Questions concerning specifications

[Cybertox]

I have some questions about computer specifications.

1. The speed of ram, is it a noticeable upgrade when upgrading from the 4 gbs 667mhz to 4gbs 1333 mhz?
How does the speed of ram affects performance?

2. If a dual core processor has 2.4 ghz, both cores run at 1.2 right?

3. The size of the cache in a processor, what does it do and is bigger cache a noticeable upgrade from a processor with a small amount of cache?

4. What is more important in a videocard (gpu), its number of cores or its vram?

5. SSD cards are faster than hard drives because they have more rpm, or there is another reason?

6. Can you explain me how integrated intel graphics work?

7. When you have two gpus both of them execute the same amount of work, is that right?

Hopefully you can answer my questions and thanks in advance!

[keefy]

1. The speed of ram, is it a noticeable upgrade when upgrading from the 4 gbs 667mhz to 4gbs 1333 mhz?
How does the speed of ram affects performance?

I have not experienced it for myself but I dont think it it would be worth the upgrade.


2. If a dual core processor has 2.4 ghz, both cores run at 1.2 right?

No both run at 2.4GHz

3. The size of the cache in a processor, what does it do and is bigger cache a noticeable upgrade from a processor with a small amount of cache?

More the better thats all I know. But i doubt it would be worth upgrading to a CPU with 4MB than from one with 2MB if the only difference is the cache.

4. What is more important in a videocard (gpu), its number of cores or its vram?



5. SSD cards are faster than hard drives because they have more rpm, or there is another reason?

No they are nto mechanical they are faster because they have no moving parts its all stored on memory chips and can be accessed almost instantly.

6. Can you explain me how integrated intel graphics work?


7. When you have two gpus both of them execute the same amount of work, is that right?

Hopefully you can answer my questions and thanks in advance!

[Vulgotha]

Ram speed affects bandwidth. Basically how quickly you can load something into memory. This will result in increased performance across the board providing your CPU can keep pace.

Solid State Drives are not mechanical. They are not magnetic based storage, more closely related to flash memory. They are faster by a large margin.

[Varsh]

1. The speed of ram, is it a noticeable upgrade when upgrading from the 4 gbs 667mhz to 4gbs 1333 mhz?
How does the speed of ram affects performance?

In terms of gaming performance very little, but if you're one to do any sort of media editing then you'll notice a massive difference.

2. If a dual core processor has 2.4 ghz, both cores run at 1.2 right?

Both run at 2.4GHz however keep in mind that apps have to take into account the different cores, so if an app isn't multi-threaded then it'll be slower than a single core that's clocked faster - not a problem nowadays.

3. The size of the cache in a processor, what does it do and is bigger cache a noticeable upgrade from a processor with a small amount of cache?

The more cache the less the CPU needs to send data to RAM for storage before being sent back to the CPU for further processing, this alleviates a strain on your FSB allowing more data being able to process through it (less of a bottleneck).

4. What is more important in a videocard (gpu), its number of cores or its vram?

For games it's the core count and clock speed, for graphic programmes like 3ds Max, AutoCAD, Photoshop and others, it's VRAM you'll need more of due to texture space. For instance I have a GTX 285 which has 3GB RAM, I wouldn't be able to do half of my 3D work if I had a standard 1GB card that's on the market.

5. SSD cards are faster than hard drives because they have more rpm, or there is another reason?

SSDs are based on NAND Flash, however because of this they are non-volatile and much more expensive. It's a trade off really, HDDs are for stability, reliability and storage while SSDs are for access speed - if used solely for an OS only installation then it's a massive benefit.

6. Can you explain me how integrated intel graphics work?

IGPs work in the same way as GFX cards, the only difference other than it being integrated on the motherboard is that its VRAM is generated from your onboard RAM, so if you had 8GB RAM in your system and you set in the BIOS that the IGP will use 2GB RAM, then your IGP will use 2GB RAM and your system has your RAM split up and have 6GB remaining for everything else. It's not ideal as GDDR is exponentially faster than DDR.

7. When you have two gpus both of them execute the same amount of work, is that right?

Yes and no, if you're playing a game that uses SLi or Crossfire then yes they both work in tandem to push the game along, but usually you gain only about 20-40% of the FPS. On the other hand if you're using multiple monitors and playing the game on a single screen then it's the GPU that is displaying that image that will be used (generally this would be the primary GPU).

[Morganator]

I have some questions about computer specifications.

1. The speed of ram, is it a noticeable upgrade when upgrading from the 4 gbs 667mhz to 4gbs 1333 mhz?
How does the speed of ram affects performance?

This is really dependent on several factors. These factors include speed, cas latency, architecture, form factor, bus speed and width and a host of other factors. While the speed of ram may be different and may have a noticeable effect on the fetching and storing of data. If I have 4GB of 667MHz and another 4GB of 1333MHz ram and the speed is all I have to go by, I'm gonna choose the 1333MHz option. However, if I have 4GB of 667MHz Quad Channel ram and another 4GB of 1333MHz of Dual Channel ram, I'm gonna choose the 667MHz ram assuming my motherboard is equipped to handle Quad Channel ram. When it comes to latency, the lower the latency, the better. Latency plays a minor but important role in the fetching of stored data in the ram. architecture is also key but not an issue unless you're looking into building a computer from scratch and I mean scratch. Form factor is important but only in regards to shopping for laptops.

2. If a dual core processor has 2.4 ghz, both cores run at 1.2 right?

No. Both cores run at 2.4GHz.

3. The size of the cache in a processor, what does it do and is bigger cache a noticeable upgrade from a processor with a small amount of cache?

The cache in a processor is essentially on-chip ram meant only for processes that are called constantly. Logic dictates that the more cache a processor has, the faster computations are done because the data which is called on constantly has no distance to travel. Cache though, has a diminishing ROI, meaning, the more you have the less that is used proportionately to how much you've spent and/or the less you see in performance gains in proportion to how much you've spent. This all boils down to how much your programs require. Ironically, gaming benefits from this more than workstation applications.

4. What is more important in a videocard (gpu), its number of cores or its vram?

Personally, I think the number of cores is more important. Obviously having more vram helps with data access times but I think that the more cores you have, the less ram is needed because while one core is focused one thing, another is focused on another.

5. SSD cards are faster than hard drives because they have more rpm, or there is another reason?

No. SSD drives are faster in that they have no moving parts, run cooler and because the memory storage architecture is set up in a way that allows for fast read and access speeds. Ironically, current SSD technology suffers from slower write speeds.

6. Can you explain me how integrated intel graphics work?

Integrated graphics is merely the graphics chip being soldiered straight to the motherboard. Integrated is just another term for embedded chip. There are several advantages to embedded technology as opposed to add-on card technology. They are fixed and have very low latency though the tradeoff is they're not upgradeable and are thus, a fixed quantity.

7. When you have two gpus both of them execute the same amount of work, is that right?

Hopefully you can answer my questions and thanks in advance!

This is all dependent on your setup. The premise is that two GPUs execute twice the amount of work as one. This isn't entirely true. The first GPU will always work under full load while the second one more or less, augments the workflow. The performance gain of two GPUs as opposed to one is actually only 50% in most cases, Depending on the setup, you could theoretically achieve up to 75% increase in performance but real world applications aren't optimized for anything higher.

[Cybertox]

Thank you all for the answers, now i can understand everything that has been asked precisely.

[keefy]

Inside a conventional hard drive versus a Solid State Drive,

The bottom is SSD.

[ex nihilo7]

I will also give my thoughts.

I have some questions about computer specifications.

1. The speed of ram, is it a noticeable upgrade when upgrading from the 4 gbs 667mhz to 4gbs 1333 mhz?
How does the speed of ram affects performance?

Yes, but some applications will not show as much benefit.

2. If a dual core processor has 2.4 ghz, both cores run at 1.2 right?

Both run at 2.4 GHz. Keep in mind, the clock speed might be slightly slower or slightly faster than 2.4GHz. They will not both have the exact clock speed, but for all practial purposes you can say the do.

3. The size of the cache in a processor, what does it do and is bigger cache a noticeable upgrade from a processor with a small amount of cache?

Cache is the second fastest in the memory hierarchy. It goes registers > cache > ram > flash > hard disk.
The larger the cache, the more information it can store in there, and the less the CPU has to go to the ram to find what it needs. Since cache is faster than ram, you notice a performance boost. If it can't find what it needs, then its called a page fault.


4. What is more important in a videocard (gpu), its number of cores or its vram?

Depends on the situation. I would have to say vram first, then cores. I had to do a lab with CUDA, and my GPU could do more work than a friends GPU because I had more vram in mine. We had the same number of cores though. There is like 32 threads doing work for each processing core in the NVIDIA cards. Warp size for CUDA cards is 32 threads, and each warp does work independently of each other. The CPU is considered the Host, and the GPU the device. There has to be communication between the GPU and the CPU. All the parallel work is done in the CUDA Kernels. That's how it works for CUDA programming, but since games use Direct X and OpenGL, I'm not sure how much different it is. I know its a lot harder to program though.

5. SSD cards are faster than hard drives because they have more rpm, or there is another reason?

Refer to memory hierarchy.

6. Can you explain me how integrated intel graphics work?

They incorporate vector and pixel-shader processors into the CPU die? Not sure actually. Never looked into it.

7. When you have two gpus both of them execute the same amount of work, is that right?

Not necessarily, no. You can get great performance benefits if the code is optimized, but it doesn't scale to be twice as good in real world performance.

Hopefully you can answer my questions and thanks in advance!

[Centurion]

Cache is the second fastest in the memory hierarchy. It goes registers > cache > ram > flash > hard disk.
The larger the cache, the more information it can store in there, and the less the CPU has to go to the ram to find what it needs. Since cache is faster than ram, you notice a performance boost. If it can't find what it needs, then its called a page fault.

Registers shouldn't be confused with memory.And you're using the term page fault incorrectly here.The term you're looking for is cache miss.

[ex nihilo7]

[/B]Registers shouldn't be confused with memory.And you're using the term page fault incorrectly here.The term you're looking for is cache miss.

Oh wait, yeah you are right about the cache miss. I was thinking of RAM and TLB, translation etc. Whenever you google memory hierarchy or look in a textbook, they always have processor registers as the fastest in the memory hierarchy. Its not main memory for sure.

[itachi73378]

Since we have this talk about registers and Cache going on I'll ask a question myself.
Is the register memory of a much faster and different variety than that of RAM? I read that cache memory is much more expensive so they use little of it, however at the same time I feel like a large chunk of the advantage the cache has over RAM is how close it is to the processor (therefore it is faster)

Am I right in thinking so?

[Centurion]

Since we have this talk about registers and Cache going on I'll ask a question myself.
Is the register memory of a much faster and different variety than that of RAM? I read that cache memory is much more expensive so they use little of it, however at the same time I feel like a large chunk of the advantage the cache has over RAM is how close it is to the processor (therefore it is faster)

Am I right in thinking so?

Registers aren't memory in the same sense as RAM and cache. They're not DRAM or SRAM. They are a series of flip-flops inside the chip. The purpose of registers is to hold instruction operands and results. On most processors, if you want to add two values, you first load the values from memory into the registers using load instructions, then add the two values using an add instruction, then write the result back to memory using a store instruction (or keep it in the register for later use). This isn't necessarily true for the Intel case, since some of its instructions can operate on memory directly.

Careful use of registers plays a crucial role in how code performs.

CPUs have very few registers. 32-bit x86 processors have 8 32-bit general-purpose registers, while x86-64/x64 processors have 16 64-bit general-purpose registers.

RISC designs tend to have more registers. For example, PowerPC processors have 32 general-purpose registers. The SPU's in the PS3's Cell processor have 128 registers.

Also, these are only the registers visible to the programmer. Modern processors do a lot of tricks behind the scenes and may have many more registers that the programmer cannot access directly (i.e. register renaming).

You are correct that cache is fast because it is on-chip. But that is also why it is so expensive. Cache consumes CPU die space, which is limited.

[itachi73378]

Registers aren't memory in the same sense as RAM and cache. They're not DRAM or SRAM. They are a series of flip-flops inside the chip. The purpose of registers is to hold instruction operands and results. On most processors, if you want to add two values, you first load the values from memory into the registers using load instructions, then add the two values using an add instruction, then write the result back to memory using a store instruction (or keep it in the register for later use). This isn't necessarily true for the Intel case, since some of its instructions can operate on memory directly.

Careful use of registers plays a crucial role in how code performs.

CPUs have very few registers. 32-bit x86 processors have 8 32-bit general-purpose registers, while x86-64/x64 processors have 16 64-bit general-purpose registers.

RISC designs tend to have more registers. For example, PowerPC processors have 32 general-purpose registers. The SPU's in the PS3's Cell processor have 128 registers.

Also, these are only the registers visible to the programmer. Modern processors do a lot of tricks behind the scenes and may have many more registers that the programmer cannot access directly (i.e. register renaming).

You are correct that cache is fast because it is on-chip. But that is also why it is so expensive. Cache consumes CPU die space, which is limited.

Oh so registers contain the data manipulation circuits. And if I'm not mistaken these circuits are located in the Arithematic Logic unit of the processor (ALU) correct? But I remember reading that there is a high speed register storage other than the ALU.

The way you put it it sounds like the registers are read only.

[Centurion]

Oh so registers contain the data manipulation circuits. And if I'm not mistaken these circuits are located in the Arithematic Logic unit of the processor (ALU) correct? But I remember reading that there is a high speed register storage other than the ALU.

The way you put it it sounds like the registers are read only.

Most registers are not read-only. You can write to them (load value from memory, copy from another register, write result from instruction) and read from them (store value to memory, use value in computation). There are read-only registers that hold status flags, on some architectures.

There are units other than the ALU, e.g. the Floating Point Unit (FPU). On Intel processors the FPU has its own set of registers. On PowerPC the general-purpose registers are also used to hold floating-point values.

Registers just hold values.. the logic for manipulating the data is in the execution units (APU,FPU etc).