TomsHardware: Core i7-3970X Extreme Review: Can It Stomp An Eight-Core Xeon?

[inf64]

We have been over that 10x already. Faster than and slower than use different bases for calculating numbers. 3970x is 21.1% faster than 8350(we use 8350's result as a base-which is 116;so 91.5/116=0.788 => 100-78.8=21.1). On the other hand 8350 is 26.8% slower than 3970x(we use a 3970x's results as a base-which is 91.5;so 116/91.5=1.26 or 26.8%). It's very easy to understand but yet somehow very complicated to some users here.

 

[TemjinGold]

We have been over that 10x already. Faster than and slower than use different bases for calculating numbers. 3970x is 21.1% faster than 8350(we use 8350's result as a base-which is 116;so 91.5/116=0.788 => 100-78.8=21.1). On the other hand 8350 is 26.8% slower than 3970x(we use a 3970x's results as a base-which is 91.5;so 116/91.5=1.26 or 26.8%). It's very easy to understand but yet somehow very complicated to some users here.

Calm down. We all get that you need a different base for faster and slower. The problem is you're wrong because you are using the wrong base both times. I believe the other guy tried to explain it but to help you understand, pretend the Intel took 50s and the AMD took 100s. We would all agree, the Intel is 100% faster than the AMD in this case (since it's twice as fast), right? But here's what happens if you use your math:

50 / 100 = .5 => 1 - .5 = .5.

That puts the Intel at 50% faster than the AMD. How do you explain that?

While this doesn't work for FASTER, this formula is fine for SLOWER (because the AMD is 50% slower than the Intel.) Do you get it now?

 

[moonbogg]

Haswell-DT vs SB-E/IB-E in singlethreaded gonna be an utter nightmare for LGA2011 owners. And SB-E atleast might easily lose more than it wins in MT.

No it won't, because I will sell my platform and buy a Haswell if it turns out to be a good benefit for gaming needs. Nightmare = avoided.

 

[AtenRa]

Calm down. We all get that you need a different base for faster and slower. The problem is you're wrong because you are using the wrong base both times. I believe the other guy tried to explain it but to help you understand, pretend the Intel took 50s and the AMD took 100s. We would all agree, the Intel is 100% faster than the AMD in this case (since it's twice as fast), right? But here's what happens if you use your math:

50 / 100 = .5 => 1 - .5 = .5.

That puts the Intel at 50% faster than the AMD. How do you explain that?

While this doesn't work for FASTER, this formula is fine for SLOWER (because the AMD is 50% slower than the Intel.) Do you get it now?

The math is correct, the interpretation is wrong,

Core i7 3970 = 50 Seconds
FX8320 = 100 Seconds

Because we measure TIME (Seconds),


50 / 100 = .5 => 1 - .5 = .5.

.5 means HALF the TIME in (seconds) or,
It takes 50% OF the TIME of the 100 seconds.

Therefore,
That means that Core i7 3970 is 50% FASTER because it will need 50% of the time to finish the same work.

Also

It will do the work at HALF THE TIME = Core i7 3970 is 100% faster. Here you can also say that Core i7 3970 is 2x faster.

Therefore,
FX8350 will need DOUBLE the TIME that means 2x SLOWER or it will be 100% slower because it needs 100% MORE TIME over the 50 Seconds that Core i7 3970 does to finish the same work

 

[inf64]

Exactly AtenRa. I cannot believe we are still debating this basic stuff lol...

 

[nitromullet]

FX8350 will need DOUBLE the TIME that means 2x SLOWER or it will be 100% slower because it needs 100% MORE TIME over the 50 Seconds that Core i7 3970 does to finish the same work

This is incorrect. The slower processor cannot be 100% slower and still be doing any work. If someone offers you a 100% discount you aren't expected to pay half are you?

...and these two statements contradict each other. The first statement is false, the second is correct.

Therefore,
That means that Core i7 3970 is 50% FASTER because it will need 50% of the time to finish the same work.

Also

It will do the work at HALF THE TIME = Core i7 3970 is 100% faster. Here you can also say that Core i7 3970 is 2x faster.

 

[inf64]

3970x(50s) is 50% faster than 8350(100s) since 50% of 100s is 50s and 100-50=50... Similarly 8350(100s) is 2x slower than 3790(50s) since 100/50=2 or exactly slower for a full amount of time 3970x takes to finish the job. It is 100% slower than 3970 which means you take the time of 3970x and multiply it by 2. Wow this is almost as hard as differential calculus!

 

[AtenRa]

...and these two statements contradict each other. The first statement is false, the second is correct.

hmmm, it is easy to make the mistake and it seams I have also done it because i have said that Core i7 is 100% faster.

It is not the Core i7 that is 100% faster, it is the 100 seconds that are 100% MORE than the 50 seconds.

The Core i7 3970 is 50% faster because it needs HALF the Time of the 100 seconds or 50% LESS TIME (50% of the 100seconds is 50 seconds)

Now, if you say that Core i7 is 100% faster because 100 seconds are 100% more than 50 seconds its wrong.

I will correct it thanks

 

[AtenRa]

Exactly AtenRa. I cannot believe we are still debating this basic stuff lol...

It is easy to make a mistake here because faster is LESS TIME and not MORE fps

 

[Gideon]

hmmm, it is easy to make the mistake and it seams I have also done it because i have said that Core i7 is 100% faster.

It is not the Core i7 that is 100% faster, it is the 100 seconds that are 100% MORE than the 50 seconds.

The Core i7 3970 is 50% faster because it needs HALF the Time of the 100 seconds or 50% LESS TIME (50% of the 100seconds is 50 seconds)

Now, if you say that Core i7 is 100% faster because 100 seconds are 100% more than 50 seconds its wrong.

I will correct it thanks

When talking about precentage, one has to be very clear, what is takne as the 100% value.

This post states it quite clearly: http://math.stackexchange.com/questions/186730/calculate-x-slower-faster

Therefore, if i7 completes something in 50 seconds and FX completes the same task in 100 seconds then:

Core i7 is 100% faster than FX AND FX is 50% slower than Core i7 at the same time.
This is so, because in the first case we take the 50 seconds as the base value, in the other case we take the 100 seconds.

 

[Gideon]

This is also why AMD is in such a bad situation single threaded performance wise (following numbers are hyphothetical, but close enough to get the idea):

If Ivy Bridge is 70% faster than Piledriver in single threaded apps and Haswell is 10% faster than Ivy, then Haswell IS NOT 80% faster than Piledriver. It's actually 87% faster (10% from the 170% value). Therefore even if Haswell only improves a "measly" 10% over Ivy Bridge, Steamroller would need to be 17% faster than Piledriver just to keep the Status Quo!

 

[antihelten]

This is also why AMD is in such a bad situation single threaded performance wise (following numbers are hyphothetical, but close enough to get the idea):

If Ivy Bridge is 70% faster than Piledriver in single threaded apps and Haswell is 10% faster than Ivy, then Haswell IS NOT 80% faster than Piledriver. It's actually 87% faster (10% from the 170% value). Therefore even if Haswell only improves a "measly" 10% over Ivy Bridge, Steamroller would need to be 17% faster than Piledriver just to keep the Status Quo!

No, Steamroller would only need to be 10% faster to keep status quo since 187%/110% = 170%

 

[Idontcare]

No, Steamroller would only need to be 10% faster to keep status quo since 187%/110% = 170%

Steamroller was never billed as being 10% faster, it was billed as being 10% more efficient in terms of performance/watt.

The assumption there is, of course, that you can just juice up the clocks and negate the power savings and turn that efficiency into higher performance...but making assumptions regarding power and clockspeed with AMD cpus lately is a bit of gamble.

 

[nitromullet]

It is 100% slower than 3970

No. 100% is all of something. To be 100% slower means the AMD chip would have to be doing 0% of the work in the same amount of time, and 0% of anything is nothing. Is the AMD chip doing nothing?

In the example of the AMD chip doing half the work of the Intel chip in the same amount of time:

• the AMD chip is 50% slower (doing half the work) than the Intel chip
• the Intel chip is 100% faster (doing twice the work) than the AMD chip.

Percentage is relative to what is being compared.

 

[AtenRa]

Core i7 is 100% faster than FX AND FX is 50% slower than Core i7 at the same time.

In the example of the AMD chip doing half the work of the Intel chip in the same amount of time:

Well, i begging to believe that people are trolling,

It was established from the start that we are measuring TIME, we dont have the same amount of TIME, we have LESS TIME.

We do not count work done at the same time but how much LESS time it takes to complete the same task or work.

We can have more than 100% SLOWER TIME because our base is not ZERO(0) but 100seconds or 50seconds.

CPU A needs 50 SECONDS to finish the task
CPU B needs 100 SECONDS to finish the task
CPU C needs 150 SECONDS to finish the task

CPU C is 200% SLOWER than CPU A, it takes 200% more than the 50 SECONDS CPU A needs, for the CPU C to finish the same job.
50 + 200% = 150

Again, we measure SECONDS not work done at the same amount of time but the opposite.

Sorry for the off topic, that was my last post about this subject.

 

[TemjinGold]

Perhaps English comprehension is the issue here (no offense meant). 100% slower means doing ZERO work, just like 100% discount means FREE. 200% slower is impossible unless the AMD cpu is doing NEGATIVE work.

Note that 2x the speed and 2x faster are not the same thing. The former refers to 200% while the latter refers to 300%. If measuring time is confusing, then measure work. In the 50/100 example, the Intel processor would do twice the work in the same amount of time OR the same work in half the time, that is why it is 100% faster.

In any case, what bothers me isn't that there are folks on both sides equally convinced the other side is wrong. What bothers me is we can't have a simple intellectual debate without all the condescending remarks about how the other side must be composed of idiots.

 

[inf64]

No. 100% is all of something. To be 100% slower means the AMD chip would have to be doing 0% of the work in the same amount of time, and 0% of anything is nothing. Is the AMD chip doing nothing?

In the example of the AMD chip doing half the work of the Intel chip in the same amount of time:

• the AMD chip is 50% slower (doing half the work) than the Intel chip
• the Intel chip is 100% faster (doing twice the work) than the AMD chip.

Percentage is relative to what is being compared.

100% slower than 3970x which scores 50s is 100s man. What can't you understand? When we say slower it means you ADD UP time,my Lord.... I cannot believe you guys still don't get it. We measure time and slower means LONGER time. Not less time as you people think. I'm amazed,truly .

If we would to measure "points"(like Cinebench),it would be different. 2600K scores 6.8pts,1100T scores 5.9pts. 1100T is 5.9/6.8=0.867 or 13.3% slower than 2600K. On the other hand 2600K is 6.8/5.9=1.15 or 15% faster than 1100T.

 

[nitromullet]

100% slower than 3970x which scores 50s is 100s man. What can't you understand? When we say slower it means you ADD UP time,my Lord.... I cannot believe you guys still don't get it. We measure time and slower means LONGER time. Not less time as you people think. I'm amazed,truly .

100% slower than something would mean it is doing no work at all.

 

[inf64]

100% slower than something would mean it is doing no work at all.

No. It means it does the job in 2x longer time. We measure time. Time. Need I write it one more time? Oh sorry,I already did
100% slower is actually 2x slower than the base you are using. Which is 50s in the (hypothetical) example original poster used.

 

[AtenRa]

No. It means it does the job in 2x longer time. We measure time. Time. Need I write it one more time? Oh sorry,I already did
100% slower is actually 2x slower than the base you are using. Which is 50s in the (hypothetical) example original poster used.

heh, Resistance is Futile

 

[Idontcare]

No. It means it does the job in 2x longer time. We measure time. Time. Need I write it one more time? Oh sorry,I already did
100% slower is actually 2x slower than the base you are using. Which is 50s in the (hypothetical) example original poster used.

All I know is that according to my wife I'm 50% shorter than I claim to be (manhood-wise), 50% slower than she'd like but I finish 50% faster than she prefers :|

Now would someone kindly tell me what base she is using for those measurements? And if the answer is something like "her ex-boyfriend Ron, who she still probably sees from time to time" then, well, then yeah you're probably right





j/k, just trying to lighten up the mood, unless your name is Ron, then I gots a beef with you!

 

[nitromullet]

No. It means it does the job in 2x longer time. We measure time. Time. Need I write it one more time? Oh sorry,I already did
100% slower is actually 2x slower than the base you are using. Which is 50s in the (hypothetical) example original poster used.

The Intel chip is twice as fast as the AMD chip, so it is 100% faster.
The AMD chip is half as fast as the Intel chip, so it is 50% slower.

Again, the AMD chip cannot be 100% slower if it is doing any work. It is not possible. A rock is 100% slower than the Intel chip, but not even a K6-II is that slow.

 

[TemjinGold]

The Intel chip is twice as fast as the AMD chip, so it is 100% faster.
The AMD chip is half as fast as the Intel chip, so it is 50% slower.

Again, the AMD chip cannot be 100% slower if it is doing any work. It is not possible. A rock is 100% slower than the Intel chip, but not even a K6-II is that slow.

Dude let's just forget it. You can't make an argument when they think we don't know math while we think they don't understand English...

 

[inf64]

@nitromullet
I give up you don't get it. I'm sorry I tried my best.

@ IDC

Man,that post made my day,seriously .

 

[Makaveli]

It just goes to show how important word choice is when writing!