AMD R9 or Nvidia Gtx. CPU Bottleneck on Core i5 750 2.66 (Lynnfield).

The GTX 760 is slightly ahead , not by much .

However if you play racing sims then most probably the 270x will serve you just fine and save some money towards a cpu/mobo upgrade.

You'll be bottlenecked to some extent using either of them, though it would help if you can oc your CPU.
Most 750s with a decent cooler (old TRUE is plenty) will do 4GHz on air. The later i5 760 will do 4.4GHz.
That will help reduce the bottleneck quite a lot. Or of course you could replace the 750 with an i7 870
which will oc to around 4.2GHz on a decent mbd with a good air cooler (4GHz on average mbds), and
the HT will help with some games aswell.

Between the two, I'd recommend the 760, if only because - testing lots of cards as I do - the lower
quality of AMD's drivers is IMO still a factor, and DX9/CF is still broken.

Btw, if you have the option, consider used items (don't know your choices given your location, I know
some S.A. nations have very high import taxes), eg. a 7970 3GB or GTX 580 are both fast; the 7970
is quicker, basically the same card as the newer 280X, but I have far less driver issues with 580s. A
580 1.5GB card from eBay should be around $150 or less (bit more for a 3GB version). 7970 will be
a bit more again (I bought two 7970s for around 145 UKP each).

Various benchmark results on my site here.

Ian.

PS. Re future upgrades, a used Z68 board and used 2600K or 2700K are good choices. I built almost
all of my gaming system from used parts, massive savings vs. buying new (it's a 2700K @ 5GHz, two
GTX 580 3GB in SLI).

Your screen is gonna determine your response. The GPU just calculates and forms the frames. For high end systems where your trying to remove frame tearing and such its a screen and GPU combination issue, where certain GPU's offer bettter technologies to help the screen stay organized. Basically Whats really important is having a nice high performance screen before considering a GPU that is designed to run "smoother". A screen at 60hz with a 5-10ms response time is fairly standard and no GPU can over come the limitations of the screen. That being said, i also notice that processor. which wont be able to keep up with those gpu's. what i get when looking at your build is that you bought this system a long time ago and never really intended to get into high end gaming. So if you really cant afford a rebuild and just want to ad a card to get smooth game play, you should look at R7 series from AMD or lower. you dont need alot of power cause it will go to waste. and spending extra money on it will be a waste. ALso if you wanna buy a card now that wont be useless if you upgrade later, get a 750ti. its cheap, its a monster at 1080, it requires no extra pcie power cables, and its on a modern architecture meaning it will support future release features. that card will be bottlenecked by your cpu, but when its time to upgrade you have a card that will continue to compete into the future thanks to driver support for new architecture

I've posted the same info a dozen times in the past couple of weeks.

It depends on the game, resolution, detail level, etc. If he's playing games like
Call of Juarez, then the CPU will be perfectly ok, even with newer cards (CoJ
doesn't need much CPU power). Other games which behave like X3TC are
entirely dependent on CPU power. And then there are infamous titles like Crysis
which need a more balanced system, but preferably as fast as possible across
the board (Stalker behaves this way, but is less intense than Crysis). Atm I'm
not familiar with how racing games behave, not a genre I've looked into much
for my benchmarking.

A good card will not be wasted in all games, and often the bottleneck will not
be that significant, especially at higher resolution/detail levels where the GPU
becomes the more important factor.

The 750 Ti is a nice card, but remember one cannot add another for SLI, so
its future usefulness will be limited to a degree. It does however have a big
advantage in power/noise behaviour.


Jean, when you're playing those games you've listed, what is the CPU doing?
Use system monitoring tools like CoreTemp, RealTemp, or even just Windows
Tasks Manager, to see what the system is up to. Likewise, you can use GPU-Z
or other tools like MSI Afterburner, GPU Shark, etc. to monitor GPU behaviour.

If the games you like to play do not show that much CPU usage, then a GPU
upgrade will work nicely. And even if those games do show a fair amount of
CPU activity, as I say you can oc the CPU, upgrade to an oc'd 870, etc. NB: don't
upgrade to an i7 860, they're hard to oc.

Have a look at my benchmark results, and if you like I'd be happy to run an
i5-760/7970 test to show what you'd get from a 280X (and hence a ceiling for
how a 270X would behave, or I could even downclock the 7970 to match a 270X).

Oh, forgot to ask: what GPU do you have at the moment?

Ian.

While the GTX 760 is better yes but the drivers problem occurs to both factors , I've owned more than 5 AMD cards in the last 5 years and all is well .

Driver issues affect NVIDIA cards to a far lesser degree. I've benchmarked dozens of cards, not just a mere 5.
This is my custom GPU storage unit, with about 20 cards not visible as they're in systems or on about 15
open-air test boards in various configs:

http://www.sgidepot.co.uk/misc/gpushelves15.jpg

I've tested 2/3-way SLI with numerous 580s, and 2-way CF with two 7970s. AMD's drivers consistently throw up a
whole range of weird oddities in behaviour that just don't happen with NVIDIA drivers, eg. GPU-Z freezing after
a settings change, very slow switch from CF to non-CF, etc. I had the same issues with the 4870, 5850, etc., though
the latest AMD drivers are easily the worst (older drivers I used with 5850s were less pernickity).

It's less of a problem when using a single card, but then for any sensible GPU investment, adding a 2nd in the
future as an intermediate speed boost is sound planning, so driver issues should be considered.

Ian.

TopLuca writes:
> Adding a 2nd card imo is a last resort or the card you already have is a card is the top of line. ..

Nonsense, going CF or SLI can give a very impressive speedup, depending on various factors.
Whether or not the OP would wish to do that is open to question, though his choice of mbd
is noteable.


> ... and yet again the CPU needs upgrade before anything ( adding a 2nd card ).

Stop posting potentially inaccurate info. The CPU only needs changing IF he sees significant
CPU loading while playing the games he likes to play. As yet he's not said whether that's the
case. And even if so, if the i5 is not oc'd then there is significant scope for easing that bottleneck.
As I said, a cheap used cooler like a TRUE will easily get the i5 750 to 4GHz.



> going to happen and the chances of occurring are fairly minimal.

I agree for 1-card it's less of a problem, but still something to bare in mind if multi-GPU is
ever going to be considered.

Can't comment on multi-screen setups, not something I've looked into much.


> Is the extra $40 in the favor of the GTX 760 worthy ? Not in the least.

Imo both prices are too high, hence my mentioning used GPU options - 7970 is much
better value, though the option may be void depending on his location.


> Unless he wants the 760 for streaming using the amazing shadowplay or he prefers Nvidia
> otherwise 40 more bucks for something fairly minimal doesn't seem a correct choice especially
> when you need a full PC upgrade

He doesn't necessarily need a full PC upgrade, that could be flat out not the case. We can't
judge until he's done a few tests. And if he's only running at stock, there's plenty of scope
for an excellent speedup.

Using your own rationale though, if he did get a 270X, then the saved $40 could be used
to fit a much better cooler with a couple of nice quiet fans, giving a major CPU boost, if
that is the i5 is currently only at stock.

Remember: P55 was a low-latency design, it was a lot more potent than X58-buyers gave
it credit for. I was able to prove over & over again from tests that P55 + Lynnfield could
beat X58 + Nehalem, for much less cost. I've proved it dozens of times. Don't write off
a P55 build just because it's "old"; indeed, an oc'd 870 gives a better 3DMark physics
score than a 3570K.

Anyway, conclusions are null until we know the CPU loading he's getting.

Ian.

The Dual-X will just do fine , The Toxic is definitely better but not really worth the extra 80 bucks.

Overclocking is a good idea, those Lynnfield CPUs will go to 3.4GHz without too much effort (mine did) but take your time and do your research, don't rush.
Even overclocked it may slightly restrict a R9 270X, but don't worry about it.
Don't bother about a 4Gb R9 270X, the extra memory and cost are not worth it.

Happy Gaming Sir , thanks for the compliment

Most welcome! 8) Not forgotten about you btw, just been loaded down with work stuff this week.
I'll be setting up the test board this afternoon, gotta vote in a referendum in the evening, so I'll
do some testing after that (first need to clone the boot SSD, change the drivers, get the 7970 up
& running). Hmm, anyone know if there's a bus width difference between the 280X and 270X?
If so, then downclocking a 7970 wouldn't fairly represent 270X performance. I'll look into it.

Note I'm ussing an 'average' mbd (Asrock P55 Extreme), so it should if anything underestimate
what you could do with your Sabertooth. And as mentioned I'll be using an old TRUE cooler with
two typical decent (but not top-end) fans, namely Gelid Wing Blue UV 120mm 3-pin. Likewise,
I'll set the RAM at medium speed even though it could go higher. One caveat: I'll leave the i5 760
at its default 2.8GHz for the stock-speed test; slightly more than the 750, but it shouldn't matter.
As for the oc, I'm aiming for between 4 and 4.4, depends what the board can handle (I know the
P55 Deluxe can run this i5 at 4.4, but the Extreme's power phases aren't as sophisticated). Oh,
the RAM is Mushkin Redline 2x4GB, capable of 2133, but it'll be set for more like 1600 or thereabouts,
which helps prevent the IMC being stressed too much, lower VTT, minimises heat, easier CPU oc,
better overall speed. The SSD is an OCZ Vertex2E 120GB, PSU is my standard testing unit, overkill
for this, but the same for most testing (Thermaltake Toughpower XT Gold 1475W), and the OS is
Win7/Pro/64.

Btw, if you know of any freely available benchmark demos of racing-type games, do let me know,
happy to try them out. None of my normal tests are racing games, but I do use a selection which
is able to show a wide variety of behaviour.

And Jean, SSDs on SATA2 platforms actually work really well. 8) I tested this, the speedups are
still very nice. Here's some SATA2/SATA3 chipset comparison data.

However, work stuff first, have to pack up a 5GHz 2700K system to send to an engineering company...

Ian.

More later (still sorting out the system), but just wanted to say, you really won't shorten the life of the CPU
at all if you limit the oc to whatever can be obtained with a low vcore of something like 1.35V. Even 1.4 to
1.45 is not that high for a Lynnfield, and many samples will happily run at 4.0 with as little as 1.3V.

I don't think I've ever seen a PCIe voltage issue, and I've meddled with a lot of P55/H55 boards (atm I have
nine of them).

Re the cards, personally I wouldn't bother with the 570 or 7950, the specs drop isn't worth it (eg. lower VRAM
with the 570). If you want to go that route, try to find either a 3GB 580 or a GHz Edition 7970. The 580 will be
cheaper, and IMO you'll have less hassle with the drivers (though as others have said it's fair to say the 7970
ought to be equally ok if used just on its own), but the 7970 is certainly faster, though I must confess I was
surprised at how big the 7970s are, actually larger than the MSI 580 3GB LX (or put another way, freakishly
huge when sat next to a Palit 460. I mean if you've not seen such a card before, at first it seems
like your brain is gawping at something like this! *grin*).

The last 3GB 580s I bought cost around 100 UKP each (Palit 783MHz). The 7970s I bought (1050MHz Sapphire and
1150MHz MSI) were about 145 UKP each. I also bought some top-end 832MHz MSI 3GB 580 LX cards for my AE system
and general testing, these were somewhat more, between 135 and 175 (I obtained six of them in all; these cards are
crazy, they oc so easily, 900MHz to 1GHz no problem with most samples - I've run mine at 950 max so far).

1.5GB 580s will cost even less of course, these days around 80 UKP or less in the UK. Fine if you're
only gaming at HD or less, but I'd still recommend hunting for the 3GB type if you can.

Quick note: the standard 3GB 580s like the Palit are definitely quite loud under heavy load (the MSI LXs are
much better), but I was surprised the 7970s didn't really make that much less noise.

Sometimes unusual cards show up, eg. llike this (580 with custom cooler, so much less noise):

http://cgi.ebay.co.uk/ws/eBayISAPI.dll?ViewItem&item=231335305851

Only thing I would say though is always buy from a seller which accepts returns, unless they
really do have a thoroughly glowing +ve and high feedback.

Ian.

PS. Oddly enough someone here listed two of the MSI LXs recently (231331994796 and 231331994796), though
I don't think I'd bid quite as much as that now (I obtained most of mine last year when newer cards cost a lot more);
for a single card a 7970 is better value.

Let's be honest , your CPU isn't really bad at all but isn't as cutting edge as the newer generations that's the problem.

Also bare in mind 3dmark11 isn't a real-world game and in real world gaming you're more likely to see some bottlenecks however if you plan on upgrading your CPU next year then your CPU will handle the GTX 970 and you'll see massive improvements over what you have currently however I'd recommend overclocking since you're upgrading anyways and its a risk worth taking.

I know that you asked Ian and not myself but I liked to help you till he responds with a comment disagreeing with me ( JK , We don't always agree but I respect his opinions )

N.B : AMD aren't a small company and while Nvidia is currently better in the Higher end card range however that doesn't mean that AMD's mid-range cards suck too , so think twice about the available funds and enjoy your upgrades

TopLuca makes some good points, and the 970 would certainly be a worthy upgrade; any future
build to which you might switch would cerainly be able to exploit it nicely. Personally, if I could
afford it in your position, I would indeed get a 970 (pity about the used pricing, but never mind;
at least with something new like a 970, you won't have any noise/power issues).

Apologies for not replying with an update sooner! I was really busy this past week. I have however
done the 7970 tests with a stock i5 760, which showed what I expected, namely GPU-bound tests
were almost identical compared to a 7970 running with a 5GHz 2700K (eg. Call of Juarez and Unigine),
while more CPU-dependent game tests (Far Cry 2, Stalker) were indeed slower than the 7970/2700K
combo, but not by as much as one might expect.

In general, older CPU-dependent games are so fast with newer cards like the 7970 that, at lower
resolutions, and despite quite large frame rate differences, the absolute frame rates are so high it
means one would never notice the difference in practice (you'll see what I mean when I've uploaded
the FC2/Stalker data). At higher resolutions, the differences shrink, but performance is still good overall.

Though I've not yet done the 7970 tests with an oc'd i5 760, in the morning I'll uploaded the results
obtained so far so you can see how a 7970 behaves with the i5 760 at stock speed. Many reviews of
the GTX 970 do include a 7970 or 280X (same thing) in their chart results, so the data should prove
useful. I hope to do the overclocked tests this week.

Ian.

PS. Be careful of 3DMark11. Just as with the old 3DMark06, modern CPU/gfx combos tend to produce
significantly higher overall results just via the use of a more powerful CPU, ie. it's possible to have
system A give a higher overall result than system B, even though system B shows higher frame rates
in the game tests than system A. I've seen this again & again with the testing I've been doing.

Update! I've uploaded stock-speed i5 760 results with 7970 and 7970/CF to my site (except for 3DMark,
doing that shortly, ie. 3DMark06, Vantage, 3DMark11, Firestrike, etc.) I don't yet have pages for most of
the 3DMark data I've accumulated, but I'll include the links here when they're ready; atm my site only has
a page for the '06 results). See:

http://www.sgidepot.co.uk/sgi.html#PC

Thus, check the data for Stalker COP, Unigine, X3TC, Call of Juarez, Far Cry 2 and PT Boats.
NB: remember the 7970 GHz Edition is basically the same as a 280X. I have two GHz Edition
7970s, one runs at 1050MHz, the other normally at 1150 but I force it to 1050 in CF so it
matches the other card.

Results are much as I expected with a 7970. At low res/detail, CPU speed does make a difference,
but the fps are so high it doesn't matter anyway. At high res/detail, the gap narrows or vanishes
completely (sans dodgy Unigine/Stalker results because of the DX9/CF issue), yet performance is
still good overall.

DX10/DX11 tests show the expected scaling, with Stalker @ 2560x1440 high-detail 7970/CF very
close to the speed of the same 7970/CF with a 5GHz 2700K. This is what I meant about the way
many games become less & less bottlenecked by CPU power as the resolution/detail/etc. increases.
The gap is a little wider for Ungine Valley @ 1920x1080 DX11, so I expect an oc'd i5 to help there
(notice how the gap shrinks when moving to 2560x1440). Some newer games can behave a bit
differently, but that's less common.

FC2 at 1920x1200 does show a gain from two 7970s even with the i5 @ stock speed, but the gap
between it and the 2700K suggests this a typical example where having the i5 oc'd would help, though
note (as above) the performance is so high one wouldn't really notice. At 2560x1440 the CF scaling gap
increases, which is good, and the divide between the i5 and 2700K sihrinks somewhat, but it's still
significant. Oc'ing the i5 would help, but again the fps is so high one would not notice in practice, except
of course for the larger degree of stutteriing which can occur with multiple GPUs for some tests (this
issue is very noticeable for the high-res Unigine Heaven/Valley tests).

Call of Juarez at low res shows a gap dependent on CPU power, but as above the absolute speed is
so high it doesn't matter. Move up to 1920x1080 though and everything changes. Now the CF scaling
is working very well; the i5 isn't far behind the 2700K, and more interestingly their results for just one
7970 are almost identical. COJ is the classic case where increasing res & detail very quickly moves the
bottleneck onto the GPU. Alas I don't know how to run this test beyond 1920x1200 (doesn't give me
the option of testing any higher), so I don't yet know how performance would scale up to 2560x1440.

PT Boats is an odd one, it seems to be CPU-limited at just about any resolution, perhaps because it
employs lots of physics calculations which normally would be accelerated via PhysX on an NVIDIA card,
but of course with a 7970 it has to be done by the main CPU.

Tropics seems to have CPU-related issues at low res or with DX9, but these factors begin to fade
at HD res and higher. Indeed, at 2560x1440 with DX10, the i5 760 beats the 2700K! (TBH it's
basically exactly the same, ie. well within error margins, but the result is nonetheless amusing,
showing how the GPU-heavy nature of Unigine sometimes only reveals itself when under heavy load).

I think the Heaven results show some potential for an oc'd CPU helping out even at medium detail,
it will be interesting to see what happens with the oc'd setup.

One thing is clear though (and the 3DMark data will likely back this up when I have the numbers),
a P55 system is perfectly capable of exploiting powerful cards quite well, much more so than
most people would believe, and remember most P55 boards are only running with x8/x8 (similar
limited PCIe issue to modern Z97, though at least the newer chipsets are V3, so x8/x8 on a Z97
is akin to x16/x16 on an older board). I do have a P55 board which supports x16/x16 (and also
x8/x8/x8/x8; it's an ASUS P7P55 WS Supercomputer, employs PCIe switches in the usual manner),
but I've not done much with it yet, and I won't use it for most results because such data wouldn't
be representative of most P55 boards.

Anyway, I'll sort out the oc later this week and update the tables with oc'd numbers. Can't do it
sooner, got to do some work stuff first.

Hope this helps!

Ian.

Hats off Ian

I must again stress that you simply overclock your cpu. Neither of those cards would be bottlenecked in any fashion if you were to push for something moderate like 3.4Ghz.

TopLuca writes:
> Hats off Ian

Thanks! Various 3DMark results coming shortly...


spentshells writes:
> I must again stress that you simply overclock your cpu. Neither of those cards would be bottlenecked in
> any fashion if you were to push for something moderate like 3.4Ghz.

Read my earlier posts, I've covered this already in great detail. Of course I'm going to do oc'd tests aswell,
but the whole point here is to first get a baseline with the CPU at stock speed so that one can compare, see
how different games/tests vary in their response to a faster CPU. Have a look at my data, some games like
CoJ are less dependent on CPU power, others like Stalker and FC2 show mixed behaviour, with the CPU being
less relevant as the resolution/detail increases, though with games like these, even one modern card is so
fast that CPU-based differences tend to be irrelevant anyway.

Games like X3TC are rare, ie. entirely CPU dependent (single threaded I believe, SLI/CF not supported). X3TC behaves
a bit like 3DMark06 wrt to CPU speed.

Last night I did some Crysis testing. Annoyingly the benchmark tool does not offer a 2560x1440 option, so I'm looking
into alternative benchmarking methods. I had hoped to do tests that one could directly compare to existing data on
toms reviews, etc., but the old site-review tradition of using 1920x1200 and 2560x1600 may make that impossible -
back then was before the market shift to 1920x1080 and 2560x1440. I also found that changing test settings didn't
really seem to affect performance that much, so I'm not sure the benchmark tool is running the tests correctly.

I had a look for newer downloadable benchmark demos, but couldn't find any. Anyone know of something relevant?
I tried the Resident Evil 6 demo, but it's rubbish. Seems like the older tradition of releasing game demos that can be
used as benchmarks has kinda faded away.

Is there a standard Crysis2 benchmark demo? I could use that I suppose as a slightly more up to date game. I do
have the full game, so no problem there.

If anyone knows of a downloadable standalone racing game benchmark, let me know.

Ian.