here are my components for my gaming rig

[mfenn]

Termie you snuck that in while I was writing my last thread, now your saying what I thought was the reasion I wanted the I7. I mean we are all talking about OCing out rigs and stuff trying to sqeeze every Mhz we can, so thats why i was thinking about the I7. $100 swing wont break me.

The question is not whether or not $100 will break you. The question is where you should spend that $100. You could spend $100 on upgrading an i5 to and i7 and get a 10-15% boost in gaming. Or you could spend that $100 on upgrading a GTX 660 Ti to a 7970 and get a 20-40% boost in gaming. The choice seems obvious to me.

 

[Termie]

The point that we both agree on is, "[The CPU] doesn't make as big a difference as an equal amount spent on a GPU". What we don't agree on is your straw man argument that "While many of the posts above suggest the 3770k doesn't make a difference in gaming, that is not true." Nobody is saying that the i7 3770 isn't faster. In fact, in direct contradiction to your premise, several people have pointed out that it is faster.

What we are saying is that in a world of finite budgets, you are better off spending money on the GPU before the CPU.



This is a misleading argument on two fronts.

1. You're making an assumption that the only difference between the 1366 and 1155 platforms is that one has HyperThreading and the other doesn't. This is absolutely false. There are tons of other platform differences that can account for the performance differential such as amount of cache, amount of memory bandwidth, etc. Additionally characterizing a 17.5% difference in average framerate as a "big difference" is somewhat misleading, especially given the cost.

2. The cost difference between an i5 750 1156 system and an i7 920 1366 system was certainly NOT $100. You have to look at the total platform cost, which would be more like $200.

People have said the 3770 is 100MHz faster, and have discounted HT as relevant to gaming. That is providing the OP incomplete information.

The 920 vs. 750 argument is for illustration purposes only. The OP isn't deciding between those two chips. But it's a great example of the value of hyperthreading - the triple-channel memory probably had nothing to do with those differences.

Today the OP has the chance to pay less than $100 TOTAL for a 3770k vs. a 3570k. I'd say beyond buying a 7970/670, any additional funds should be directed to getting that 3770k, rather than, say, a 680 or 7970 GHz edition. If he doesn't want to or can't spend the extra ~$100, he'll still be fine with the 3570k, but a few years down the line, the lack of hyperthreading may mean it drops further behind the 3770k.

 

[Cerb]

The question is not whether or not $100 will break you. The question is where you should spend that $100. You could spend $100 on upgrading an i5 to and i7 and get a 10-15% boost in gaming.

http://www.xbitlabs.com/articles/cpu/display/core-i7-3770k-i5-3570k_6.html#sect0

Usually half of even that much, even if you only go by min FPS (which I like to, but was having trouble rounding up a suitable review with these two CPUs, and they aren't on AT's bench gizmo).

I mean we are all talking about OCing out rigs and stuff trying to sqeeze every Mhz we can, so thats why i was thinking about the I7.

But, at the expense of smoother gameplay? A 670, 7950, or 7970, if you so much as turn MSAA on, will quickly become worth the cost over the 660 Ti; and 7950 cards are getting pretty righteous, right now.

IMO, go with a 7950 (or 7970 on sale at 7950 cost) and an i5-3570K, and put some of the saved money into a nice CPU heatsink and fan ($30-60).

but a few years down the line, the lack of hyperthreading may mean it drops further behind the 3770k.

HT will need double the threads, lower average IPC/thread, and a low total cache footprint. HT today doesn't have a significant downside, finally, but SCII, your example, does not take advantage of it. The very page you linked to even makes mention of that, and shows a quad-core CPU use graph.

The extra threads of the Core i7 920 processor are no advantage when compared to the Core i5 750 in this game, but the additional memory capacity and bandwidth is. The Core i7 920 was 11% faster when comparing the clock for clock data at 3.70GHz which is quite significant.

HT has benefited a handful of games, but the difference there was the platform, not HT. If a game could use 3+ channels of DDR3 with a IB CPU, it won't be doing it on 1155.

 

[lehtv]

My monitor that I just bought and have yet to recieve is a Asus VE258Q 25-Inch Full-HD LED with a res of 1920x1080P. So I think what you said above is its not worth the 670 card becasue my new monitor is a 1080 correct?

The part about the 670 not being worth it over 660 Ti was an opinion which I put in brackets. I said either is fine... If you're OK spending the extra $100 now then by all means go for the 670 and enjoy slightly higher framerates / settings

I see that the 660Ti has a 3G SC card that is $335. Is the extra gig worth $35?

No, you'd be much better off with a 7950 3GB for less $ if you needed that much VRAM

256k SSD card leaving room for an addition.

Just nitpicking here - we all know what you mean but the SSD is (a) not 256k, it's 256GB - gigabytes, not kilos - and (b) not a card, it is a drive, and the 'drive' part is already part of SSD - Solid State Drive.

thats the P8Z77-V, I think they all have about the same arrangment of PCIe slotts for the GPU. I know that they don't have duel PCIe 16x16 if I run SLI, but with a high end GPU figuring I wont need too.

There's no PCIe 16x16. Maybe you mean 16x / 16x? As in two slots both running at 16x? That's only possibel in very high end boards and also not needed at all. 8x / 8x is fine for SLI.

It's unlikely you will need SLI for 1080p when getting a $300+ graphics card. But you bought an X-750 so I think it's not a bad idea paying an extra couple of dozen bucks for a SLI compatible board as well. The P8Z77-V LK is SLI-compatible.

 

[mfenn]

Cerb and lehtv handled the other points nicely, but I wanted to point one more thing out.

The 920 vs. 750 argument is for illustration purposes only. The OP isn't deciding between those two chips. But it's a great example of the value of hyperthreading - the triple-channel memory probably had nothing to do with those differences.

You cannot simply make statements like the bolded without support. Obviously data would be the best, but I am OK with even some logical reasoning.

 

[Termie]

Cerb and lehtv handled the other points nicely, but I wanted to point one more thing out.



You cannot simply make statements like the bolded without support. Obviously data would be the best, but I am OK with even some logical reasoning.

Ok, forget the 750 vs. 920, because I really don't care too much about that benchmark. Here's all the proof you need.

The 2500k gets killed by an older, "slower" CPU with hyperthreading.

 

[mfenn]

Ok, forget the 750 vs. 920, because I really don't care too much about that benchmark. Here's all the proof you need.

The 2500k gets killed by an older, "slower" CPU with hyperthreading.

Thank you for digging up a benchmark. However, I don't think it quite shows what you think it does.

The i7 975 is a 1366 chip with the same base clock (ok 33MHz different) and only 100Mhz slower max turbo. It also has more memory bandwidth and more cache, just like the 1366 vs 1156 example earlier. From the comparison of those two chips, you cannot say whether or not it was HT that made the difference. (Note that this is different from saying that HT didn't make a difference. Science!)

Now, lets take the more direct comparison. i5 2500K vs. i7 2600K. First lets normalize for clock speed because those effects are well understood. The i7 2600K has 3% more clock speed than the i5, so lets increase the i5's score by 3% to simulate a hypothetical 3.4 GHz i5. That puts the scores at 98 for the i5 2500K and 111 for the i7 2600K (i7 2600K is 13% faster). The i7 has more cache than the i5 and has HT. We know that some combination of those two makes a 13% difference, but the data really isn't enough to say which.

 

[lehtv]

I think it's pretty clear that the difference in lowest FPS between 2500K and 2600K is caused by lack of HT, assuming no experimental errors during benchmarking the 2500K. I find it very hard to believe that a 40% increase could be caused by having a bit more cache. The lowest FPS depicts a scenario where the game is simultaneously very GPU and CPU heavy. That's where the extra threads benefit the game the most. That the difference in average FPS is quite a bit smaller just shows that these CPU heavy situations are not so common

 

[Termie]

I think it's pretty clear that the difference in lowest FPS between 2500K and 2600K is caused by lack of HT, assuming no experimental errors during benchmarking the 2500K. I find it very hard to believe that a 40% increase could be caused by having a bit more cache. The lowest FPS depicts a scenario where the game is simultaneously very GPU and CPU heavy. That's where the extra threads benefit the game the most. That the difference in average FPS is quite a bit smaller just shows that these CPU heavy situations are not so common

This has me thinking that's it's about time someone put together a thread on this forum with benchmarks for HT and non-HT processors. I will get some benchmarks together as I have time, maybe on 5 or 6 games at the settings I normally play at, changing only my HT setting in the BIOS.

I began last night with BF3, but honestly my testing at ultra 1920x1200 was inconclusive, not to mention incredibly time-consuming since it was multiplayer.

I'll start with single-player run-throughs, not canned benchmarks, as I don't believe built-in benchmarks sufficiently stress the processor. By the way, I've found that setting processor affinity does not work correctly and the HT must be disabled in the BIOS, adding to the difficulty of back-to-back testing.

Watch the CPU forum for a new thread in the next week or so.

 

[lehtv]

Single-player run-throughs probably don't stress the CPU enough to see the difference between HT and not

 

[Termie]

Single-player run-throughs probably don't stress the CPU enough to see the difference between HT and not

Well, that's what I'm going to start with. The only multiplayer game I actually play is BF3, and getting consistent runs is nearly impossible, even when not changing any settings. I'll save it for last, because I have to do something drastically different than what I did last night. My 5 runs each on HT and non-HT produced averages that varied by as much as 20fps for each setting. Impossible to come to any conclusions with data like that.

 

[lehtv]

My 5 runs each on HT and non-HT produced averages that varied by as much as 20fps for each setting. Impossible to come to any conclusions with data like that.

Were they all on the same map with the same amount of players? I would probably try to play an open map like Firestorm with 64 players and use the same playstyle for each run, e.g. infantry guy holding objectives mid-map

 

[Termie]

Were they all on the same map with the same amount of players? I would probably try to play an open map like Firestorm with 64 players and use the same playstyle for each run, e.g. infantry guy holding objectives mid-map

That's exactly what I did. 64-player Firestorm server that wasn't full, because I had to get back into it after adjusting settings. The simplest thing like switching sides after the round completely messed up my testing, because my team no longer had the flag I'd benched on. I eventually captured it on my own, but was indoors while benching, as opposed to outdoors in the previous round. Simply impossible to use the data I collected over 45 minutes of playing/rebooting to adjust settings/logging back in/playing.

Frankly, I really don't like benchmarking with BF3. It's a huge time-waster, and it means I'm not enjoying the game.

Anyway, forum user Moonbogg has already documented the effects of HT in this game through very careful and time-intensive benchmarking, and I don't feel a strong need to repeat it: http://forums.anandtech.com/showpost.php?p=33814087&postcount=174

 

[BTRY B 529th FA BN]

someone did a comparison post using th me UE3 engine on a single platform showing benches with differences on a few things, HT being one of them. on my phone - ill dig it up later at home

 

[Termie]

someone did a comparison post using th me UE3 engine on a single platform showing benches with differences on a few things, HT being one of them. on my phone - ill dig it up later at home

I did that for Batman. If that's what you're thinking of, save yourself the digging!

 

[postmortemIA]

Thank you for digging up a benchmark. However, I don't think it quite shows what you think it does.

The i7 975 is a 1366 chip with the same base clock (ok 33MHz different) and only 100Mhz slower max turbo. It also has more memory bandwidth and more cache, just like the 1366 vs 1156 example earlier. From the comparison of those two chips, you cannot say whether or not it was HT that made the difference. (Note that this is different from saying that HT didn't make a difference. Science!)

Now, lets take the more direct comparison. i5 2500K vs. i7 2600K. First lets normalize for clock speed because those effects are well understood. The i7 2600K has 3% more clock speed than the i5, so lets increase the i5's score by 3% to simulate a hypothetical 3.4 GHz i5. That puts the scores at 98 for the i5 2500K and 111 for the i7 2600K (i7 2600K is 13% faster). The i7 has more cache than the i5 and has HT. We know that some combination of those two makes a 13% difference, but the data really isn't enough to say which.

Point is, all 3 make difference, and you can't get either of them on i5.
Suppose I'm building a new PC today (I actually am, and I already got i7, which makes me biased). I am looking for decent board, memory, and such, so I would get same parts, regardless of if I select i5 or i7. Now after all these are taken into consideration, price difference becomes 15% (i5 is $220, i7 is $290. And looks like form benchmarks that it is worth it.

Now, once I assemble my system tomorrow I will run benchmark HT on vs HT off and we'll see if it makes any difference in single threaded games.

 

[BTRY B 529th FA BN]

try the UE3 engine aswell

 

[mfenn]

Point is, all 3 make difference, and you can't get either of them on i5.

Never said they didn't. What I am saying is that you can't just pin it all on HT. Termie was taking the easy way out initially and not considering that other factors could have an effect. I was simply pointing that out.

Suppose I'm building a new PC today (I actually am, and I already got i7, which makes me biased). I am looking for decent board, memory, and such, so I would get same parts, regardless of if I select i5 or i7. Now after all these are taken into consideration, price difference becomes 15% (i5 is $220, i7 is $290. And looks like form benchmarks that it is worth it.

I think you're looking at it the wrong way. First of all, the difference is only 15% in a few CPU limited games. In most games, you are looking at a much smaller difference.

Second, the cost difference is not just 15%. The difference is $70 which is nearly 7950 -> 7970 or GTX 660 -> GTX 670, both of which make a bigger difference in gaming.

 

[BTRY B 529th FA BN]

I did that for Batman. If that's what you're thinking of, save yourself the digging!

Food for thought

 

[fng01]

OK I have already ordered and recieved some of this stuff. I took a lot of what some of you folks suggested seruoisly, I thank you all for that. You'll find some of which I did not. I did my own research and came to my own conclutions baised on what I will be using the rig for. I know many of you guys wont approve of my CPU and motherboard choice and probly not even my GPU but I know I will be happy and the equipment it top quality. Its been a long time since I have upgraded and build a new rig, I am using a E8500 C2D CPU with an old Asus P5N32-E board and they have been good to me with no problems but this new rig will feel like a cadilac.

Corsair Obsidian Case
Asus P8Z77-V Pro
I7-3770K
Hyper 212 Plus
Corsair Vengance Blue 16 GB DDR3 1600
EVGA GeForce GTX 660Ti SuperClocked 3072MB GDDR5
Samsung 830 256GB SSD
Asus DVD burner
Seasonic X750W Gold Power Supply
Asus VE285Q-25" LED monitor
Win7 64 OEM home Premium
Also got the Battlefield 3 Premium addition
my graphics card comes with Borderlands 2

 

[Termie]

OK I have already ordered and recieved some of this stuff. I took a lot of what some of you folks suggested seruoisly, I thank you all for that. You'll find some of which I did not. I did my own research and came to my own conclutions baised on what I will be using the rig for. I know many of you guys wont approve of my CPU and motherboard choice and probly not even my GPU but I know I will be happy and the equipment it top quality. Its been a long time since I have upgraded and build a new rig, I am using a E8500 C2D CPU with an old Asus P5N32-E board and they have been good to me with no problems but this new rig will feel like a cadilac.

Corsair Obsidian Case
Asus P8Z77-V Pro
I7-3770K
Hyper 212 Plus
Corsair Vengance Blue 16 GB DDR3 1600
EVGA GeForce GTX 660Ti SuperClocked 3072MB GDDR5
Samsung 830 256GB SSD
Asus DVD burner
Seasonic X750W Gold Power Supply
Asus VE285Q-25" LED monitor
Win7 64 OEM home Premium
Also got the Battlefield 3 Premium addition
my graphics card comes with Borderlands 2

That is an excellent build. Have fun with it!