Poll: A few questions regarding Threadripper vs i9

[Enderman]

10 minutes ago, He_162 said:

If you put both CPU's at the same clock speed and measure single core performance, you're getting the IPC. That's benchmarks.

Now if you take the overall score and divide it by the cores, you can get a similar metric for IPC, although not as accurate.
i9-7900x = 2167 in cinebench. / 10 = 216.7 (@3.3ghz)

TR1920x= 2431 in cinebench. / 12 = 202.6 (@3.5ghz)

 

Threadripper was clocked slightly higher across all cores, but it seems like it's within 11% of X299 / Skylake-X's IPC if it's anything close to accurate.

And you're literally telling people not to compare products when one is clearly better price to performance, and you can figure out things like relative IPC, overall IPC, and how well it does in the cherry picked benchmarks, which is silly, conversations should be had period, and it's not something we should just "stop and accept whatever the reviewers say".

I was told Ryzen sucks at gaming until I bought it, and found out not only was it smoother, but no one mentioned the fact that it doesn't matter what the max framerate is if you don't have a GTX 1080 Ti, or 240hz monitor.

1) no, you don't divide by the number of cores, you run a benchmark that is single threaded, like R15 single thread.

 

2) Threadripper hasn't even been released yet. You have no idea if the rumours of the price are accurate at all, or if intel plans to reduce MSRP on i9s, or if retailers will sell at whatever price they want. Since they don't exist in stores yet, no you cannot say anything about price performance.

 

3) Whoever told you ryzen sucks for gaming was completely wrong, it is fine for gaming. What is not wrong is that on average a 7700k will get higher frame rates, since only a minority of games use more than 8 threads.

 

4) in competitive games, yes the max framerate matters even if it is above your monitor's refresh rate.

 It will reduce the delay between input and visible display on the screen.

[HKZeroFive]

Just now, He_162 said:

No clue, optimism I guess?

It clearly doesn't.

Or delusional... there's a very thin line.

[porina]

Random thoughts on this thread:

 

Skylake-X problems seem comparable to existing 115x CPUs. I don't think it is particularly worse, and serious overclockers will delid anyway.

 

X299 problems similarly are overstated. Heat problems are only due to heavily overclocked high load conditions.

 

Zen can have more IPC than Intel... if you compare two threads per core. Intel low thread performance advantage is mostly from higher clock potential.

 

I wouldn't count on TR OC'ing more than already seen. With adequate cooling, you might get all cores around 4 GHz or so, which should eat threaded work for breakfast. The clue is in the name!

 

My value judgement is not on cores or IPC or clocks in isolation, but overall throughput for a given task.

 

My next system build will be X299 based, waiting on a micro-ATX board to come out that looks nice. Will get a 7800X because for this build, I do want to optimise for gaming, thus prefer the higher clock over the number of cores but still having more than 4 cores I do currently. Why not Ryzen? I already have a 1700, and with a low voltage OC it was only reaching around 3.7 GHz before hitting the voltage wall. I haven't tried a yolo OC on it yet. Assuming Skylake-X runs similarly to Skylake-S, 4.2 should be a doddle at low voltage with option of perhaps 4.7 ball park possible if I decide to throw enough cooling at it.

 

It has been bought to my attention that the L3 cache on Skylake-X runs at around half clock compared Skylake-S. Under some HPC like scenarios this has resulted in lower than expected performance, and it may require retuning of code to work better with the cache rebalance.

[Skanky Sylveon]

36 minutes ago, Enderman said:

4) in competitive games, yes the max framerate matters even if it is above your monitor's refresh rate.

 It will reduce the delay between input and visible display on the screen.

As someone that has done some hobbyist game creation, you are correct.

Even if a game's physics and movement system is unbound to the frame rate, Input is a lot of times still frame rate bound.

Thus a faster frame rate will cause less delay.

 

Edit: This is a huge oversimplification, however if I were to explain in detail, it would be multiple paragraphs long.

[HKZeroFive]

1 hour ago, Cinnabar Sonar said:

Better single threaded performance if and after Skylake x thermal throttles.  

While this is an issue already, the higher core Skylake x chips that are coming out may have this problem even more.

Skylake x has better single threaded performance when running at full speed, that is clear.

Even at 4GHz, you still have a significant amount of headroom with Skylake-X as opposed to Ryzen/Threadripper which is limited by something process-related. At the same clocks, Skylake-X uses less power than Ryzen and combined with the fact that SKL-X's IPC is higher, it's a fair assumption to say single-threaded performance will still be better than Threadripper... even before it starts to thermal throttle. It's only once you start overclocking that things start to get messy.

[Skanky Sylveon]

16 hours ago, HKZeroFive said:

Even at 4GHz, you still have a significant amount of headroom with Skylake-X as opposed to Ryzen/Threadripper which is limited by something process-related. At the same clocks, Skylake-X uses less power than Ryzen and combined with the fact that SKL-X's IPC is higher, it's a fair assumption to say single-threaded performance will still be better than Threadripper... even before it starts to thermal throttle. It's only once you start overclocking that things start to get messy.

AMD measures TDP differently then Intel, it's not necessarily an apples to apples comparison, but yeah, the higher TDP Intel chips seem to use less power.

 

I said after thermal throttling, not before.  Skylake x clearly has better single threaded performance before throttling.

 

Edit: I meant to say TDP where I said IPC.

[He_162]

11 hours ago, Enderman said:

1) no, you don't divide by the number of cores, you run a benchmark that is single threaded, like R15 single thread.

 

2) Threadripper hasn't even been released yet. You have no idea if the rumours of the price are accurate at all, or if intel plans to reduce MSRP on i9s, or if retailers will sell at whatever price they want. Since they don't exist in stores yet, no you cannot say anything about price performance.

 

3) Whoever told you ryzen sucks for gaming was completely wrong, it is fine for gaming. What is not wrong is that on average a 7700k will get higher frame rates, since only a minority of games use more than 8 threads.

 

4) in competitive games, yes the max framerate matters even if it is above your monitor's refresh rate.

 It will reduce the delay between input and visible display on the screen.

1. I understand this, I explained it was not accurate before going on with it.

2. All specs given were REAL BENCHMARKS, and ACTUAL INFORMATION FROM AND AND INTEL
 

3. The i7-7700k only gets higher frames in 1080p, and if you have something better than a GTX 1060, so people getting a Ryzen 6 core, and a budget GPU like the GTX 1060 / RX-480 will not see a difference, except for the FACT that Ryzen gets better frametimes due to more cores, and threads, and a smoother delivery of frames (on games that support 8 or more threads)

4. Let's not get off topic into nitty gritty details about how high frames on a crap monitor is bigger than the issue of having a crap monitor.

[Enderman]

1 minute ago, He_162 said:

2. All specs given were REAL BENCHMARKS, and ACTUAL INFORMATION FROM AND AND INTEL
 

3. The i7-7700k only gets higher frames in 1080p, and if you have something better than a GTX 1060, so people getting a Ryzen 6 core, and a budget GPU like the GTX 1060 / RX-480 will not see a difference, except for the FACT that Ryzen gets better frametimes due to more cores, and threads, and a smoother delivery of frames (on games that support 8 or more threads)

4. Let's not get off topic into nitty gritty details about how high frames on a crap monitor is bigger than the issue of having a crap monitor.

2) I haven't seen any real benchmarks of threadripper from any reviewer. I haven't seen any Threadripper CPUs in any store.

Where do you live that you are in the future and this stuff exists already??

 

3) no, 4k too, but obviously you wouldn't play 4k with a 1060.

Ryzen gets better frame times when streamign and doing other intensive tasks in the background, don't confuse that with "just gaming"

 

4) Just because a monitor is 60hz or 144hz does not mean it's crap. It's still beneficial to get 300fps+ in competitive games.

Try it yourself, you should notice a difference.

[He_162]

3 minutes ago, Enderman said:

2) I haven't seen any real benchmarks of threadripper from any reviewer. I haven't seen any Threadripper CPUs in any store.

Where do you live that you are in the future and this stuff exists already??

 

3) no, 4k too, but obviously you wouldn't play 4k with a 1060.

Ryzen gets better frame times when streamign and doing other intensive tasks in the background, don't confuse that with "just gaming"

 

4) Just because a monitor is 60hz or 144hz does not mean it's crap. It's still beneficial to get 300fps+ in competitive games.

Try it yourself, you should notice a difference.

1. Are you not listening? I just told you the benchmarks were from AMD and Intel, they are real benchmarks, done in real time, with their processors, and every time you claim we need to wait for third party reviewers, they got the same, or better scores, so please, go do some googling like I asked before you embarrass yourself.

2. I just can't see anyone spending more money for unnoticeable amounts of FPS.

 

3. I have, on a 144hz monitor I noticed much better gameplay, and it felt smoother vs a 60hz monitor, and when I put a GT-720 in there to test what it felt like at just 60FPS, it was so similar to running my monitor at 60hz and 200+ FPS that I have decided you must not have done this yourself, or you'd know...

[Damascus]

Lol, love the last question, what do you want, the most expensive hedt cpu or its price to performance opponent?"

 

I'll take the i9 if it's 7900x or higher and I'm not paying

[Enderman]

2 minutes ago, He_162 said:

1. Are you not listening? I just told you the benchmarks were from AMD and Intel, they are real benchmarks, done in real time, with their processors, and every time you claim we need to wait for third party reviewers, they got the same, or better scores, so please, go do some googling like I asked before you embarrass yourself.

2. I just can't see anyone spending more money for unnoticeable amounts of FPS.

 

3. I have, on a 144hz monitor I noticed much better gameplay, and it felt smoother vs a 60hz monitor, and when I put a GT-720 in there to test what it felt like at just 60FPS, it was so similar to running my monitor at 60hz and 200+ FPS that I have decided you must not have done this yourself, or you'd know...

1) lol, using benchmarks from AMD  Have you not seen all the previous BS stuff they're released?

Linus said so himself, never trust the benchmarks from the company itself.

They're always been biased in some way, whether it is from nvidia or AMD.

 

2) The 7700k is the same price as 8 core Ryzen...

It's not unnoticeable either, go check out the benchmarks.

 

3) Take your game and lock it at 144fps, then unlock it and go to 300fps or more, you will see a difference.

Everyone knows that games are more responsive when running at a high refresh rate regardless of what your monitor's refresh rate is.

[jjohnthedon1]

13 hours ago, Enderman said:

Why would you ever compare a throttling CPU to a non-throttling one??

It should never throttle unless you make very bad buying decisions.

 

"hmm i wonder if my honda goes faster than ferrari with no wheels..."

Because the current 10 core throttles with a h115 that's why

its pegged at 95 degrees with a top aio cooler 

if it had a better tim or solder it would be a lot better 

 

stop treating op like a idiot he knows what he is on about 

[Enderman]

5 minutes ago, jjohnthedon1 said:

Because the current 10 core throttles with a h115 that's why

its pegged at 95 degrees with a top aio cooler 

if it had a better tim or solder it would be a lot better 

 

stop treating op like a idiot he knows what he is on about 

Solution: don't overclock as much as toms hardware tried.

 

Obviously if you increase a 140W CPU to 300W with 1.4 core voltage it's gonna get hot.

The trick is to NOT use 1.4v when using an AIO.

 

Quote

Realistically, 4.5 GHz should be achievable with an all-in-one liquid cooler. 

 

[jjohnthedon1]

2 minutes ago, Enderman said:

Solution: don't overclock as much as toms hardware tried.

 

Obviously if you increase a 140W CPU to 300W with 1.4 core voltage it's gonna get hot.

The trick is to NOT use 1.4v when using an AIO.

I'm running 1.45v with a h110 with temps in the 50's oh wait my chip is soldered

 

the video I saw was at 1.29 

[jjohnthedon1]

3 minutes ago, Enderman said:

Solution: don't overclock as much as toms hardware tried.

 

Obviously if you increase a 140W CPU to 300W with 1.4 core voltage it's gonna get hot.

The trick is to NOT use 1.4v when using an AIO.

Also the point of ops post 

will thread ripper perform better if it can oc higher due to not being thermally limited 

 

answer absolutely 

[Enderman]

2 minutes ago, jjohnthedon1 said:

I'm running 1.45v with a h110 with temps in the 50's oh wait my chip is soldered

 

the video I saw was at 1.29 

Ok so? You don't have an intel CPU...

If you make a 6950X run at 1.4v then obviously an AIO will also not be able to keep up, regardless if it is soldered or not. (fyi it is)

Why are you even arguing this? Everyone knows that if your temps are getting too high, you don't keep pushing the overclock.

 

An AIO is fine for a stock or slightly OCd 7900X.

http://www.tomshardware.com/reviews/intel-core-i9-7900x-skylake-x,5092-10.html

 

1 minute ago, jjohnthedon1 said:

Also the point of ops post 

will thread ripper perform better if it can oc higher due to not being thermally limited 

 

answer absolutely 

You don't know how well threadripper overclocks.

You also don't know much much worse it's single core performance is, and if an overclock will compensate for that.

Answer: we know absolutely nothing.

[jjohnthedon1]

6 minutes ago, Enderman said:

Ok so? You don't have an intel CPU...

If you make a 6950X run at 1.4v then obviously an AIO will also not be able to keep up, regardless if it is soldered or not. (fyi it is)

Why are you even arguing this? Everyone knows that if your temps are getting too high, you don't keep pushing the overclock.

 

An AIO is fine for a stock or slightly OCd 7900X.

http://www.tomshardware.com/reviews/intel-core-i9-7900x-skylake-x,5092-10.html

 

You don't know how well threadripper overclocks.

You also don't know much much worse it's single core performance is, and if an overclock will compensate for that.

Answer: we know absolutely nothing.

We know it's 2 ryzen chips on one die ? 

We know that ryzen tops out at 4 ghz 

 

well whats the pont in delidding then 

trutu is the x299 cpus would be better soldered 

 

solder is a lot better than intell current cooling solution we all know that 

end of story 

[Enderman]

8 minutes ago, jjohnthedon1 said:

We know it's 2 ryzen chips on one die ? 

We know that ryzen tops out at 4 ghz 

 

well whats the pont in delidding then 

trutu is the x299 cpus would be better soldered 

 

solder is a lot better than intell current cooling solution we all know that 

end of story 

When you have two CPUs meshed together I highly doubt the overclocking will be the same as with one CPU, especially with double the heat.

 

Is ryzen at 4GHz better IPC than skylake/kaby lake stock? Nope.

 

Yes it would be better if i9s were soldered, not disagreeing with that at all.

I guess it makes the delidding process easier though for those that want to do it.

 

My point is that you need to overclock a CPU to the capacity of cooling you have, not OVER it.

Taking an i5 and putting 1.8v into it and then saying "oh no its throttling on my custom loop" is not the proper way of using a CPU.

[jjohnthedon1]

9 minutes ago, Enderman said:

When you have two CPUs meshed together I highly doubt the overclocking will be the same as with one CPU, especially with double the heat.

 

Is ryzen at 4GHz better IPC than skylake/kaby lake stock? Nope.

 

Yes it would be better if i9s were soldered, not disagreeing with that at all.

I guess it makes the delidding process easier though for those that want to do it.

 

My point is that you need to overclock a CPU to the capacity of cooling you have, not OVER it.

Taking an i5 and putting 1.8v into it and then saying "oh no its throttling on my custom loop" is not the proper way of using a CPU.

Well they would oc better and safer with no need to delid at all if they were soldered 

 

I didn't even price up a 7700k system because it wasn't soldered 

[Enderman]

1 minute ago, jjohnthedon1 said:

Well they would oc better and safer with no need to delid at all if they were soldered 

 

I didn't even price up a 7700k system because it wasn't soldered 

Ok, but you understand that if you leave it at stock or overclock reasonably, the 7900X will not throttle, right?

Whether Threadripper will be able to have better IPC with similar cooling is currently impossible to know.

 

And no, never trust benchmarks from the manufacturer.

Spoiler

Remember this?

lol it was hilarious when all reviewers' benchmarks showed the opposite.

 

[jjohnthedon1]

9 minutes ago, Enderman said:

Ok, but you understand that if you leave it at stock or overclock reasonably, the 7900X will not throttle, right?

Whether Threadripper will be able to have better IPC with similar cooling is currently impossible to know.

 

And no, never trust benchmarks from the manufacturer.

  Reveal hidden contents

Remember this?

lol it was hilarious when all reviewers' benchmarks showed the opposite.

 

I get that 

but it wouldnt throttle oced if it were soldered 

 

[TahoeDust]

Threadripper is probably going to be a multicore performance monster...but it's single core performance will still be garbage.  My 7820x at its 4.7GHz 1.24v 24/7 overclock has single core performance almost 20% better than the highest AMD score ever recorded in the Cinebench thread.

[Enderman]

4 minutes ago, jjohnthedon1 said:

I get that 

but it wouldnt throttle oced if it were soldered 

 

It would, it would just take more voltage or a worse cooler to get there.

You can make any CPU throttle regardless of soldering. You just need enough voltage.

 

What you're trying to say is that with the same cooling, it would overclock more before throttling if it was soldered.

[jjohnthedon1]

5 minutes ago, Enderman said:

It would, it would just take more voltage or a worse cooler to get there.

You can make any CPU throttle regardless of soldering. You just need enough voltage.

 

What you're trying to say is that with the same cooling, it would overclock more before throttling if it was soldered.

Evidently Sherlock 

 

[Enderman]

1 minute ago, jjohnthedon1 said:

Evidently Sherlock 

 

Ok great, now we're on the same page.