 core should be equal to a 2ghz 8 core in some of these games, and yet we see the 8 core will run much lower fps and struggle where as the 4 core will be much more efficient and give better performance.

I'm not saying in years to come it won't be like that, but 4 cores is more than enough for now and for years to come, I would imagine a 6 core becoming necessary around 2018/19 and 8 cores maybe 2/3 years after that, it's not as easy for them to just program "more cores"

even look at a game like the witcher 3, it utilisied 8 threads, yet the 1800x is some 40+ fps lowercompared to the 7700k, even a 6900k which is close in IPC to kabylake is slower than the 7700k, when all facts tell us it *should* be quicker.

But like I said Intel have a big clock advantage over ryzen yet that clock advantage doesn't always translate into FPS. We're seeing a trend as demonstrated by adored tv where the newer games seem to be able to utilize ryzen better than older games. This is the future of which AMD will be apart of. Ryzen with all things taken into consideration, is a very capable cpu and is a massive step in the right direction for amd. It isn't for you, well so be it you pay your money for whatever take your fancy, it doesn't make ryzen a bad choice it just means that perhaps others have different priority's when it comes to purchasing a new cpu.

 

[smilingcrow]

Another article from 3rd of April -_-

Have a look here last week "news" not from 3 months ago, why is irrelevant to post anything older than June.....

Here is a good example how unrealistic all benchmarks are from previous months today.

Here is one using the 1700X with GTX1080 same game and scene from March 3rd to run a comparison.

Check the perf difference between the "before the game update" 1600 (but with all windows, BIOS etc) and the 1700X perf from March......

Yes the 1700X runs at 1920x1080 and the 1600 at 2560x1080, but the latter is faster because of those updates. (on the before the game update benchmark).

On the post game update benchmark, as you see the performance of the $200 cheaper 1600 is in par with the 7700K @ 4.8Ghz+

And that money difference makes the argument 7700K + GTX1080 +Z270 or 1600 and GTX1080Ti +B350? Since they cost the same more or less.


Many here had the question in this very forum "what to buy as future proof system".
The answer isn't Intel 4 core CPU today.


And something newer comparing actually against intel CPUs albeit with a GTX970OC.

I am more interested in looking at the data averaged across a wide range of games as logged properly showing average and minimum frame rates and latencies.
A couple of snippets of YT clips just aren't enough to be considered objective.
Another issue when looking at games tested months apart is you need to make sure that they are using the same GPU driver and game versions.
Has any well regarded site done all this to show the true state of play?

 

[zornyan]

I am more interested in looking at the data averaged across a wide range of games as logged properly showing average and minimum frame rates and latencies.
A couple of snippets of YT clips just aren't enough to be considered objective.
Another issue when looking at games tested months apart is you need to make sure that they are using the same GPU driver and game versions.
Has any well regarded site done all this to show the true state of play?

gamernexus lists what drivers/Windows version/pretty much every tiny detail of their setup, and they mostly have it all stay the same, only changing things like gpu drivers when it's specifically needed for a new game/benchmark.

 

[Doobedoo]

I am more interested in looking at the data averaged across a wide range of games as logged properly showing average and minimum frame rates and latencies.
A couple of snippets of YT clips just aren't enough to be considered objective.
Another issue when looking at games tested months apart is you need to make sure that they are using the same GPU driver and game versions.
Has any well regarded site done all this to show the true state of play?

But who is doing this? Most of the benchmark reviews show an outdated suite of in game benchmarks rather than real world testing.

 

[Doobedoo]

gamernexus lists what drivers/Windows version/pretty much every tiny detail of their setup, and they mostly have it all stay the same, only changing things like gpu drivers when it's specifically needed for a new game/benchmark.

Have a look at this vid because he too shows ryzen on par with a 7700k including witcher 3 that doesn't show 40+ fps advantage that you said it does.

 

[zornyan]

Have a look at this vid because he too shows ryzen on par with a 7700k including witcher 3 that doesn't show 40+ fps advantage that you said it does.

which is funny because here the 1800x is 20-30 fps behind the 7700k at stock

https://youtu.be/HZPr-gNWdvI

the 7700k at stock speeds is still a decent amount ahead of the 1800x oc'd to 1800x, but compare stock for stock in novigrad and there's a clear 20+ fps lead.

before anyone says anything, both videos are 3 months old so it's a fair comparison.

 

[humbug]

What is it about Gamersnexus that makes it so attractive to those trying to sell Intel Uber Alles?

You can add 20% performance to all of their Ryzen results given they were all done with 2400Mhz Ram and we all know running Ram at that speed with Ryzen gimps its performance by 20% vs 3200Mhz.

This from TPU was across 20 Games, a selection 4 times bigger than Gamersnexus and yes that is a £200 Ryzen 5 at worst 7% slower than a £350 7700K

 

[Rroff]

and yes that is a £200 Ryzen 5 at worst 7% slower than a £350 7700K

But its also not really pulling away from a CPU that can be found quite a bit cheaper than £200 atleast in those images you posted so I'm not sure I'd be emphasis that aspect too much.

 

[humbug]

But its also not really pulling away from a CPU that can be found quite a bit cheaper than £200 atleast in those images you posted so I'm not sure I'd be emphasis that aspect too much.

The i5? don't make me post the one where Ryzen is 60% faster in Metro LL.

 

[Rroff]

Go ahead and post it those graphs are kind of meh where everything is within a couple or percent or so for the most part and doesn't really tell you anything about their performance capability or cost benefit.

 

[humbug]

Can't find it this late at night but rest assured it'll get posted again some time. ^^^^^
--------------

Impact of memory speed on gaming performance.

Most of these reviews were done on 2400Mhz RAM, the difference in performance between 2400Mhz and 3200Mhz as shown here is 15%.
If you add 15% to all those results suddenly the 7700K looks a lot less like the winner.

Another thing about all these reviews that are being banded about also make up their performance almost entirely from Ashes of the Singularity and Rise of the Tombraider performance which was at the time very much below par.
Both those games have since had a patch with has brought the performance at least in line with the 7700K, some are even reporting higher performance than that. In the case of Rise of the Tombraider Ryzen performance is up as much as 40%.

https://www.eteknix.com/memory-speed-large-impact-ryzen-performance/

Edit: Rise of the Tombraider pre and post Ryzen patch performance side by side, 1700X, performance after patch is consistently about 30 FPS higher.

 

[Rroff]

I'm surprised to see no one talking about the dismal storage IOPS (which can significantly slowdown OS booting and game load times, etc.) and poor random small read/write performance - going to be quite an issue if it bears out in enterprise environments where IOPS performance can be critical in systems utilising a lot of cores i.e. database management or backend services.

 

[FredFlint]

I'm surprised to see no one talking about the dismal storage IOPS (which can significantly slowdown OS booting and game load times, etc.) and poor random small read/write performance - going to be quite an issue if it bears out in enterprise environments where IOPS performance can be critical in systems utilising a lot of cores i.e. database management or backend services.

IOPS dont look that bad on my AMD PC:

Spoiler

I tested the Intel when the drive was new and not used as the boot drive, in the AMD test its the boot drive and 55% full.

 

[FredFlint]

No they all have a maximum of ## PCIe lanes, its not that Intel are not listing some, what, modesty? no they are listing all they have.

We actually have 52 but we don't like to tute our own horn so we just list 44.

It looks like Threadripper uses 48 lanes from the CPU to PCI-E slots and the rest for PCH (storage, USB...). My X99 has 40 from the CPU and 8 Gen 2 to the PCH and it was not advertised as a 44 Lane (8 Gen 2 = 4 Gen 3). Time will tell how mobo’s distribution the lanes but it looks like it will be 48 from the CPU(still more than Intel).
https://videocardz.com/70093/amd-ryzen-threadripper-with-64-pcie-lanes

 

[DragonQ]

I'm thinking the next 2500k will be a 8 core Skylake-X chip, as in that if you get one when they launch it will do you for the next 6 years.

I doubt it; it won't be £165 (and it was only that much because VAT went up to 20% just as it came out).

 

[zornyan]

What is it about Gamersnexus that makes it so attractive to those trying to sell Intel Uber Alles?

You can add 20% performance to all of their Ryzen results given they were all done with 2400Mhz Ram and we all know running Ram at that speed with Ryzen gimps its performance by 20% vs 3200Mhz.

This from TPU was across 20 Games, a selection 4 times bigger than Gamersnexus and yes that is a £200 Ryzen 5 at worst 7% slower than a £350 7700K

the gamernexus benchmarks were done with 2996mhz ram.