Intel Core i7 Review Thread

[savantu]

lol, where did i say that?



because k10 is a slow cpu....

dont forget all memory read/writes are placed on the FSB, if all data of 3 GTX280 have to placed on that same FSB, a serious bottleneck is born
i also thought the fsb isnt bidrectional, so you have to wait before you can send something in a direction

http://www.xtremesystems.org/forums/...&postcount=128

Are you dumb, disingenuous or a combination of both ??

[Bellisimo]

You didnt, but you obviously dont understand it either for what the GPU needs.

The only large bandiwdth requiring part is textures. And that doesnt go via the FSB on a Core 2. Hence why JJs test showed no difference between 200 and 400.

5x vs 70 fps is a large difference if you ask me
but he showed benches with a single card, i comment on a system with 3-way SLI => FSB becomes to crowded

http://www.xtremesystems.org/forums/...&postcount=128

Are you dumb, disingenuous or a combination of both ??

Simplification process for savantu:
memory in my post = main memory for the cpu
data from 3*gtx280 = cpu-gpu communication

[Shintai]

5x vs 70 fps is a large difference if you ask me
but he showed benches with a single card, i comment on a system with 3-way SLI => FSB becomes to crowded

Prove it. You dont understand it and obviously are out on deep water where you cant stand.

The FSB is NOT crowded for graphics. Also improving FSB speeds shows virtually nothing. Unless the CPU itself needs the bandwidth. Like photoshop.

[gosh]

Nowhere does the GPu go over the CPU for its textures.

Texture data is probably processed by the processor before going to det video card. If it goes using the processor or direct from memory doesn't matter, I think that that type of action needs the cpu to be synchronized. There are some operatioins that need this. Memory needs to be in the videocardmem if you want good performance

[Bellisimo]

Prove it. You dont understand it and obviously are out on deep water where you cant stand.

The FSB is NOT crowded for graphics. Also improving FSB speeds shows virtually nothing. Unless the CPU itself needs the bandwidth. Like photoshop.

i said FSB becomes crowded with 3 powerfull gpu's communicating with the cpu, or do you think gpu's dont communicate with the cpu?
if you took a look at my anandtechh link, you see 2 cards scale fine, so how come 3 cards dont?

You have 3 people including proof against you. I think its time for you to show the edvidence for your claims.

i dont see any proof, do you?

[gosh]

The FSB is NOT crowded for graphics. Also improving FSB speeds shows virtually nothing. Unless the CPU itself needs the bandwidth. Like photoshop.

There are a lot of dots sent to the video card, also on Core 2 the FSB handles both memory and video traffic. Haven't read about i7 but a assume that it can handle ram and video card independently.

[savantu]

i said FSB becomes crowded with 3 powerfull gpu's communicating with the cpu, or do you think gpu's dont communicate with the cpu?
if you took a look at my anandtechh link, you see 2 cards scale fine, so how come 3 cards dont?

It's the drivers Einstein.

[Shintai]

There are a lot of dots sent to the video card, also on Core 2 the FSB handles both memory and video traffic. Haven't read about i7 but a assume that it can handle ram and video card independently.

Hence back to the K8/K10 issue.

Also alot of dots? Do your 1Gbit line gets lagged by 1000 10Kbit connections? Even in half duplex?

[Bellisimo]

It's the drivers Einstein.

so 3-way sli performs better on x58 because of the graphics drivers?

[gosh]

It's the drivers Einstein.

Driver is executed by the CPU

[savantu]

There are a lot of dots sent to the video card, also on Core 2 the FSB handles both memory and video traffic. Haven't read about i7 but a assume that it can handle ram and video card independently.

I'm pretty sure GPUs tax memory a lot more than the CPU in a graphic intensive game.

In FSB topology you have :

-CPU-FSB-RAM
-GPU-FSB-CPU
-GPU-RAM

With QPI topology you have
-CPU-RAM
-GPU-QPI-CPU
-GPU-QPI-CPU-RAM

Basically FSB has the upper hand in GPU accessing memory while the QPI system is in CPU accessing memory.
Core beating the crap out of K8/10 in games is proof that the FSB isn't really a bottleneck because GPUs care more about accessing memory than flooding the FSB with data ( which btw , I believe isn't exactly large in size , I'd even venture to say that latency is more important )

Disclaimer - All of the above is a guess and should be taken with a spoon of salt.

[savantu]

so 3-way sli performs better on x58 because of the graphics drivers?

Performs better because Nehalem can run the crappy driver faster than Core.Had you bother to interpret the results , you would have known that at low resolutions you end up CPU bottlenecked.The fact that Nehalem offers 30-50% better results at low resolution is evidence of increased computing power.

[gosh]

I'm pretty sure GPUs tax memory a lot more than the CPU in a graphic intensive game.

If it does then performance will be very slow.
But if you were right about this you still need to consider that the video card and CPU runs asynchronously. That means that if the video card or the CPU is talking to some area that they both are using that type of operation needs to be synchronized. I think that game developers tries to avoid that type of operation as much as they can.

[Bellisimo]

As you can see on this link

http://www.anandtech.com/cpuchipsets...spx?i=3265&p=5

With systems containing multiple GPUs, the CPU often needs to send the same data to each GPU - for example, when texture, geometry, or other common rendering data must be transfer over the FSB, ultimately intended for receipt by all GPUs. Normally the CPU must send a separate set of data packages in order to target each GPU. This causes FSB congestion due to the increased CPU-to-GPU traffic, which could be better used for other purposes such as main memory read or write requests or cache coherency with quad-core CPUs.

In addition to the direct link technology previously explained, Broadcast technology allows only one message to be sent by the CPU where it is then received, replicated, and broadcasted to all GPUs, eliminating the need for multiple, near-identical transfers over the FSB. This allows for more timely access to other resources that must share this common interface. We're not exactly sure how the SPP can control what the CPU does and does not send over the FSB, and it seems likely that the system drivers also have changes to accommodate this new technology; either way, there's definitely more to this than we have shared so far. We will be sure to pass along anything we learn

Nvida tries to reduce gpu-cpu traffic here, because the fsb gets crowded...
again i repeat, i am talking about multi-gpu systems

[AKM]

The fact that Nehalem offers 30-50% better results at low resolution is evidence of increased computing power.

I don't see anything close to 30-50%/clock improvement at low res.

http://www.legionhardware.com/document.php?id=786&p=10

[savantu]

I don't see anything close to 30-50%/clock improvement at low res.

http://www.legionhardware.com/document.php?id=786&p=10

I'm talking about the graphs with multi GPU configs.Single GPU can't unleash Nehalem.

[AKM]

Not even at lowest quality setting and 800x600 res? I doubt that.

[Bellisimo]

I'm talking about the graphs with multi GPU configs.Single GPU can't unleash Nehalem.

not even at low res configs? quite funny, cause low res is always cpu-limited
it doenst matter if you have 3 or 2 GTX280 on low res

[Shintai]

Not even at lowest quality setting and 800x600 res? I doubt that.

[kl0012]

Looks like Nehalem is a folding monster.
Core i7-920 almost twice as fast as Q6600:

[Hornet331]

Looks like Nehalem is a folding monster.
Core i7-920 almost twice as fast as Q6600:

Exactly what im looking for.

[WheelMan]

Norwegian review for those who might be interrested:
http://www.hardware.no/artikler/intel_core_i7_920/57245

Translated to Engrish ( ) by Google:
http://translate.google.com/translat...-8&sl=no&tl=en
Worst translation in years, i think..

[Shintai]

Norwegian review for those who might be interrested:
http://www.hardware.no/artikler/intel_core_i7_920/57245

Translated to Engrish ( ) by Google:
http://translate.google.com/translat...-8&sl=no&tl=en
Worst translation in years, i think..

Nej, det værre med kinesisk

[kl0012]

Initial test: One Core i7-965 overtakes two quad-core Opterons (2360 SE) in fp_rate_base_2006 (SPEC CPU2006):
http://www.heise-online.co.uk/news/I...e-x86--/111860

[Shintai]

Initial test: One Core i7-965 overtakes two quad-core Opterons (2360 SE) in fp_rate_base_2006 (SPEC CPU2006):
http://www.heise-online.co.uk/news/I...e-x86--/111860