Would an I7920@stock bottleneck a 7970?

[crazekid362]

What about two 7970's in crossfire? Mind you, I have no idea how to overclock a cpu so that is out of the question.

 

[monsta]

Don't think it will , wait till they are released first to see how good they perform on different systems.

 

[refillable]

Maybe, but nobody know it yet. But I think it shouldn't with minimum overclocking. And also it depends on the programs.

 

[deadjon]

Theres always a first time for everything!
Once you overclock , you can never stop
Its a lot of fun and its VERY rewarding. Theres threads and stickys allover tomshardware and every other PC enthusiast forum for OCing

If you really dont want to OC, then I am going to has at a guess and say yes.
My Q6600@3.6 (back when it was nice to me) used to outperform a stock i7 920, clock for clock performance was only around 20 - 25% faster on the Nehalems.

Only a guess though, but if someone asked the same question about a Q6600@3.6, they would probably say yes.

 

[vitornob]

Hard to say about intel.

For sure the AMD FX line would bottleneck a 7970 crossfire, explanation below:

Based in recently Tomshardware System Builder Marathon, Dec. 2011: $1200 Enthusiast PC, and System Builder Marathon, Dec. 2011: System Value Compared show us that a FX-6100 bottlenecks A LOT even the 6950 Crossfire (mind you the 2*7970), bottlenecks enough to make an i5-2400 with a SINGLE 6870 be a better overall system for gaming.

PS: even a FX-6100 @4,5GHZ + 2*6950 isn't much faster than i5-2400 stock + single 6870 (1080p) in many games.
Except Metro 2033

 

[Wucked]

At stock yes, it probably will. But an overclocked 920 (3.6 - 4GHz+) is still a strong CPU. So if you're considering getting a couple of 7970's, you really need to be considering an overclock as well. Its not difficult to do, you just need a decent cooler and a couple of hours for stability testing.

 

[deadjon]

Hard to say about intel.

For sure the AMD FX line would bottleneck a 7970 crossfire

Okay here we go.

The baseline is Deneb (Phenom II)

Deneb 100% Per Clock
Kentsfield ~105% Per Clock
Bulldozer ~85% Per Clock
Nehalem ~ 130% Per Clock
Sandy Bridge ~150% Per Clock

If a Bulldozer 6 Core is Bottlnecking 2 6950s @ 45% Slower than the Nehalem Clock for Clock (Lets assume its not using 2 cores for games) We can take that number and compare the clocks.
(4.6/100) x 45 = 2.07 In this case we do 4.6 - 2.07 and we get a 2.53ghz Nehalem.
A Stock i7 920 is 2.66ghz, not much of an increase.

I conclude that a stock i7 920 will bottlneck even 2x 6950s - from that analysis.

The analysis is rudementry and quite crude, but it effectivley shows the whereabouts of a stock i7 920.

[EDIT] If the i7 was clocked at 4.0ghz its a whole different ballgame. It'll beat a stock 2600k at that speed and not bottleneck anything

 

[vitornob]

Okay here we go.

The baseline is Deneb (Phenom II)

Deneb 100% Per Clock
Kentsfield ~105% Per Clock
Bulldozer ~85% Per Clock
Nehalem ~ 130% Per Clock
Sandy Bridge ~150% Per Clock

If a Bulldozer 6 Core is Bottlnecking 2 6950s @ 45% Slower than the Nehalem Clock for Clock (Lets assume its not using 2 cores for games) We can take that number and compare the clocks.
(4.6/100) x 45 = 2.07 In this case we do 4.6 - 2.07 and we get a 2.53ghz Nehalem.
A Stock i7 920 is 2.66ghz, not much of an increase.

I conclude that a stock i7 920 will bottlneck even 2x 6950s - from that analysis.

The analysis is rudementry and quite crude, but it effectivley shows the whereabouts of a stock i7 920.

[EDIT] If the i7 was clocked at 4.0ghz its a whole different ballgame. It'll beat a stock 2600k at that speed and not bottleneck anything

I do agree that is a rudimentary analysis, but still it's a analysis with nice premisses. Sadly bulldozer is a completely different architecture where their 6 cores it's more likely to be a 3 core with hyperthreading (due to each 2 cores - 1 module - shares a lot of resources). But still from a preliminary analysis it's a good start.
Said that: +1 for you!

 

[deadjon]

I do agree that is a rudimentary analysis, but still it's a analysis with nice premisses. Sadly bulldozer is a completely different architecture where their 6 cores it's more likely to be a 3 core with hyperthreading (due to each 2 cores - 1 module - shares a lot of resources). But still from a preliminary analysis it's a good start.
Said that: +1 for you!

Whilst the cores are completley different - Im going for a clock for clock in general analysis. So it would work for a single core Pentium 4 against a 16 core Opteron, Im not talking perf per core, its perf per clock

And thanks, its nice to know my work is appreciated, rudementry or not