Tech-Hounds.com

Looking at the Intel P35 Express chipset

The next few months left in 2007 will certainly be anything but dull. First on the list is the launch of AMD's very late Barcelona processors. Hopefully, we're going to see desktop variants by the end of the year. AMD's graphic division is aslo scheduled to release a refresh of their Radeon HD2900XT. On the other side of the fence, rumor has it NVIDIA is already prepping up G92 for November launch. And of course, semiconductor giant hopes to maintain their lead with the release of Yorkfield and Wolfdale.

Out of these three manufacturers, Intel is by far the one with the clearest roadmap. Dual core Wolfdale and the quad core Yorkfiled, desktop members of the new Penryn family of processors are refreshes of Intel's very successful Core 2 Duo family of processors. Intel announced these processors several months ago, offering juicy tidbits - performance improvements between 10 to 40 percent with certain application workloads compared to existing Core 2 Duo processors, the use of 45 nm high k manufacturing and expected spees in excess of 3 GHz.

With the annoucements of new processors Intel also made announcements of new chipsets.  Generation 3 Series chipsets include Q33, Q35, G31, G33, G35, P31, P35 Express chipsets. However, unlike previous products, these chipsets (atl least some of them) and motherboards using them are already on the market for tsome time. First announced at Computex this year, these chipsets are successor to Intel's own P965 and G965 chipsets. What sets apart these chipsets from the previous generation are support for 1333 MHz bus, a reworked memory controller allowing DDR3-SDRAM support (only for P35 and G33 at this time) and of course, support for the newly announced 45 nm Penryn family of processors.

Today, we're going to take a look at a member of the Generation 3 Series - the Intel P35 Express chipset.

Setting Up

Thanks to Gigabyte, we managed to get samples of two boards, the Gigabyte P35-DS3P with DDR2 and P35-DQ6 for this test. For comparison, we're going to use our trusty Gigabyte DS3P to represent P965. We flash the BIOS of each respective boards to their latest, final version BIOS.Since all three boards came from the same manufacturer, we were able to use the same BIOS settings for all boards, not to mention the same drivers. This should minimize other influencing factors such as different hardware / drivers etc.

Our test setup
Intel Core 2 Duo E6300 and E6550 socket LGA-775
4 x 512 MB A-DATA Vitesta 5-5-5-18 PC6400 DDR2-SDRAM
2 x 1024 MB Samsung  7-7-7-20 PC8500 DDR3-SDRAM
Gigabyte Radeon X1950 Pro 256 MB graphics card
Gigabyte P965-DS3P Intel P965, P35-DS3P Intel P35 (DDR2) and P35-DQ6 Intel P35 (DDR3) motherboard
Hitachi Deskstar 250 GB Serial ATA 16 MB buffer 7200 rpm hard drives
LiteOn 1673S DVD-RW
FSP Epsilon 800 watts ATX/BTX power supply

Settings:
Core: 1.25 volt
DDR2: 1.9 volt (default motherboard voltage)
DDR3: 1.5 volt (default motherboard voltage)
PCI-E clock: 100 MHz
DDR clock: 800 MHz for DDR2 (except when testing E6550 with P965 - memory is synchronous with FSB)
Memory Timings: SPD except otherwise specified

As we were setting up, we found that the Gigabyte P35-DS3P we're using won't boot at all with all memory slots populated. We can see and hear fans spinning up, lights coming on but no BIOS screen. So we removed two modules and wala, the board boots and everything work properly. From experience, we usually see this kind of behaviour if the memory controller is set to use Command Rate 1T. Unfortunately, we were unable to check if P35 uses Command Rate 1T by default, however we were able to confirm that P965 does use Command Rate 2T if all slots are populated. We did update to a newer BIOS, but no dice.

Well, what's wrong with 1 GB of RAM? The problem is that you really want 2 GBs of RAM with Vista. With just 1 GB of memory, Vista downgrade memory performance of our system from 5.4 to 4.5, most likely since it has to fetch data from virtual memory / page file more often. It shouldn't be a problem with small benchmarks like Sciencemark and SuperPI whose memory usage are fairly low, but it can be a problem with games who uses more than 1 GB of memory. Though performance will not be different, we'll have to run more test to 'buffer' data in RAM. Vista's user address space problems also didn't make it any easier. This means we had to look around for some 1 GB memory modules.

Thankfully we came across a pair from TeamExtreme, on loan from a friend. One thing we noticed is that these modules are also double bank / double sided, so even we were to use four of this modules together with the P35, we'll likely experience the same problem again. Single sided memory modules usually costs extra and sometimes uses higher latency. P35 insistence on single sided modules is really disappointing, considering no such restrictions with P965. Just to be clear, we're still not sure whether or not this is the nature of the P35 or the motherboard, but we are completely sure such behaviour does not exists with P965 boards we tested. We'll have to see until our P35 motherboard round up to see whether or the board or the chipset is to blame here.

We did some test to see whether or not there are performance differences between 1 GB and 2 GB and found the difference to be so small that they can be attributed to variations between runs. We saw the same thing during testing for the article Quest for Lag Free Gaming.

Preliminary Tests

Much like our last article looking at Core 2 Duo E6300, we're going to use some synthetic benchmarks to examine the memory subsystem bandwidth and latency: SuperPi 1.1 (no mod) and Sciencemark 2.0's Membench. Keep in mind, these are only preliminary tests results, from which we're hoping to see whether or not there are differences between the two chipset (P965 and P35) with basically the same memory (and timing) and then once again with DDR3 memory modules.

Here are the Core 2 Duo E6300 and E6550 processors we used today

 

Here are the SPD data from the A-DATA Vitesta modules

 

And here are the SPD data from the DDR3 Samsung modules we used in this test

 

Bandwidth & Latency

First, let's look at latency. One thing we saw with our last article is that generally latency gets higher with higher clock - thankfully bandwidth also goes up. We ran tests under several conditions, using different FSBs - 266 MHz (default on the E6300), 333 MHz (default on the E6550) and 400 MHz. We also change the processor multipliers - from experience, using a higher multiplier help reduce latency (somewhat). For this test, we've decided to let the motherboard apply SPD timing values, but unlike our last article we set the memory to 800 MHz (or as close as possible to 800 MHz)- which means its running asynchronous at 266 and 333 MHz.

Core 2 Duo E6300 & P965

	7x266 MHz	7x333 MHz	6x400 MHz	7x400 MHz
Bandwidth	5151.21 MB/s	5490.95 MB/s	5893.33 MB/s	6355.65 MB/s
Latency
4 byte stride	3 cycles	3 cycles	3 cycles	3 cycles
16 byte stride	6	8	9	9
64 byte stride	26	32	35	35
256 byte stride	87	111	128	124
512 byte stride	99	126	147	141

Core 2 Duo E6300 & P35 - DDR2

	7x266 MHz	7x333 MHz	6x400 MHz	7x400 MHz
Bandwidth	5043.08 MB/s	5663.68 MB/s	5947.79 MB/s	6285.54 MB/s
Latency
4 byte stride	3 cycles	3 cycles	3 cycles	3 cycles
16 byte stride	6	8	9	9
64 byte stride	26	31	36	35
256 byte stride	88	110	128	127
512 byte stride	99	128	148	147

As expected, increased FSBs means higher latencies and higher bandwidth. Latency increases with increased FSBs are pretty different between the P965 and P35. Look at latencies on 256 byte stride on the table above - with the same multiplier, the P965 jumped from 87 to 111 to 124 (24 and 13 cycles, respectively). The P35 has different 'jumps' - 88 to 110 to 127 (22 and 17 cycles respectively). The same holds true to 512 byte stride as well.

At 7x400, the P35 has slightly higher latencies but surprisingly less bandwidth than the P965. Notice the increase in bandwidth is slightly different on both chipsets. On the P965, we saw an increase of around 6 percent (266 to 333 MHz) to 15 percent (333 to 400 MHz), but on the P35 its more like 12 percent (266 to 333 MHz) and 10 percent (333 MHz to 400 MHz). Now, you might say - isn't that bad? Well, not necessarily. Intel very likely tweaked the P35 memory controller so it behaves more 'predictably' with FSBs higher than 400 MHz and also works better with asynchronous memory settings - important for DDR3 modules.

Core 2 Duo E6300 & P35 - DDR3

	7x266 MHz	7x333 MHz	6x400 MHz
Bandwidth	4893.08 MB/s	5457.03 MB/s	5849.65MB/s
Latency
4 byte stride	3 cycles	3 cycles	3 cycles
16 byte stride	6	8	9
64 byte stride	25	32	36
256 byte stride	85	111	129
512 byte stride	96	130	153

We didn't ran 7x400 since from the results, it's already apparent that strapping a pair of DDR3 modules isn't going to offer much improvement. Although the DDR3 we're using ran at slightly higher clocks (480 MHz) than our DDR2 PC6400 (400 MHz), the higher latencies of the modules kept it from offering more bandwidth. Keep in mind that that official DDR3 spec memory runs at 533 MHz (1066 MHz effective). It's possible at that speed the situation is different - the additional clock should mean latencies are comparable to DDR2 800 with lower latencies.

Core 2 Duo E6550 & P965

	7x333 MHz	6x400 MHz	7x400 MHz
Bandwidth	5829.77 MB/s	5849.68 MB/s	6299.01 MB/s
Latency
4 byte stride	3 cycles	3 cycles	3 cycles
16 byte stride	8	9	9
64 byte stride	30	36	35
256 byte stride	108	132	127
512 byte stride	124	150	146

Core 2 Duo E6550 & P35 - DDR2

	7x333 MHz	6x400 MHz	7x400 MHz
Bandwidth	5747.82 MB/s	5986.77 MB/s	6324.57 MB/s
Latency
4 byte stride	3 cycles	3 cycles	3 cycles
16 byte stride	8	10	9
64 byte stride	31	42	35
256 byte stride	106	129	124
512 byte stride	129	144	139

The primary difference between the E6300 and E6550 is cache size. With higher FSBs and memory running synchronously, cache becomes less important because the penalty of fetching data from memory is smaller. However, the opposite is also true - a large cache (with the same clock and timing) should be more important when there is a difference between memory and FSB bandwidth - like in an asynchronous memory environment. At least, that's the theory. Do remember that we had to use synchronous memory settings for the E6550 with the P965. On the P35, the DDR2 memory is running at 400 MHz (actually 415 MHz at 333 MHz FSB)  but timings should not change much.

We ran the E6550 from 333 MHz upwards since the processor's default FSB is 333 MHz - no sense in running the processor slower than default. Now, lets look at latencies on different FSBs without changing multipliers. At default FSB of 333 MHz, the P965 generally has higher  latencies - we expected that after looking at the E6300 results. On a side note, compare the bandwidth of a synchronous bus - memory on the P965 and asynchronous bus - memory on the P35. Synchronous is still the best way to go. But at 400 MHz, running synchronously, the P35 is the one with lower latencies. Comparing bandwidth increases, the P965 offers about 8 percent increase (from 333 MHz to 400 MHz) while the P35 is about 10 percent. Latencies 'jumps' are 108 to 127 (19 cycles) for 256 byte stride and 124 to 146 (22 cycles) for 512 byte stride on the P965, while on the P35 its 106 to 124 (18 cycles) for 256 byte stride and 129 to 139 (10 cycles) for 512 byte stride.

This is where it's get interesting. You'll no doubt notice bandwidth increases (in percent) are actually smaller than what we saw with E6300 with the P35 (10 - 12 percent) and P965 (6 - 15 percent). Also, with the E6300 latencies jumped 13 cycles with P965 and 17 cycles with P35 for 256 byte stride. That's, well, that's a 6 cycle difference on the P965 and 1 cycle on the P35 with 256 byte stride and 7 and 9 cycles with the bigger, 512 byte stride. The kicker? P35 512 byte stride latency is LOWER by 9 cycles.Asynchronous / synchronous settings are not the factor here, since we're comparing results on the same chipset but different processors - both were run asynchronously at 333 MHz. With the FSB effectively the same, we wondered if cache is the only difference between these two processors. If cache is the differentiating factor, we wondered how the same P35 will handle Wolfsdale's and Yorkfield's bigger yet faster (lower latency) cache .

Core 2 Duo E6550 & P35 - DDR3

	7x333 MHz	6x400 MHz	7x400 MHz
Bandwidth	5541.27 MB/s	5685.75 MB/s	6217.81 MB/s
Latency
4 byte stride	3 cycles	3 cycles	3 cycles
16 byte stride	8	9	9
64 byte stride	32	37	35
256 byte stride	111	130	125
512 byte stride	127	150	145

Well, no surprises here. DDR3's higher latency means we're getting slightly less bandwidth than if we were to pair the P35 chipset with lower latency DDR2 modules (at relatively the same clock). 

SuperPi 8M Results

Generally speaking, latencies between the two chipsets - P965 and P35 - are pretty similar.  at default FSBs of each respective processor. Twiddle a bit with higher FSBs, and we began to see some interesting twists.P35 seem to favor large cache processors, according to the results we saw with our Sciencemark Membench. Naturally we are intrigued - how does this peculiar behavior in other applications. We turned to SuperPi 8M for our second test.

Core 2 Duo E6300 & P965

	7x266 MHz SPD	7x333 MHz	6x400 MHz	7x400 MHz
Iterations	seconds	seconds	seconds	seconds
1	15	13	13	11
2	30	25	24	21
3	44	37	36	31
4	58	49	47	42
5	72	61	59	52
6	86	73	70	62
7	100	85	82	72
8	114	97	94	82
9	128	109	105	93
10	142	121	117	103
11	156	132	128	113
12	170	145	140	123
13	184	157	151	133
14	198	168	163	144
15	212	180	174	154
16	226	192	186	164
17	240	204	197	174
18	254	216	209	184
19	268	228	220	194
20	282	239	231	204
21	295	251	242	214
22	307	261	252	223

Core 2 Duo E6300 & P35 - DDR2

	7x266 MHz SPD	7x333 MHz	6x400 MHz	7x400 MHz
Iterations	seconds	seconds	seconds	seconds
1	16	13	13	11
2	30	25	25	21
3	44	37	36	31
4	58	49	47	41
5	72	61	59	51
6	86	72	70	61
7	100	84	81	72
8	114	96	93	82
9	129	108	104	92
10	143	120	116	102
11	157	132	127	112
12	171	143	138	122
13	185	155	149	132
14	199	167	161	142
15	213	179	172	152
16	227	191	183	163
17	241	202	195	173
18	255	214	209	183
19	269	226	217	193
20	283	237	228	202
21	296	248	239	212
22	308	259	249	221

Core 2 Duo E6300 & P35 - DDR3

	7x266 MHz SPD	7x333 MHz	6x400 MHz	7x400 MHz
Iterations	seconds	seconds	seconds	seconds
1	15	13	13	11
2	30	25	24	21
3	44	37	35	31
4	58	49	47	42
5	72	61	59	52
6	86	72	70	62
7	100	84	81	72
8	114	96	93	83
9	128	108	104	93
10	142	120	116	103
11	156	132	127	113
12	170	144	139	123
13	184	155	150	134
14	199	167	162	144
15	213	179	173	154
16	227	191	184	164
17	241	203	196	174
18	255	214	207	184
19	268	226	219	194
20	282	238	230	204
21	295	249	240	214
22	308	259	250	223

The difference is small, but its there even at default clocks. As we increase FSB and clock, the difference increased from 1 to 2 / 3 seconds. Much like we saw with Sciencemark Membench, we did not see much of a difference between P965 and P35 with the Core 2 Duo E6300. Now, let's see how does the same test do fare with E6550.

Core 2 Duo E6550 & P965

	7x333 MHz	6x400 MHz	7x400 MHz
Iterations	seconds	seconds	seconds
1	13	13	11
2	25	24	21
3	37	36	31
4	48	47	41
5	60	58	51
6	72	70	61
7	83	81	71
8	95	93	81
9	107	104	91
10	119	115	101
11	130	127	111
12	142	138	121
13	154	149	131
14	165	161	141
15	177	174	151
16	189	183	161
17	200	195	171
18	212	206	181
19	223	217	191
20	235	228	201
21	246	239	210
22	266	249	219

Core 2 Duo E6550 & P35 - DDR2

	7x333 MHz	6x400 MHz	7x400 MHz
Iterations	seconds	seconds	seconds
1	13	12	11
2	24	23	21
3	36	35	31
4	47	46	41
5	59	57	51
6	71	68	61
7	82	79	72
8	94	91	82
9	105	102	92
10	117	113	102
11	129	124	112
12	140	138	122
13	152	147	132
14	164	158	142
15	175	169	152
16	187	180	163
17	198	191	173
18	210	202	183
19	221	213	193
20	233	224	202
21	244	235	212
22	254	245	221

Core 2 Duo E6550 & P35 - DDR3

	7x333 MHz	6x400 MHz	7x400 MHz
Iterations	seconds	seconds	seconds
1	13	12	11
2	24	24	21
3	36	35	31
4	48	47	42
5	59	59	52
6	71	69	62
7	83	81	72
8	94	92	83
9	106	103	93
10	118	115	103
11	129	126	113
12	141	138	123
13	153	149	134
14	164	160	144
15	176	172	154
16	188	183	164
17	199	194	174
18	211	205	184
19	222	217	194
20	234	228	204
21	245	238	214
22	255	248	223

Needless to say, the results here are pretty interesting. Look at the difference at E6550 default FSB between the P965 and P35 - thats 11 to 12 seconds, definitely far bigger than the 2 second difference we saw between both chipsets with E6300 on the same speed (7x333 MHz). The lack of asynchronous memory setting support for 1333 MHz FSB processor on the P965 really hurt its performance. At 333 MHz the difference is smaller - only about 4 seconds. At 400 MHz FSB, we saw the difference between P965 and P35 dwindle to 2 seconds. At this speed, the E6300 at 2.8 GHz is actually neck to neck with the E6550 at the same clock - if you're using a P35 equipped motherboard with DDR2 and DDR3. The fastest combination? The E6550 with P965 - 2 seconds faster at 219 seconds.

FSB Higher than 400 MHz

We saw telltale signs of memory controller tweaks and perhaps others with Sciencemark Membench, but our SuperPi 8M results seem to show that performance wise, the P35 is no faster than its older sibling, P965. The only new 'wrinkle' it has for the 1333 MHz processor is the (re)introduction of asynchronous memory support - which is not all that great anyway. We still saw that synchronous memory settings is still the way to go with Intel chipsets (memory controller). Is this all there is to it? Well, not quite. The major concern with P965, especially for overclockers is its well known FSB limitation. The P965 was clearly not designed to go far beyond 400 MHz. From what we saw in our article concerning the matter, we think the P965 chipset simply couldn't 'loosen' internal chipset timings to handle FSBs higher than 400 MHz. One of the key features Intel promised overclockers everywhere with P35 (and X38) is no more FSB limits.

Let's see if this is true. Unfortunately, we only managed to top off at 466 MHz with this setup. Higher FSBs are definitely possible on the P35, but we lack the necessary processor cooling, so we'll have to settle for 466 MHz or 1864 MHz effective according to Intel's FSB mantra. We used TeamExtreme DDR2 1200 MHz 1 GB memory modules for this test, rated at 5-5-5-15 at 1200 MHz with 2.35 - 2.45 volts. Since 6 x 466 MHz still equals to 2.8 GHz, we kept the processor's voltage, but raised FSB, MCH and PCI-E voltage until the setup was stable (enough for our preliminary tests).

Core 2 Duo E6550 & P965

	6x400 MHz	7x400 MHz	6x466 MHz
Bandwidth	5849.68 MB/s	6299.01 MB/s	6594.95 MB/s
Latency
4 byte stride	3 cycles	3 cycles	3 cycles
16 byte stride	9	9	11
64 byte stride	36	35	38
256 byte stride	132	127	141
512 byte stride	150	146	157

Core 2 Duo E6550 & P35 - DDR2

	6x400 MHz	7x400 MHz	6x466 MHz
Bandwidth	5986.77 MB/s	6324.57 MB/s	6835.86 MB/s
Latency
4 byte stride	3 cycles	3 cycles	3 cycles
16 byte stride	10	9	10
64 byte stride	42	35	40
256 byte stride	129	124	129
512 byte stride	144	139	150

In addition to overall better stability at this FSB, the P35 also offer higher bandwidth than the P965. Interestingly, latencies are lower - 129 cycles to 141 cycles for 256 byte stride and 150 cycles to 157 cycles on the P965. How does it perform on SuperPi 8M? Let's find out.

Core 2 Duo E6550 & P965

	6x400 MHz	7x400 MHz	6x466 MHz
Iterations	seconds	seconds	seconds
1	13	11	11
2	24	21	21
3	36	31	30
4	47	41	40
5	58	51	50
6	70	61	60
7	81	71	70
8	93	81	80
9	104	91	90
10	115	101	100
11	127	111	110
12	138	121	119
13	149	131	129
14	161	141	139
15	174	151	149
16	183	161	159
17	195	171	169
18	206	181	179
19	217	191	189
20	228	201	198
21	239	210	208
22	249	219	216

.
Core 2 Duo E6550 & P35 - DDR2

	6x400 MHz	7x400 MHz	6x466 MHz
Iterations	seconds	seconds	seconds
1	12	11	11
2	23	21	20
3	35	31	30
4	46	41	40
5	57	51	50
6	68	61	60
7	79	72	70
8	91	82	79
9	102	92	89
10	113	102	99
11	124	112	109
12	138	122	119
13	147	132	129
14	158	142	139
15	169	152	149
16	180	163	158
17	191	173	168
18	202	183	178
19	213	193	188
20	224	202	197
21	235	212	207
22	245	221	215

At 466 MHz FSB, the additional bandwidth on the P35 allows it to finish SuperPi 8M calculation faster by 1 second. Well, P35 does offer more performance on higher FSBs and there's likely more frequency headroom if you're lucky or brave enough. That's good news for overclockers dissatisfied with P965 FSB wall.

Performance

Although we did get some useful information from our preliminary tests, we'd still like to see whether or not they  translate into more applicable situations - games. So we ran some of our usual game benchmarks at default settings - 1024 x 768 without AA and AF. These tests are ran under three different configurations - 7 x 400 MHz, 6 x 400 MHz and 7 x 333 MHz. This is the order the results are arranged - the first group is results we got with 7 x 400 MHz and so on. An extra benefit of doing these tests is that we can see how much of an impact lowering processor clock and FSB have on performance (with these benchmarks).

Here is again our test setup - it's still the same setup as before.

Our test setup
Intel Core 2 Duo E6300 and E6550 socket LGA-775
4 x 512 MB A-DATA Vitesta 5-5-5-18 PC6400 DDR2-SDRAM
2 x 1024 MB Samsung  7-7-7-20 PC8500 DDR3-SDRAM
Gigabyte Radeon X1950 Pro 256 MB graphics card
Gigabyte P965-DS3P Intel P965, P35-DS3P Intel P35 (DDR2) and P35-DQ6 Intel P35 (DDR3) motherboard
Hitachi Deskstar 250 GB Serial ATA 16 MB buffer 7200 rpm hard drives
LiteOn 1673S DVD-RW
FSP Epsilon 800 watts ATX/BTX power supply

Settings:
Core: 1.25 volt
DDR: 1.9 volt (default motherboard voltage)
DDR3: 1.5 volt (default motherboard voltage)
PCI-E clock: 100 MHz
DDR clock: 800 MHz for DDR2 (except when testing E6550 with P965 - memory is synchronous with FSB)
Memory Timings: SPD except otherwise specified

Windows Vista Home Premium Edition with the latest updates installed
ATI Catalyst 7.8 reference driver
Intel Chipset Software Installation Utility 8.3.0.1011
Realtek Semiconductor High Definition Audio System Software Ver:R1.72
DirectX 9.0c
all respected games used for benchmarks have been updated to their latest, final builds.

The results:

P35 to P965 - DDR2	E6300			E6550
	7x400	6x400	7x333	7x400	6x400	7x333
	Difference	Difference	Difference	Difference	Difference	Difference
Min	0.82%	2.00%	0.00%	-0.80%	7.21%	8.57%
Avg	1.52%	0.56%	2.14%	0.80%	2.86%	3.58%
Max	2.66%	1.92%	2.40%	4.98%	7.70%	9.67%

This game maybe old, but it is one game that have proven sensitive to FSB increases. The graph does show this, but as we can see this happens mostly on maximum fps. We'll have to dig a little deeper and look at the scores. The table above shed some additional light on the situation. From looking at average and minimum frame rates, the difference between the P965 and P35 is small if you pair it with the E6300. With the E6550, it's a different story. Remember, the larger difference at 7x333 MHz with the E6550 is likely due to synchronous memory settings - the memory is only running at DDR2 667 MHz rather than its rated speed of DDR2 800 MHz. However, that doesn't explain the difference at 6 x 400 MHz, which is still on the same level as the 7 x 333 MHz results. So, chances are this behavior is not entirely related to asynchronous memory settings.

P965	E6300		E6550
	7x400	6x400	7x400	6x400
	Compared to 7x333	Compared to 7x333	Compared to 7x333	Compared to 7x333
Min	31.18%	7.53%	19.05%	5.71%
Avg	6.95%	3.45%	6.66%	1.87%
Max	13.03%	4.61%	23.16%	4.36%
P35	E6300		E6550
	7x400	6x400	7x400	6x400
	Compared to 7x333	Compared to 7x333	Compared to 7x333	Compared to 7x333
Min	32.26%	9.68%	8.77%	4.39%
Avg	6.30%	1.85%	3.79%	1.17%
Max	13.31%	4.11%	17.89%	2.48%
P35 DDR3	E6300		E6550
	7x400	6x400	7x400	6x400
	Compared to 7x333	Compared to 7x333	Compared to 7x333	Compared to 7x333
Min	29.17%	7.29%	11.61%	5.36%
Avg	6.80%	1.99%	4.36%	0.90%
Max	11.65%	3.11%	18.15%	2.35%

The E6300 with the smaller cache appreciates the additional bandwidth from overclocking the FSB far greater than the E6550 on both the P35 and P965 chipset. This is expected since its very likely have to fetch data from memory more often (due to its smaller cache). The numbers on the table above illustrates this better than the graph earlier. Of course, this are in percentages - the additional frame rates are nice to have but we actually don't really need it anymore since the game is already very fast for fluid gameplay.

P965	E6550 to E6300
	7x400	6x400
	Difference	Difference
Min	2.46%	11.00%
Avg	9.20%	7.83%
Max	60.28%	46.74%
P35	E6550 to E6300
	7x400	6x400
	Difference	Difference
Min	0.81%	16.67%
Avg	8.43%	10.30%
Max	63.90%	55.08%
P35 DDR3	E6550 to E6300
	7x400	6x400
	Difference	Difference
Min	0.81%	14.56%
Avg	8.45%	9.79%
Max	66.43%	56.12%

In this table, we compare the E6550 performance relative to the E6300. We like to see if the behavior we saw with Sciencemark's Membench and SuperPi manifest again in this scenario. Of course, the E6550 with the larger cache is faster, but the question here is just how much faster? And more importantly, is there a difference between the P965 and P35 here. We also saw with higher FSB the role of a larger cache becomes less important - is that also true here?

Obviously, the difference can be clearly seen on the maximum frame rates. Its obvious the E6550 does run slightly faster with the P35 chipset. Interestingly enough, on lower clocks (6 x 400 MHz), we can also see differences on average and minimum frame rates. This confirmed our assessment that given a high enough FSB and bandwidth (and also processor clock apparently), you don't actually need a  larger cache.

P35 to P965 - DDR2	E6300			E6550
	7x400	6x400	7x333	7x400	6x400	7x333
	Difference	Difference	Difference	Difference	Difference	Difference
Min	2.08%	0.00%	2.08%	2.08%	0.00%	0.00%
Avg	1.88%	1.73%	1.37%	1.09%	2.45%	4.83%
Max	1.05%	1.82%	1.85%	0.68%	0.34%	1.76%

F.E.A.R is a graphically rich game, which make it more of a graphics benchmark than a system benchmark. For those who don't talk tech, it simply means we're not seeing any differences here, at least nothing noticeable when you're actually playing. The differences likely occur during areas where the processor is 'holding up' the graphics card from reaching its maximum performance - maximum frame rates. We do not see a difference between P965 and P35 here, but then again we also didn't see much difference between the processors (at the same FSB and multiplier) either. The E6300 and E6550 practically offer the same level of performance. Any differences we're seeing is very likely just normal variations between runs.

P965	E6300		E6550
	7x400	6x400	7x400	6x400
	Compared to 7x333	Compared to 7x333	Compared to 7x333	Compared to 7x333
Min	0.00%	0.00%	-2.04%	0.00%
Avg	4.23%	0.57%	6.45%	2.90%
Max	5.90%	1.48%	4.23%	3.17%
P35	E6300		E6550
	7x400	6x400	7x400	6x400
	Compared to 7x333	Compared to 7x333	Compared to 7x333	Compared to 7x333
Min	0.00%	-2.04%	0.00%	0.00%
Avg	4.76%	0.93%	2.65%	0.56%
Max	5.07%	1.45%	3.11%	1.73%
P35 DDR3	E6300		E6550
	7x400	6x400	7x400	6x400
	Compared to 7x333	Compared to 7x333	Compared to 7x333	Compared to 7x333
Min	-2.04%	0.00%	0.00%	0.00%
Avg	3.81%	2.10%	3.02%	0.36%
Max	4.36%	0.73%	0.68%	0.00%

P965	E6550 to E6300
	7x400	6x400
	Difference	Difference
Min	0.00%	2.08%
Avg	3.01%	3.20%
Max	3.14%	6.55%
P35	E6550 to E6300
	7x400	6x400
	Difference	Difference
Min	0.00%	2.08%
Avg	2.20%	3.92%
Max	2.76%	5.00%
P35 DDR3	E6550 to E6300
	7x400	6x400
	Difference	Difference
Min	2.08%	0.00%
Avg	3.64%	2.66%
Max	3.14%	6.14%

This table confirms what we saw before. F.E.A.R is more of a graphical benchmark than a system benchmark (with this setup). Although it doesn't add anything to this article, its actually pretty nice to know this fact. Even with very 'light' settings - no AA and AF - F.E.A.R is more graphically bound than system bound. Lets move on now.

P35 to P965 - DDR2	E6300			E6550
	7x400	6x400	7x333	7x400	6x400	7x333
	Difference	Difference	Difference	Difference	Difference	Difference
Min	0.00%	2.06%	0.00%	0.00%	1.02%	1.02%
Avg	1.65%	2.00%	2.60%	0.59%	3.25%	2.37%
Max	14.89%	16.89%	11.76%	16.33%	7.54%	10.42%

Being lighter than F.E.A.R, we certainly hoped this game and scene scale better with system (FSB, multiplier, memory timings) changes. Well, it does, but again like Homeworld 2, its limited to parts where the system is holding up the graphics card. The differences we're seeing between the P35 and P965 are significant in percent, but not significant in gameplay. It does seem to appear that both processors benefit at about the same level.

P965	E6300		E6550
	7x400	6x400	7x400	6x400
	Compared to 7x333	Compared to 7x333	Compared to 7x333	Compared to 7x333
Min	0.00%	-1.02%	1.02%	0.00%
Avg	3.77%	1.26%	4.75%	1.59%
Max	6.33%	-0.90%	2.08%	5.00%
P35	E6300		E6550
	7x400	6x400	7x400	6x400
	Compared to 7x333	Compared to 7x333	Compared to 7x333	Compared to 7x333
Min	0.00%	1.02%	0.00%	0.00%
Avg	2.81%	0.66%	2.93%	2.47%
Max	9.31%	3.64%	7.55%	2.26%
P35 DDR3	E6300		E6550
	7x400	6x400	7x400	6x400
	Compared to 7x333	Compared to 7x333	Compared to 7x333	Compared to 7x333
Min	1.02%	-1.02%	0.00%	0.00%
Avg	2.38%	0.13%	0.40%	0.40%
Max	1.63%	3.25%	-4.58%	4.58%

P965	E6550 to E6300
	7x400	6x400
	Difference	Difference
Min	1.02%	1.03%
Avg	2.87%	2.24%
Max	4.26%	15.07%
P35	E6550 to E6300
	7x400	6x400
	Difference	Difference
Min	1.02%	0.00%
Avg	1.80%	3.50%
Max	5.56%	5.86%
P35 DDR3	E6550 to E6300
	7x400	6x400
	Difference	Difference
Min	0.00%	2.06%
Avg	0.80%	3.06%
Max	0.00%	7.87%

The results are seen from a different perspective in the table above. These tables look more like the ones we saw with F.E.A.R earlier. This led us to conclude that its pretty likely the difference between P35 and P965 we saw earlier with this game is 'superficial'. Remember, we're using an AMD / ATI Radeon X1950 Pro and this is an OpenGL game. Although ATI have revamped their OpenGL driver for Vista, its still generally not as 'lightweight' as their competitors. What we saw in this benchmark seem to indicate the problem are related to memory bandwidth / latency. Great to know and the extra frame are also nice, but you won't notice them at all in game.

P35 to P965 - DDR2	E6300			E6550
	7x400	6x400	7x333	7x400	6x400	7x333
	Difference	Difference	Difference	Difference	Difference	Difference
Min	10.94%	10.94%	2.94%	-5.80%	9.09%	2.90%
Avg	0.77%	3.30%	1.48%	2.63%	-0.59%	0.35%
Max	0.00%	3.81%	0.92%	3.54%	-1.79%	-1.79%

Serious Sam II is still a little bit twitchy - frame rate wise - with this setup. You'll no doubt notice this if you looked at minimum and maximum frame rates in the graph above. Thankfully, overall the average frame rates are still what we expect - they scale with FSB / processor clock increases. Not by much, but they still do. How do we know this? Well, if you look at the results with P965 and P35, you might think the differences are variations between runs. But if you look at the DDR3 results, you'll see there an unmistakable indication that this benchmark / game still benefit from extra bandwidth. It's more pronounced on the slightly more limited E6300. The increase is high enough that at 7 x 400 on a P35 motherboard with DD3 that the E6300 is able to match the E6550 performance.

P965	E6300		E6550
	7x400	6x400	7x400	6x400
	Compared to 7x333	Compared to 7x333	Compared to 7x333	Compared to 7x333
Min	-5.88%	-5.88%	0.00%	-4.35%
Avg	1.01%	-1.95%	0.17%	-0.04%
Max	-0.92%	-3.67%	0.89%	0.00%
P35	E6300		E6550
	7x400	6x400	7x400	6x400
	Compared to 7x333	Compared to 7x333	Compared to 7x333	Compared to 7x333
Min	1.43%	1.43%	-8.45%	1.41%
Avg	0.31%	-0.19%	2.44%	-0.98%
Max	-1.82%	-0.91%	6.36%	0.00%
P35 DDR3	E6300		E6550
	7x400	6x400	7x400	6x400
	Compared to 7x333	Compared to 7x333	Compared to 7x333	Compared to 7x333
Min	1.54%	-3.08%	-1.45%	0.00%
Avg	7.38%	2.82%	3.21%	-0.35%
Max	8.41%	1.87%	6.42%	-0.92%

P965	E6550 to E6300
	7x400	6x400
	Difference	Difference
Min	7.81%	3.13%
Avg	2.49%	5.35%
Max	4.63%	6.67%
P35	E6550 to E6300
	7x400	6x400
	Difference	Difference
Min	-8.45%	1.41%
Avg	4.37%	1.39%
Max	8.33%	0.92%
P35 DDR3	E6550 to E6300
	7x400	6x400
	Difference	Difference
Min	3.03%	9.52%
Avg	-1.62%	-0.81%
Max	0.00%	-0.92%

Nothing new to see here. The differences are still so small that for practical purposes, this benchmark is still more graphically bound than system bound like F.E.A.R and Quake 4.

P35 to P965 - DDR2	E6300			E6550
	7x400	6x400	7x333	7x400	6x400	7x333
	Difference	Difference	Difference	Difference	Difference	Difference
Min	13.33%	7.14%	4.88%	8.16%	6.67%	9.30%
Avg	10.54%	5.48%	13.32%	12.81%	6.68%	8.53%
Max	8.82%	5.06%	13.29%	13.66%	6.98%	7.78%